WO2017155421A1

WO2017155421A1 - Method and system for forecasting the power output of a group of photovoltaic power plants and managing the integration of said power output into a power grid

Info

Publication number: WO2017155421A1
Application number: PCT/PT2017/050006
Authority: WO
Inventors: Rui José NÓBREGA PESTANA; Nuno Miguel PINHO DA SILVA; André Gabriel CASACA DE ROCHA VAZ; Zhibao CHEN; Luís Miguel MARREIROS ROSA
Original assignee: Centro De Investigação Em Energia Ren - State Grid, S.A
Priority date: 2016-03-07
Filing date: 2017-03-07
Publication date: 2017-09-14
Also published as: PT109213B; PT109213A

Abstract

The present invention is related to a method and system for the management of a power grid, and more specifically a method and system for forecasting the power output of a group of photovoltaic power plants and integrating said power output into a power grid, said system and method being capable of aggregating the forecasts of the photovoltaic power output of various photovoltaic power plants (i.e., a group of photovoltaic power plants) so that the predicted combined photovoltaic power output for the group (normally related to a predetermined geographic area) is identical or very approximate to the effective combined photovoltaic power output of said group. The invention reduces the cost of planning and deploying photovoltaic power plants and improves the efficiency of managing those photovoltaic power plants through more accurate forecasts. As a result of the invention, the attractiveness of renewable solar technology as a power source will increase.

Description

DESCRIPTION

METHOD AND SYSTEM FOR FORECASTING THE POWER OUTPUT OF A GROUP OF PHOTOVOLTAIC POWER PLANTS AND MANAGING THE INTEGRATION OF SAID POWER OUTPUT INTO A POWER GRID

Technical Field

The present invention is related to a method and system for the management of a power grid, and more specifically a method and system for forecasting the power output of a group of photovoltaic power plants and integrating said power output into a power grid. Background Art

The sharp growth rate of photovoltaic power plants installed capacity in some countries has increased the need to have much more efficient and adaptable power grid management systems, a need that is particularly felt by the Transmission System Operator (hereinafter "TSO") , since the TSO typically has to have accurate predictions of this type of renewable energy due to their dispatch requirements.

In countries where this type of power production exists (i.e. photovoltaic power), the TSO has to cope daily with the fluctuating input from this type of renewable energy source .

Currently, many TSOs try to cope with these fluctuations and lack of accurate predictions by using only electrical variables (e.g., voltage, current, active power and frequency) and, by and large, ignore the advantages of factoring in the weather forecast. Indeed, in order to manage the integration of photovoltaic power output into the power grid, the traditional approach of TSOs has been to simply curtail the integration of photovoltaic power output when there is more power output than had been expected. This is known as the corrective control method or mode and it has the consequence of weakening the system since the TSO needs power generation to back-up the load. The problem is compounded when the TSO has to manage the integration of photovoltaic power output from several different photovoltaic power plants.

In short, up to now, TSOs have been focussed primarily on the safety of the system in terms of power balance (frequency equilibrium) and of grid limits (avoiding overloads) .

Those TSOs that do factor in weather forecasts in the management of the power grid typically do so only to obtain a general notion of the possible impacts on power production within a certain medium to long time frame. Indeed, TSOs are currently unable to adequately harness the available information on the weather and use it for the purpose of the efficient management of the power grid they are responsible for.

There are a few examples in the prior art of methods and systems for forecasting photovoltaic power production. However, these methods and systems have, by and large, been unsatisfactory.

Indeed, in general terms, the systems and the methods of the prior art have been unable to provide very accurate and updated forecasts. Furthermore, the systems and methods of the prior art focus on systems and methods for forecasting the power output of a single photovoltaic power plant and have not taken on an integrated global approach, in other words, the forecast of the power output of a group of photovoltaic power plants and the integration of the combined power output of that group of photovoltaic power plants into a power grid. The term "group" means, within the context of the present invention, two or more photovoltaic power plants.

Objectives of the invention

In light of the shortcomings of the prior art, the present invention seeks to provide a method and a system to produce forecasts in which the predicted photovoltaic power output of a photovoltaic power plant turns out to be identical or very approximate to the effective photovoltaic power output

It is also an objective of this invention to have a system and method that is capable of aggregating the forecasts of the photovoltaic power output of various photovoltaic power plants (i.e., a group of photovoltaic power plants) so that the predicted combined photovoltaic power output for the group (normally related to a predetermined geographic area) is identical or very approximate to the effective combined photovoltaic power output of said group (which is normally related to a predetermined geographic area) .

It is also an objective of this invention to have a system and method of forecasting the photovoltaic power output of a photovoltaic power plant or of a group of photovoltaic power plants in which the forecasts become increasingly accurate over time.

Disclosure of the Invention The method of the invention consists of a computer implemented method of forecasting power output of a group of photovoltaic power plants integrated in a power grid, each photovoltaic power plant comprising photovoltaic panels, said computer implemented method characterized by comprising the following steps:

forecasting a power output result for each individual photovoltaic power plant of said group by :

- inputting into a data processor a mesoscale meteorological forecast covering a pre-defined extended period of time in the future and an area corresponding to the photovoltaic power plant's geographic location,

- said mesoscale meteorological forecast comprising data on solar irradiance on a horizontal plane with cloudiness, ambient temperature, wind speed, wind direction and air density;

- inputting into said data processor real-time data, provided by a SCADA system, on solar radiance at the photovoltaic power plant's geographic location and applying, by means of the data processor, a solar radiation persistence algorithm to said real-time data in order to determine a solar radiation forecast for the near-future;

- inputting into the data processor real-time data on ambient temperature provided by a pyrometer placed proximate to said photovoltaic power plant's geographic location and applying, by means of the data processor, an ambient temperature persistence algorithm to said real-time data in order to determine an ambient temperature forecast for the near- future ;

inputting into the data processor real-time imagery provided by a sky-camera located proximate to the photovoltaic power plant's geographic location and near-time imagery, provided by a satellite, of the area corresponding to the photovoltaic power plant's geographic location and applying, by means of the data processor, a cloud coverage index algorithm to said imagery in order to determine a cloudiness forecast for the near- future ;

calculating, by means of the data processor, a solar radiation to real power output forecast (in accordance with the known formula of converting solar radiation into real power output) for said photovoltaic power plant for the extended period of time in the future which additionally factors in :

- the forecasts of the previous steps;

- data on tilt angle and orientation of the photovoltaic panels; data on surface area of the photovoltaic panels;

- an efficiency decay factor related to the photovoltaic panels;

- a loss of efficiency factor of the photovoltaic panels related to said ambient temperature forecast for the near-future;

