Photovoltaic power generation system power prediction method and system combined with qualitative weather index

Abstract

The invention discloses a photovoltaic power generation system power prediction method and system combined with a qualitative weather index. The method comprises the following steps: constructing a medium and short term photovoltaic power prediction model; the input of the medium-short-term photovoltaic power prediction model is weather forecast data, and the output of the medium-short-term photovoltaic power prediction model is power data; obtaining real weather forecast data, and obtaining a weather type according to the real weather forecast data; constructing a loss function according to the weather type, and optimizing the training process of the medium-short-term photovoltaic power prediction model to obtain a final power prediction model; and predicting the power of the photovoltaic power generation system based on the final power prediction model. The method has the advantages of high prediction precision and the like.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of photovoltaic power generation systems, and in particular to a power prediction method and system for photovoltaic power generation systems combined with qualitative weather indicators. Background Art

[0002] With the increasing application of photovoltaic power generation systems, more and more photovoltaic power sources are being connected to the distribution network, posing significant challenges to the planning, operation, and control of power systems. Since solar radiation is closely related to meteorological conditions, the output power of photovoltaic power generation systems is inherently random and volatile. Given that the mismatch between power storage facilities and renewable energy grid-connected power is difficult to change in the short term, the integration of large-scale photovoltaic power generation systems into the grid will have a significant impact on the safe and stable operation of the power system. This is also a key technical issue that needs to be addressed for the large-scale integration of photovoltaic power generation into the grid. Countries around the world have successively conducted technical research on photovoltaic power generation power prediction, which is of great significance to the stable operation of power systems. It helps power system dispatching departments coordinate the power generation planning of conventional energy and photovoltaic power generation, and rationally arrange the operation mode of the power grid.

[0003] Based on the time scale, photovoltaic power generation forecasts can be divided into ultra-short-term forecasts, short-term forecasts, and medium-term forecasts. The time scale for ultra-short-term forecasts is 15 minutes to 4 hours into the future, while the time scale for medium-term forecasts is 0:00 to 240:00 the next day. Currently, medium-term and short-term photovoltaic power generation forecasts mainly rely on machine learning and deep learning models. These models predict the power generation capacity of photovoltaic stations within a specific time period in the future by establishing a mapping relationship between meteorological conditions and power generation capacity of power stations. When building these models, commonly used meteorological data sources include the European Center for Climate Data (ECNC), information provided by the Global Forecast System (GFS), data released by the China Meteorological Administration, and data from some commercial meteorological applications. However, how to efficiently integrate and utilize these diverse meteorological data sources to further enhance the accuracy and performance of photovoltaic power prediction models is a key issue that needs to be addressed.

[0004] For application scenarios with multiple meteorological sources, the current processing method is mostly to fuse multiple meteorological sources together to train medium- and short-term models. The method of using multi-source meteorological fusion to train models is limited to the application of meteorological forecasting, and the data for training models is also limited to historical actual power generation, resulting in low prediction accuracy. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art, the present invention provides a method and system for predicting the power of a photovoltaic power generation system in combination with qualitative weather indicators with high prediction accuracy.

[0006] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0007] A photovoltaic power generation system power prediction method combined with qualitative weather indicators includes the following steps:

[0008] Constructing a short- to medium-term photovoltaic power prediction model; the input of the short- to medium-term photovoltaic power prediction model is weather forecast data, and the output is power data;

[0009] Obtain real weather forecast data and obtain weather type based on the real weather forecast data;

[0010] A loss function is constructed based on weather types to optimize the training process of the medium- and short-term photovoltaic power prediction model to obtain the final power prediction model.

[0011] The power of the photovoltaic power generation system is predicted based on the final power prediction model.

[0012] Preferably, the weather types include sunny, cloudy, overcast, light rain, heavy rain and snowfall.

[0013] Preferably, a loss function is constructed according to the weather type, specifically:

[0014]

[0015] Where n is the number of days covered by the training data, y i,k is the actual power generation at the kth moment on the i-th day, is the model's predicted value of power generation at the kth moment on the i-th day; C is the number of weather types, q i,j is the value of the jth element of the true label corresponding to the i-th day, p i,j is the value of the jth element of the predicted probability vector for the i-th day; λ is the weight; and m is the number of predicted values ​​of generated power in a day.

[0016] Preferably, the weather forecast data includes forecast radiation, temperature, humidity, air pressure and wind speed.

[0017] Preferably, the medium- to short-term photovoltaic power prediction model is a deep learning model capable of processing sequence data, including a long short-term memory network or a time domain convolutional network.

[0018] Preferably, a weather type model is constructed, the input of the weather type model is real weather forecast data, and the output is the weather type.

[0019] Preferably, the weather type model is a deep learning model, including a fully connected neural network, a long short-term memory network, a time domain convolutional network or a graph neural network model.

[0020] The present invention also discloses a computer program product, comprising a computer program, which executes the steps of the above method when executed by a processor.

[0021] The present invention further discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described above are executed.

[0022] The present invention also discloses a photovoltaic power generation system power prediction system combined with qualitative weather indicators, including an interconnected memory and a processor, wherein the memory stores a computer program, and when the computer program is run by the processor, the steps of the above method are executed.

[0023] Compared with the prior art, the advantages of the present invention are:

[0024] This invention builds a weather type model that categorizes meteorological data into different types, such as sunny, cloudy, and rainy, providing more refined input features for the power prediction model. This classification method enables the model to better understand the impact of different weather conditions on power output. For example, sunny days have higher sunlight intensity, which improves photovoltaic power generation efficiency, while rainy days may lead to a decrease in power generation efficiency. By identifying these weather types, the model can adjust its prediction strategy based on different meteorological conditions, thereby improving prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