- a loss of efficiency factor of the power grid in accordance with pre-determined grid loss indexes; - and a loss of efficiency factor related to time of day shading; adjusting the solar radiation to real power output forecast of the previous step by factoring in data on predetermined legal and technical limitations associated with the photovoltaic power plant;

further adjusting the solar radiation to real power output forecast determined in accordance with the previous step by factoring in data on predetermined maintenance limitations related to the photovoltaic power plant and the power grid;

validating the solar radiation to real power output forecast of the previous step by comparing it, by means of the data processor, with a near-future SCADA forecast, said near- future SCADA forecast being determined by applying a power persistence algorithm to real-time data on active power of the photovoltaic power plant as provided by the SCADA system;

further adjusting the solar radiation to real power output forecast of the previous step by factoring in real-time data on wind speed and wind direction at the photovoltaic power plant's geographic location in accordance with a time and spatial algorithm;

summing up the power output forecasts of each individual photovoltaic power plant integrated in the group, as determined in accordance with the previous steps, and applying a simultaneous factor algorithm to the total result in order to obtain a combined power output forecast;

- merging the combined power output forecast of the previous step with one or more power output predictions provided by one or more external information providers using a merging method in order to obtain a combined merged power output forecast;

- and determining a cumulative photovoltaic power output forecast for the power grid by applying an upscaling algorithm to the combined merged power output forecast and , by means of the data processor, using said cumulative photovoltaic power output forecast to model power injection at the power grid's nodal level .

In a preferred embodiment of the method of the invention, the mesoscale meteorological forecast is based on a Numerical Weather Prediction model and the extended period of time in the future is between 15 minutes and 168 hours and the near-future is between 15 minutes and 10 hours.

Also in a preferred embodiment of the method of the invention the step of applying, by means of the data processor, a solar radiation persistence algorithm to realtime data provided by the SCADA system in order to determine a solar radiation forecast for the near-future is done in accordance with the formula:

ftw_E = ttkPt + (i - «k)p.

where is a weight coefficient of persistence value that will vary with the change of being a combination forecast result of moment at time and P_* being a prediction result of the mesoscale meteorological forecast, the value of the coefficient being:

a_¾ =^■ exp{—k/ε)

where ^ε is a constant index which is determined according to weather conditions and the mesoscale meteorological forecast .

Also in a preferred embodiment of the method of the invention, the step of applying, by means of the data processor, an ambient temperature persistence algorithm to the real-time data provided by the pyrometer in order to determine an ambient temperature forecast for the near- future is done in accordance with the formula:

where *¾ is a weight coefficient of persistence value that will vary with the change of Pt÷kt being a combination forecast result of t+k moment at time and fit being a prediction result of the mesoscale meteorological forecast, the value of the coefficient ^tTfe being:

«*¾. = expC—k ε)

where ^ε is a constant index which is determined according to weather conditions and the mesoscale meteorological forecast . In a preferred embodiment of the method of the invention, the step of the sky-camera providing the data processor with imagery is done in regular time intervals within the range of every one minute to every thirty minutes and the step of the satellite providing the data processor with imagery is done in regular time intervals within the range of every thirty minutes to every ninety minutes. In a preferred embodiment of the method of the invention, the sky-camera has a fisheye lens and the step of applying, by means of the data processor, a cloud coverage index algorithm to the sky-camera imagery, further includes a step of correcting image distortion resulting from the sky- camera's fisheye lens by rectifying radius distortion in each raw image obtained by the sky-camera so as to have an initial distorted image and a calibrated image, the calibration being done according to a formula where:

i As ¾i^* and i^u£> ^v& are image coordinates of point in the calibrated image and of point ^J^ in the distorted image provided by the sky camera and which the polar coordinates are described by

where

In this step, the calibration coefficients and *½ _are computed by curve fitting, using manual matching between an orthographic image from a calibration grid and a sky camera image from the same grid.

In a preferred embodiment of the method of the invention, the step of inputting into the data processor the real-time imagery provided by a sky-camera located proximate to the photovoltaic power plant's geographic location and the near-time imagery provided by a satellite of the area corresponding to the photovoltaic power plant's geographic location includes the additional step of transforming the calibrated image from a colour image to a grey-scale image. In a preferred embodiment of the method of the invention, the step of transforming the calibrated image from a colour image to a grey-scale image is done using colour space R, G, B with the following greyscale transform formula:

B{ .v)—Ri.u,v} £{u,v$— £? , t?)

I(U.V) =— ; ς ; -f ; - ; -,

'^' B{ _tv) + R{u,v) SiUy-v) + Giu,,v)

where W is pixel image coordinates.

In a preferred embodiment of the method of the invention, the step of transforming the calibrated image from a colour image to a grey-scale image further includes the step of calculating, by means of the data processor, an energy result for the greyscale image which is done by summing up power spectrum in accordance with the Fast Fourier Transform algorithm.

In a preferred embodiment of the method of the invention, in the step of inputting into the data processor near-time imagery provided by a satellite, said imagery is infrared imagery .

In a preferred embodiment of the method of the invention, the cloudiness index is determined, by means of the data processor, from the satellite imagery every 15 minutes in accordance with the formula:

for a given time ^s the cloudiness by the Earth Mover' s

Distance between two probability density functions is lEMD(t) = EMD(P^{R, G_SB}_S PJ{R, 6_fU}} _where P* _{IS THE} probability density function of the most recently acquired infrared satellite image with respect to time * and ^rl* is the time tag closer to time In a preferred embodiment of the method of the invention, the mapping between cloudiness indexes and solar irradiance relies on irradiance measurements from a pyranometer located proximate to the sky-camera and on computation of clear sky irradiance.

In a preferred embodiment of the method of the invention, the clear sky irradiance is given by the formula:

I... = 5^■ ™^¾ - sin .

where ⁵— 1367Wm j__s ^ e solar constant, ⁿ is the ^{¾1 '} day of the current year, ^a is the solar altitude and the air mass ^mt∑ is given by

Considering a forecast horizon ^Ά , the forecasted solar irradiance at time ^{f ft} is given by the decrease of the clear sky irradiance at time - ^fi- due to the impact of ^{i± i ij} at time

corresponds to a 15 minutes lag with respect to ^f ) and to the impact of the

1E D f_or time ^f and time ^f— ¾s(

corresponds to a 1 hour lag with respect to {V*^s-^■ ) , it being defined by I_sc(t + k) = i_c._s(t + k} · l(t_f h)^, _{f wh}ere + *) is the theoretic value of the clear sky irradiance at time ^*~ ~ ·^ί?- and ^?i is

t -L h

the estimated loss of irradiance for time " ^ ^>s , whereby said loss is computed by where

are the parameters for horizon ^Sl, estimated using log-linear regression. Let E be the BRBG image energy, computed by summing up the power spectrum obtained with the Fast Fourier Transform algorithm, and let L be the mean luminance of the calibrated colour image. The cloudiness index computed from the sky camera imagery at time t is given by

In a preferred embodiment of the method of the invention, the mapping encodes cloud motion by considering cloud indexes at different times.

In a preferred embodiment of the method of the invention, the data on tilt angle and orientation of the photovoltaic panels is determined using a solar tracking algorithm.

In a preferred embodiment of the invention, the data on predetermined legal and technical limitations associated with the photovoltaic power plant are pre-loaded parameters stored in memory means that are in communication with the data processor. One of the pre-loaded parameters can be, by way of example, the inverter limit and another can be, by way of example, the transformer capacity.

In a preferred embodiment of the method of the invention, the data on loss of efficiency of the power grid in accordance with pre-determined grid loss indexes is data related to DC wiring, DC/AC wiring, AC wiring and transformers .

In a preferred embodiment of the method of the invention, the simultaneous factor algorithm is between 95% and 99%. In a preferred embodiment of the method of the invention, the merging method is a deterministic approach based on the skill of the external forecast provider.

In another embodiment of the method of the invention, the merging method is a dynamic approach based on a moving time window that minimizes the error of the forecast values and the real values.

In a preferred embodiment of the method of the invention, the merging method assigns to the external information providers a weight, said weight resulting from minimizing the least squares error between a variable forecast value and an observed value, said forecast value being defined by a linear combination of the forecast values from the external information providers, where the weights are the unknowns in the aforementioned minimization. The weights are updated every 6 hours .

In a preferred embodiment of the method of the invention, the upscaling algorithm compares the sum of the forecasted power outputs for the group of photovoltaic power plants with the real power measured by metering devices. This upscaling algorithm corresponds to an upscaling curve which is a polynomial of sixth order.

In a preferred embodiment of the method of the invention, the data is communicated from the external services and hardware on the ground to the data processor by means of a file transfer protocol server.

The invention also consists of a system for implementing the method of the invention, said system comprising, in its simplest embodiment: - a data processor capable of receiving data from various inputs and sources and processing said data in accordance with algorithms;

- a memory means capable of storing data and algorithms, said memory means being in communication with said data processor;

- at least one SCADA system capable of monitoring and controlling one or more photovoltaic power plants and communicating data to and receiving data from the data processor;

- at least one sky camera fitted with a fisheye lens for each photovoltaic power plant of the group, each sky camera placed proximate to its respective photovoltaic power plant and in communication with the data processor;

- at least one pyranometer for each photovoltaic power plant of the group, each pyranometer placed proximate to its respective photovoltaic power plant and in communication with the SCADA system;

- at least one ambient temperature sensor for each photovoltaic power plant of the group, each ambient temperature sensor placed proximate to its respective photovoltaic power plant and in communication with the SCADA system;

- and a main control center in communication with the SCADA system.

Brief Description of Drawings

To assist in the comprehension of the invention, the attached drawings are provided. It should be noted that these drawings are representations of one or more specific embodiments of the invention and should not to be interpreted as having any limitative effect on the scope of protection . Accordingly :

Figure 1 is flow chart of preferred embodiment of the method of the invention;

Figure 2 is a raw image from a sky camera;

Figure 3 is the same image of Figure 2 in color duly calibrated by the process identified in Step 4 of the method shown in Figure 1;

Figure 4 is the same image shown in Figure 3 but in BRBG (as defined in Step 4 of the method of the invention) ;

Figure 5 is an example of the type of satellite images uploaded into the data processor of the system of the invention in Step 1 of the preferred embodiment of the method of the invention shown in Figure 1;

Figure 6 is a flow chart showing types of sun trackers (source: [Mousazadeh, H.;Keyhani, A.; Javadi, A.; Mobli, H. ;Abrinia, K. ; Sharifi, A., "A review of principle and sun-tracking methods for maximizing solar systems output, " Renewable and sustainable Energy Reviews, vol. 13, pp. 1800-1818, 2009.]);

Figure 7 is a flow data diagram that illustrates the interrelationship between the methodology shown in Figure 1 and its operation applicability in the energy transmission management system;

Best Mode for Carrying Out the Invention

The preferred embodiment of the invention will now be described in further detail below.

The present invention can best be described as a new methodology for the forecasting photovoltaic power output of a group of photovoltaic power plants. For the implementation of said methodology, a system is needed to collect and process data and, subsequently, provide commands to automatically manage the power grid.

The expression "power grid", as would be generally understood by a skilled person in the art, means the facilities and infrastructure that exist for the purpose of transmitting and distributing electricity to the users. In the description of this preferred embodiment of the invention, only the facilities and infrastructure connected directly to the transmission network (very high voltage) and the distribution network (high voltage and medial voltage) are included in the concept of power grid. However, as would be understood by a person skilled in the art, a power grid may also include infrastructures at low voltage, such as photovoltaic infrastructures installed on rooftops or for domestic use that currently tend to be behind the meter with no measurements. It should be noted that the method and system of the present invention may be adapted to also include low voltage infrastructures.

The system of the invention

The system of the invention consists of the interconnection of various hardware. The system has, at its core, a data processor capable of receiving data from various inputs and sources and processing said data in accordance with certain algorithms. In a preferred embodiment of the invention, the data processor comprises one or more computers connected to one or more file transfer protocol servers. The data processor is connected to memory means capable of storing data (e.g., algorithms, meteorological data, real^¬ time grid data, forecasted power, etc.), including in the form of software. The data processor is also connected to data input devices capable of obtaining data from the surrounding environment. Each photovoltaic power plant will have associated to it specific data input devices, including at least one sky camera, at least one pyranometer and at least one ambient temperature sensor (also known as pyrometer) .

Each data input device associated to each photovoltaic power plant is installed in the vicinity of (or proximate to) the respective photovoltaic power plant so as to obtain environment readings specific to the location of said photovoltaic power plant. In a particularly preferred embodiment of the invention, each data input device will be installed in an appropriate location and within a radius of less than 5 km from the center of the photovoltaic power plant. The term "center" is used in a non-precise manner, meaning the general area considered to be at the core of the group of photovoltaic panels. In a preferred embodiment of the invention, it is possible to have more than one of the aforementioned different input devices place in the vicinity of (or proximate to) each photovoltaic power plant, each device being connected to the data processor.

The term "sky camera", as generally understood by a person skilled in the art, consists of an apparatus capable of capturing images of the sky in digital format. In a preferred embodiment of the invention, the sky camera is specifically adapted and designed for capturing images of the sky, namely by having a fisheye lens.

As will be described below, each sky camera serves the purpose of collecting real-time data that is used by the data processor for the estimation of a cloudiness index. As also described below, the real-time estimation of the cloudiness index is used to adjust the prediction of the solar irradiance provided by the mesoscale model.

The term "pyranometer", as generally understood by a skilled expert, is a type of actinometer used for measuring solar irradiance on a planar surface and it is designed to measure the solar radiation flux density (W/m2) from the hemisphere above within a wavelength range 0.3 μπι to 3 μπι. The expression "ambient temperature sensor" is, as generally understood by a skilled expert, a digital weather thermometer (also known as a pyrometer) . In the present invention, the ambient temperature sensor is capable of converting temperature readings into digital data which is communicated to the data processor for processing.

The data processor is also connected to a "SCADA system". The SCADA (Supervisory Control and Data Acquisition) system is a system that is capable of providing real-time datasets (i.e., real-time monitoring) of the actual power output of the photovoltaic power plants that are connected to the power grid. More specifically, the SCADA obtains and communicates to the data processor of the invention data on active power, reactive power, voltage, wind speed, solar irradiance and ambient temperature. The SCADA system also interacts with and controls and devices. Indeed, data captured by temperature sensors, pyranometers and a photovoltaic power plant's real-time power output is collected by the SCADA system. This real-time data is fundamental for the operation of an Energy Management System (EMS), which comprises supervisory control and data acquisition, functions for dispatch and loop digital control of generators, dynamic schedules, interchange scheduling, weather adaptive demand forecasting and the

State Estimator (SE) that allows the analysis and optimization of the transmission network in a reliable and secure manner.

The data of the SCADA system is communicated from a Remote Terminal Unit ("RTU") to a main control center using the IEC 60870-5-101/104 protocol. The main control center consists of servers used for the control of the system.

The SCADA also receives real-time information from other control centers from other renewable energy generation plants using the ICCP (Inter-Control Center Communications Protocol) . Other renewable energy generation plants may be wind farms, hydroelectric power plants or other currently known technologies.

The SCADA system regularly communicates the necessary predefined data to the data processor. In a particularly preferred embodiment of the invention, the SCADA system communicates the necessary predefined data to the data processor once every minute.

The SCADA system is used by the TSO to remotely control the power grid.

The data processor updates power output limit to avoid the overloads in the power grid. The SCADA communicates to each photovoltaic power plant the updated power output limit and the each photovoltaic power plant processor uses this new setpoint to limit the power output of said photovoltaic power plant. The data processor is also connected to external weather data repositories, such as databases or other information sources that contain satellite images. As will be described in further detail below, the satellite images provided by the weather data repositories are processed by the data processor of the invention so as to adjust the cloudiness index for the long-term, provided by the mesoscale forecast The transfer of data between the aforementioned devices and means is accomplished by either wired or wireless communication means capable of communicating in real-time (as the expression is commonly understood by a skilled person of the art) .

It should be noted that none of the systems of the prior art for the forecasting of solar power output have all the features described above, although some combine some of these features. The system of the invention has these features and is configured in this way in order to enable the implementation of the novel method of photovoltaic power forecasting and control, described below.

The methodology of the invention

In addition to the system, the present invention consists of a methodology for the forecasting of the photovoltaic power output of a group of photovoltaic power plants and the integration of said power output into a power grid.

In summarized terms, the methodology of the invention comprises the processing of real-time, near-real-time and long term data (information) received from several data sources, including peripheral data input devices placed in the vicinity of (or proximate to) each photovoltaic power plant, and then, based on the results of the processing, the data processor will automatically instruct the SCADA system to make the necessary adjustments to the management of the power grid.

The method of the invention also aggregates the forecasted power output of each photovoltaic power plant connected to the power grid in order to obtain a forecast of the photovoltaic power output of a group of photovoltaic power plants (usually corresponding to a certain predetermined geographic territory) .

In more detailed terms, the methodology of the preferred embodiment the invention comprises the following steps shown in Figure 1, namely: Step 1:

The first step of the method consists of the receipt of a mesoscale meteorological forecast for a certain pre- established period of time by the data processor using the file transfer protocol server to receive the input mesoscale data.

The mesoscale meteorological forecast is provided by an external service provider. Typically, the mesoscale meteorological forecasts are provided at pre-established times, such as, every six hours. The mesoscale meteorological forecasts are sent by the external service providers to the data processor, by file transfer protocol using any public communication means, like the internet. The time horizon of the mesoscale meteorological forecast can, naturally, vary. In a preferred embodiment of the invention, the mesoscale meteorological forecast is for the next seven days (i.e. a week) . It should be noted that the mesoscale meteorological forecasts may be based on different Numerical Weather Prediction ("NWP") models such as, by way of example, MM5 (Penn State University/National Center for Atmospheric Research) or WRF (National Center for Atmospheric Research/National Oceanic and Atmospheric

Administration/Air Force Weather Agency) .

These NWP models use physical models of the atmosphere and oceans to predict the weather based on the current weather conditions. A mesoscale meteorological forecast will include at least the solar irradiance in a horizontal plane with cloudiness, ambient temperature, wind speed and direction and also the air density for each site previously selected by the TSO.

The sites selected by the TSO will be the general location of each photovoltaic power plant connected and controlled by the system.

In a particularly preferred embodiment of the invention, the mesoscale meteorological forecast received from the external service provider by the data processor will be based on a NWP model and will be in a computer readable format. In a particularly preferred embodiment of the invention, the mesoscale meteorological forecast files are sent to the data processor in ASCII format using the file transfer protocol server. The NWP models are initialized on the basis of meteorological observations and allow calculations on a finer grid covering selected regions. The global horizontal solar irradiance and ambient temperature forecasts are the inputs of the system, which subsequently are subjected to conversion to power and estimation of the loss of power due to ambient temperature, respectively.

Step 2 :

The second step of the method consists of inputting into the data processor a real-time solar radiation forecast for a certain pre-established period of time. This pre- established period of time can, in a particularly preferred embodiment of the invention, be the next 10 hours. This data, which includes data on solar irradiance, is obtained by the SCADA system of each photovoltaic power plant connected to the system or from the substations of the TSO. This real-time data communicated to the data processor by the SCADA is processed by the data processor in accordance with a solar radiation persistence algorithm.

The expression "persistence algorithm" in this invention means a mathematical formula for a rule that states that future power generation will be the same as the last measured power. The persistence rule is a proven, common and simple approach which is known and applied in the state of the art. The persistence algorithm used in this step is the merging of the mesoscale meteorological forecast from NWP models with the persistence. The merging model is:

fttvt = ¾Pt + <1 - «k)Pt

where *¾· is the weight coefficient of persistence value and it will vary with the change of Pt*kft is the combination forecast result of moment at time and Pt is the prediction result of NWP model. The value of the coefficient is:

s¾.— exp(— k ε)

where ^ε is a constant index which is determined according to weather situation and NWP model. Step 3:

The third step of the method consists of inputting into the data processor a real-time ambient temperature forecast for a certain pre-established period of time. In a particularly preferred embodiment of the invention, this pre-established period of time can be the next 10 hours. This real-time data, which includes data on ambient irradiance, is obtained by the SCADA system of each photovoltaic power plant or from the substations of the TSO. This real-time ambient temperature data is communicated by the SCADA to the data processor and is processed by the data processor of the invention in accordance with an ambient temperature persistence algorithm.

The persistence algorithm applied on the real-time ambient temperature forecast is the same as that described above in

Step 2, with the exception that the constant ^ε is different since the dynamic of ambient temperature is different from solar radiation.

Step :

The next step consists of inputting the real-time sky camera imagery and near-time satellite imagery into the data processor.

In a particularly preferred embodiment of the invention, the sky-camera collects and communicates images of the sky above the respective photovoltaic power plant every 5 minutes while the satellite images of the territory where the respective photovoltaic power plant is located are communicated to the data processor with 1 hour intervals between images. The imagery from the sky camera and the imagery from the satellite camera have different impact on the accuracy of the forecast as will be discussed below. At least one sky camera is installed in fairly close proximity to each photovoltaic power plant in order to capture images of the sky above said photovoltaic power plant. The system of the invention foresees at least one sky camera per photovoltaic power plant but, naturally, it is possible for the system to have two or more sky cameras covering different areas in the locality where the photovoltaic power plant is located. Given that the sky camera or sky cameras are installed in close proximity to each photovoltaic power plant, the imagery collected from the sky camera or sky cameras is useful for now-casting (up to one hour ahead in time) .

The satellite imagery is provided by external providers (such as, for example, a national weather institute) and is inputted into the data processor by automatic means. The satellite imagery is useful for short-term forecasts (up to six hours ahead in time) .

The sky camera and satellite imagery are processed by the data processor in accordance with a cloud coverage index algorithm. This produces two cloudiness indexes that characterize the impact of the clouds in the solar irradiance above the photovoltaic power plant. The method of determining the cloudiness index from the sky imagery will now be described in more detail:

The sky camera raw images are taken, in a particularly preferred embodiment of the invention, through a fisheye lens. This lens introduces image distortion which is corrected by rectifying the radius distortion in the raw images .

Let ¾ ' and ^iu*> ¾^ be the image coordinates of point ^ in the rectified image and of point ^ in the distorted image from the sky camera, In polar coordinates, these coordinates are described by

RssmB

where

R_A = a^R + s_s¾_f i?i = i ² + ¾^{2 id} 8 = arctaii—

The calibration coefficients ¾^aaSfio are computed by curve fitting, using manual matching between an orthographic image from a calibration grid and a sky camera image from the same grid. The calibration coefficients are applied to each sky camera raw imagery which results in calibrated imagery .

The calibrated imagery is transformed from colour imagery to grey imagery using the colour space R, G, B (red channel value, green channel value and blue channel value) with the following new grayscale transform

Biu_tv)— R{u,\?) S{ti_y v)— G{u_tv) where ^^J is the pixel image coordinates. For the purposes of this invention, the resulting grayscale image is called a BRBG image .

Let ^& be the BRBG image energy, computed by summing up the power spectrum obtained with the Fast Fourier Transform algorithm, and let ^ be the mean luminance of the calibrated colour image. The cloudiness index computed from the sky camera imagery at time is given by

. , E(t

TFMLit) =——.

^" ' Lit)

The more overcast the sky is, the higher the index value will be.

As mentioned above, the satellite imagery provided by the external service provider is also processed in order to determine the respective cloudiness index. This procedure will now be described in more detail:

Satellite imagery of a particular territory, region or location is something that can be obtained with relative ease from services available in the market. Furthermore, the satellite imagery can be obtained with great frequency. As a skilled expert will know, it is possible to obtain satellite images of a particular territory, region or location in computer readable format (e.g., jpeg format) as often, or more often, than one per hour.

The data processor of the invention can be programmed, via an algorithm that is triggered every hour or with a different time frequency, to download the satellite images from the external service provider's website or database. More specifically, this algorithm is capable of reading an Internet webpage, identifying a new infrared satellite image and downloading it.

The satellite imagery provided by the external service provider is made available in various forms. For the method of the invention, infrared satellite imagery is required. As is known in the state of the art, infrared images measure thermal radiation: warmer bodies (e.g. Earth surface) radiate more heat than colder bodies (e.g. clouds) Hence these images encode temperature in the brightness values, thus the use of the expression "brightness temperature images".

Due to the different temperatures of clouds and earth surface, the brightness temperature images encode the sky cloudiness. Therefore, a cloudiness index may be inferred from the pixel values of the brightness of the satellite images.

In a particularly preferred embodiment of the invention, cloudiness index from the infrared satellite images relies on the statistical characterization of the brightness temperature pixel distribution under clear sky condition against which the brightness temperature distribution of the new images is compared.

The Earth Mover's Distance ("EMD") is used for measuring the distance between the statistical characterization of the new image and the clear sky brightness temperature image. The clear sky brightness temperature probability density function is computed from a clear sky image library The clear sky image library is a collection of 100 images for each 15 minutes of the day. This library is dynamically updated according to the error between the theoretic value of the clear sky photovoltaic power production and the measured photovoltaic power production for a region observed by the satellite. For example, it may be the photovoltaic power produced in a country.

In a day, there are 96 time tags of 15 minutes. Let the time tag within a day be denoted by ...,96 . T e clear sky library for time tag ^τ is a set of £=100 pairs {!*«*),*£ {!,_.,£},

where

is the corresponding error between the clear sky theoretical photovoltaic power production and the measured photovoltaic power production for a region observed by the satellite.

The clear sky image library is the collection of the 96 sets of time tagged images ^ . The clear sky library is updated according to the following criterion

slew image . &

The clear sky brightness temperature probability density function accounts for all images in the library for a given time tag * and it is given by

F*iR = _r,G = _SlB = h} = " " ^ ^k~^{L τ} where ^ is the number of images of size in the library for each time tag ⁷, is the pixel count with value ^r in the red channel of image . Similar definitions apply to and regarding the green and the blue channel, respectively .

For one image of size nxm _f the brightness temperature probability density function is given by

. #r - #o - #ib

P_tiR = r_tG = g,B = b}= , where is the pixel count with value ^r in the red channel. Similar definitions apply to ^9 and regarding the green and the blue channel, respectively.

The cloudiness index from the infrared satellite imagery is computed by the data processor every 15 minutes. For a given * the cloudiness by the EMD between the two previous probability density functions, i.e. $ EM Bit). = £>(pJ^f¾ G_f Β},Ρ*_τ{ϋ B})

where * is the probability density function of the most recently acquired infrared satellite image with respect to time and is the time tag closer to time

The mapping between the cloudiness indexes and solar irradiance relies on the irradiance measurements from a pyranometer near the sky camera and on the computation of the clear sky irradiance, which is a theoretical parameter.

The clear sky irradiance is given by

in - l x

/„ = S - ! 1 + 0.033 - 2π \ - ^m& < sina,

V ^ / ^■

where = i367I m^{. *}· j__s the solar constant, ^¾ is the ^?r" day of the current year, is the solar altitude and the air mass ^m« is given by

f?½ = Csm s - ΐ5·(« - 3,88S)^_i-^2S3:3^~3,

Considering a forecast horizon , the forecasted solar irradiance at time ^ + ^f's is given by the decrease of the clear sky irradiance at time t + k due to the impact of at time and time * - ( ^si^ corresponds to a 15 minutes lag with respect to ^ ) and to the impact of the iEMD for time * and time *^— *ι$Φ (

corresponds to a 1 hour lag with respect to *) , i.e. ¾c(t+¾) = i_ffS( + ¾ ·&¾).

where *1* +AJ is the theoretic value of the clear sky irradiance at time and is the estimated loss of irradiance for time This loss is computed by where *AA*!†=s are the parameters for horizon , estimated using log-linear regression. The mapping encodes the cloud motion by considering cloud indexes at different times, which would otherwise be needed to be explicitly computed from the image pixel displacement, a task which is prone to errors due to the suboptimal nature of the optical flow computation .

Step 5 :

The next step in the method of the invention is determining the solar radiation to power. Estimating the real power output of photovoltaic power plants requires the conversion of global horizontal solar irradiance forecasts according to the specific orientation of the module in the photovoltaic power plant.

Consequently, the tilt angle, orientation and tracking algorithm are inputs of the estimation of the angle of incidence in the photovoltaic modules. To conclude the conversion, the area and efficiency of the photovoltaic cell provided by manufacturers are introduced to adjust the magnitude of the curves.

More specifically, to adjust the photovoltaic power output forecasts it is necessary to estimate the solar irradiance in the inclined plane of each panel of the predetermined photovoltaic power plant. To do this, the data processor will factor in the efficiency of each photovoltaic panel by connecting to the memory means and retrieving the technical specifications of the photovoltaic cells provided by the manufacturer, the efficiency decay factor based on the age of the panel, the area and tilt of the photovoltaic panel and the loss of efficiency due to the ambient temperature forecasted at the location of the photovoltaic panel.

The efficiency decay factor is the loss of efficiency due to the aging of the photovoltaic panels. By way of example, the manufacturer of photovoltaic panels guarantees an efficiency of 90% during the first 12 years or operation and thereafter and until 25 years and efficiency of 80%.

The data processor will also, during this step, adjust the photovoltaic power output forecast by applying coefficients to account for the loss of efficiency due to grid losses (e.g., DC wiring, DC/AC conversion, AC wiring, transformers, etc.) and loss of efficiency due to shading in the early morning and late afternoon.

Step 6:

The next step in the method is to verify the limitations associated to each photovoltaic power plant. The data processor does this by connecting to the memory means of the system that contain pre-loaded parameters of the limitations associated to each photovoltaic power plant. These limitations may be legal or technical (e.g., inverter limit, transformer capacity, etc.) but they are included in the memory means of the system in a manner that allows the data processor to automatically adjust the photovoltaic power output forecast based on the values associated to each photovoltaic power plant.

This step avoids discrepancies in the system' s results and improves the accuracy of the forecasts.

In addition, the system is able to verify Curtailment Selection needs based upon the Technical Verification of Operational Planning ("VTP") and update the power output forecasts if required, in order to provide reliable information to the TSO.

In the event necessary, power output restrictions for a particular photovoltaic power plant, or for a particular line or transformer, may be inserted in a memory means of the system (e.g., a database In this manner, when the data processor connects to th memory means of the system during this step it is possibl to automatically adjust the photovoltaic power output forecast to take into consideration these technical legal limitations.

Step 7:

The next step in the method is to input into the data processor the power grid limitations due to maintenance work on the power grid that can limit the power output of each photovoltaic power plant. If necessary, the manager of each photovoltaic power plant may also share his maintenance plan that limits the power output. By the data processor connecting to the memory means (e.g., a database) of the system during this step, it is possible for the data processor to automatically factor in these technical limitations and adjust the photovoltaic power output forecast .

Step 8:

In the next step of the method of the invention, the adjusted photovoltaic power output forecast resulting from the above steps is validated and tested by the data processor with real-time data provided by a SCADA of power output from each photovoltaic power plant. The validation consists of verifying that the values provided by the SCADA are not superior to those of the photovoltaic power plant's limitations. The SCADA in question is the same as the one mentioned in the steps above.

Following this (i.e. after the testing and validation with the real time data from the SCADA system) , a power persistence algorithm is applied to the forecast, using the same technique describe in Step 2, with the exception that the constant ^ε is different since the dynamic of power output is different.

Step 9:

The next step in the method of the invention is the process of inputting real-time wind speed and wind direction into the adjustment of the photovoltaic power output of the photovoltaic power plant. This is done by inputting a time and spatial algorithm.

The wind speed and direction forecasts are used to anticipate cloud motions and thus adjust the overall photovoltaic power output forecast, making it more reliable Indeed, based upon the wind speed and direction forecasts in a certain region, the data processor, by applying the time and spatial algorithm, can evaluate the probable future impact of the clouds on solar irradiance in nearby regions and adjust the final forecast by reducing its phase error .

Time and spatial algorithms are known in the prior art and the method of the invention will function with either of those known algorithms. Step 10:

The next step in sum up all the individual photovoltaic power outputs from each photovoltaic power plant connected to the system in order to arrive at a combined photovoltaic power output forecast for a predetermined geographic region Each photovoltaic power plant can achieve its installed power capacity when the weather conditions are favorable. However, it is has been observed that the photovoltaic power plants of a predetermined geographic region are not all at their installed power capacity at the same time. Evaluations have shown that the maximum simultaneous factor for photovoltaic power plants located in national power grids is approximately 95% to 99%. This step of the method of the invention limits the total photovoltaic power output (normally related to a predetermined geographic region) to the maximum simultaneous factor related to the total installed capacity of the group of photovoltaic power plants (normally related to a predetermined geographic region) .

The step consists in evaluating the maximum instantaneous power output of solar production, based on metering values and the total installed capacity of the group of photovoltaic power plants (normally related to a predefined region) . The maximum simultaneous factor is less than 100%.

Step 11:

The next step of the method of the invention consists of combining the forecasts from different providers of photovoltaic power output forecasts.

The photovoltaic power output forecast of each third party provider is communicated to the data processor by the communication means and the data processor combines all these third party photovoltaic power output forecasts with the photovoltaic power output forecast that results from Steps 1 through 10, determines the discrepancies between them and averages them out.

The merging method can be deterministic based on the skill of each forecast provider or it is possible to use a dynamic method for merging the forecast based on moving time window that minimizes the error of the forecast values and the real values.

The merging method assigns to each provider a weight. The weights result from minimizing the least squares error between a variable forecast value and the observed value. The forecast value is defined by a linear combination of the forecast values from the providers, where the weights are the unknowns in the aforementioned minimization. In a particularly preferred embodiment of the invention, the weights are updated every 6 hours.

The systems' output characterizes the outcome of deterministic approaches, which can then become an input for probabilistic methods. Probabilistic methods allow assessing the confidence interval of the deterministic forecasts .

By using several sources/methods of forecast and merging them into only one, the photovoltaic power output forecast error is reduced.

Step 12:

The final step of the method of the invention consists of determining the cumulative photovoltaic power output for the group of photovoltaic power plants (normally associated to a predetermined geographic area (i.e. region) that is being controlled by the system of the invention) by applying an upscaling algorithm.

The upscaling algorithm compares the sum of the forecasted power for the group of photovoltaic power plants (normally one or more geographic regions) with the real power obtained by the metering devices of each energy producer connected to the group (normally belonging to the geographic region) . The upscaling curve is a polynomial of sixth order.

This determination method reduces computational and data handling effort, providing almost no loss of accuracy because the representative set approximates the basic properties of the total data. This approach solves the problem of the lack of detailed system information usually available .

This step of the method of the system is mandatory for a TSO managing a power grid in many countries.

The final adjusted forecast (normally for a predetermined geographic territory) is then used to model the injection of power at the nodal level in the power grid.

Advantages of the method and system of the invention

The method and system of the invention reduces the cost of planning and deploying photovoltaic power plants as well as improving the efficiency of managing those photovoltaic power plants through more accurate forecasts.

Indeed, by way of example, by developing more accurate forecasts for each photovoltaic power plant and the group of photovoltaic power plants, it is possible to have more accurate modelling of the injection of power at the nodal level in the power grid and, thus, the Technical Verification of Programming can be performed for the next day of an energy market (e.g., the Iberian energy market named MIBEL) . In this way, grid management decisions can be more informed and efficient and lead to better use of energy in the system. Indeed, proper accurate forecasts have the potential of offering TSOs (and others) information related to the short-term solar resource variability, in particular fast ramp rates, which allows for cost-effective integration of large penetration of photovoltaic power output in the power grid and the respective electricity supply system.

Consequently, as a result of the invention, the attractiveness of renewable solar technology as a power source will increase.

Contrary to what is foreseen in the present invention, the methods and systems of the prior art do not enable the integration of different inputs from different data sources so as to adjust the NWP models and, in doing so, increase their accuracy. In no method or system of the prior art is there an integration of the data from such a variety of peripheral input data devices (such at least one pyranometer, at least one temperature sensor and at least one sky camera) installed in the vicinity of (proximate to) each photovoltaic power plant connected to the power grid so as to supplement and adjust the data provided by external service providers (such as satellite images and mesoscale weather forecasts) which is also integrated into the method.

As a skilled person in the art would know, it is possible to make small deviations to the method and system mentioned above and yet not depart from the inventive aspects that characterise the present invention.

Claims

A computer implemented method of forecasting power output of a group of photovoltaic power plants integrated in a power grid, each photovoltaic power plant comprising photovoltaic panels, said computer implemented method characterized by comprising the steps of: forecasting a power output result for each individual photovoltaic power plant of said group by

inputting into a data processor a mesoscale meteorological forecast covering a pre-defined extended period of time in the future and an area corresponding to the photovoltaic power plant's geographic location,

said mesoscale meteorological forecast comprising data on solar irradiance on a horizontal plane with cloudiness, ambient temperature, wind speed, wind direction and air density;

inputting into said data processor real-time data, provided by a SCADA system, on solar radiance at the photovoltaic power plant's geographic location and applying, by means of the data processor, a solar radiation persistence algorithm to said real-time data in order to determine a solar radiation forecast for the near-future;

inputting into the data processor real-time data on ambient temperature provided by a pyrometer placed proximate to said photovoltaic power plant's geographic location and applying, by means of the data processor, an ambient temperature persistence algorithm to said real-time data in order to determine an ambient temperature forecast for the near- future ;

calculating, by means of the data processor, a solar radiation to real power output forecast for said photovoltaic power plant for the extended period of time in the future which factors in

the forecasts of the previous steps;

data on tilt angle and orientation of the photovoltaic panels;

data on surface area of the photovoltaic panels ;

an efficiency decay factor related to the photovoltaic panels;

a loss of efficiency factor of the photovoltaic panels related to said ambient temperature forecast for the near-future;

a loss of efficiency factor of the power grid in accordance with pre-determined grid loss indexes; and a loss of efficiency factor related to time of day shading;

adjusting the solar radiation to real power output forecast of the previous step by factoring in data on predetermined legal and technical limitations associated with the photovoltaic power plant;

validating the solar radiation to real power output forecast of the previous step by comparing it, by means of the data processor, with a near-future SCADA forecast;

said near-future SCADA forecast being determined by applying a power persistence algorithm to real-time data on active power of the photovoltaic power plant as provided by the SCADA system; further adjusting the solar radiation to real power output forecast of the previous step by factoring in real-time data on wind speed and wind direction at the photovoltaic power plant's geographic location in accordance with a time and spatial algorithm;

summing up the power output forecasts of each individual photovoltaic power plant integrated in the group, as determined in accordance with the previous steps, and applying a simultaneous factor algorithm to the total result in order to obtain a combined power output forecast; merging the combined power output forecast of the previous step with one or more power output predictions provided by one or more external information providers using a merging method in order to obtain a combined merged power output forecast;

determining a cumulative photovoltaic power output forecast for the power grid by applying an upscaling algorithm to the combined merged power output forecast and, by means of the data processor, using said cumulative photovoltaic power output forecast to model power injection at the power grid's nodal level.

2. The method of claim 1, characterized in that the mesoscale meteorological forecast is based on a Numerical Weather Prediction model.

3 . The method of claim 1, characterized in that the extended period of time in the future is between 15 minutes and 168 hours.

4. The method of claim 1, characterized in that the near- future is between 15 minutes and 10 hours.

5. The method of claim 1, characterized in that the step of applying, by means of the data processor, a solar radiation persistence algorithm to real-time data provided by the SCADA system in order to determine a solar radiation forecast for the near-future is done in accordance with the formula:

where ^■£% is a weight coefficient of persistence value that will vary with the change of , _{! ¾ f} being a combination forecast result of t + k moment at time £, and $¾ being a prediction result of the mesoscale meteorological forecast, the value of the coefficient *¾ being: t¾ = exp(— k ε)

where ε is a constant index which is determined according to weather conditions and the mesoscale meteorological forecast.

The method of claim 1, characterized in that the step of applying, by means of the data processor, an ambient temperature persistence algorithm to the real^¬ time data provided by the pyrometer in order to determine an ambient temperature forecast for the near-future is done in accordance with the formula:

where «¾ is a weight coefficient of persistence value that will vary with the change of k , p_t~k t being a combination forecast result of t+k moment at time £, and i¾ being a prediction result of the mesoscale meteorological forecast, the value of the coefficient «¾ being:

¾ = exp(—k/ε_.)

The method of claim 1, characterized in that the step of the sky-camera providing the data processor with imagery is done in regular time intervals within the range of every one minute to every thirty minutes. The method of claim 1, characterized in that the step of the satellite providing the data processor with imagery is done in regular time intervals within the range of every thirty minutes to every ninety minutes.

The method of claim 1, characterized in that the sky- camera has a fisheye lens and the step of applying, by means of the data processor, a cloud coverage index algorithm to the sky-camera imagery, further includes a step of correcting image distortion resulting from the sky-camera's fisheye lens by rectifying radius distortion in each raw image obtained by the sky- camera so as to have an initial distorted image and a calibrated image, the calibration being done according to a formula where:

[u_Af v_A and |s¾. are image coordinates of point A in the calibrated image and of point in the distorted image provided by the sky camera and which the polar coordinates are described by

where

10. The method of claim 9, characterised in that calibration coefficients i¾ and s_e are computed by curve fitting, using manual matching between an orthographic image from a calibration grid and a sky camera image from the same grid. 11. The method of claim 1, characterized in that the step of inputting into the data processor the real^¬ time imagery provided by a sky-camera located proximate to the photovoltaic power plant's geographic location and the near-time imagery provided by a satellite of the area corresponding to the photovoltaic power plant's geographic location includes the additional step of transforming the calibrated image from a colour image to a grey-scale image .

12. The method of claim 11, characterized in that the step of transforming the calibrated image from a colour image to a grey-scale image is done using colour space R, G, B with the following greyscale transform formula:

_T, , B{u,v) -R( _sv B(u-,v) - G{u,v)

ί^{ίί: ~} 8{u,.v) + R(u_>v) B( _yv) + Gin, ;·)^'

where {u_>v) are pixel image coordinates.

13. The method of claims 11 and 12, characterized in that the step of transforming the calibrated image from a colour image to a grey-scale image further includes the step of calculating, by means of the data processor, an energy result for the greyscale image which is done by summing up power spectrum in accordance with the Fast Fourier Transform algorithm. 14. The method of claim 1, characterized in that in the step of inputting into the data processor near- time imagery provided by a satellite, said imagery is infrared imagery. 15. The method of claim 14, characterized in that a cloudiness index is determined, by means of the data processor, from the satellite imagery every 15 minutes in accordance with the formula: for a given time ί the cloudiness by the Earth Mover' s Distance between two probability density functions is lEMD(t

where P* is the probability density function of the most recently acquired infrared satellite image with respect to time t and x{t) is the time tag closer to time t . The method of claims 1 and 15, characterized in that mapping between cloudiness indexes and solar irradiance relies on irradiance measurements from a pyranometer located proximate to the sky-camera and on computation of clear sky irradiance. The method of claim 16 characterized in that the clear sky irradiance is given by the formula:

i_c, = S - q^m* · sin. ,

where 5 = 1367 m^_i is the solar constant, n is the n^th day of the current year, a is the solar altitude and air mass <?_-K is given by

The method of claims 16 and 17, characterized in that, considering a forecast horizon h , the forecasted solar irradiance at time £-§-&■ is given by the decrease of the clear sky irradiance at time M ¾ due to the impact of TFM.L at time t and time t— k_sc(t} ( k_sc(i) corresponds to a 15 minutes lag with respect to t ) and to the impact of the 1EMD for time t and time t— k^it) (fe_/siA) corresponds to a 1 hour lag with respect to (έ) , it being defined by ½:(£ + ·¾)—

' , where I_c._s{t - h} is the theoretic value of the clear sky irradiance at time t - h and t£_s¾) is the estimated loss of irradiance for time 14- k , whereby said loss is computed by

where j¾_=c are the parameters for horizon h , estimated using log-linear regression; E being the greyscale ("BRBG") image energy, computed by summing up the power spectrum obtained with the Fast Fourier Transform algorithm, and L being the mean luminance of the calibrated colour image, the cloudiness index computed from the sky camera imagery at time t being given by

A t ;

19. The method of claim 16, characterised in that the mapping encodes cloud motion by considering cloud indexes at different times.

20. The method of claim 1, characterized in that the data on tilt angle and orientation of the photovoltaic panels is determined using a solar tracking algorithm.

The method of claim 1, characterized in that the data on predetermined legal and technical limitations associated with the photovoltaic power plant are pre-loaded parameters stored in memory means that are in communication with the data processor.

The method of claim 21, characterized in that a pre-loaded parameter is the inverter limit.

23. The method of claim 21, characterized in that a pre-loaded parameter is the transformer capacity.

24. The method of claim 1, characterized in that the data on loss of efficiency of the power grid in accordance with pre-determined grid loss indexes is data related to DC wiring, DC/AC wiring, AC wiring and transformers .

25. The method of claim 1, characterized in that the simultaneous factor algorithm is between 95% and 99%.

26. The method of claim 1, characterized in that the merging method is a deterministic approach based on the skill of the external forecast provider.

The method of claim 1, characterized in that the merging method is a dynamic approach based on a moving time window that minimizes the error of the forecast values and the real values.

28. The method of claim 1 characterized in that the merging method assigns to the external information providers a weight, said weight resulting from minimizing the least squares error between a variable forecast value and an observed value, said forecast value being defined by a linear combination of the forecast values from the external information providers, where the weights are the unknowns in the aforementioned minimization.

29. The method of claim 28, characterized in that the weights are updated every 6 hours.

30. The method of claim 1, characterized in that the upscaling algorithm compares the sum of the forecasted power outputs for the group of photovoltaic power plants with the real power measured by metering devices .

31. The method of claims 1 and 30, characterized in that the upscaling algorithm corresponds to an upscaling curve which is a polynomial of sixth order.

32. The method of claim 1, characterized in that the data is communicated to the data processor by means of a file transfer protocol server.

33. A system for implementing the method of claim 1, characterized in that it comprises:

a data processor capable of receiving data from various inputs and sources and processing said data in accordance with algorithms;

a memory means capable of storing data and algorithms, said memory means being in communication with said data processor;

at least one SCADA system capable of monitoring and controlling one or more photovoltaic power plants and communicating data to and receiving data from the data processor;

at least one sky camera fitted with a fisheye lens for each photovoltaic power plant of the group, each sky camera placed proximate to its respective photovoltaic power plant and in communication with the data processor; at least one pyranometer for each photovoltaic power plant of the group, each pyranometer placed proximate to its respective photovoltaic power plant and in communication with the SCADA system;

at least one ambient temperature sensor for each photovoltaic power plant of the group, each ambient temperature sensor placed proximate to its respective photovoltaic power plant and in communication with the SCADA system;

and a main control center in communication with the SCADA system.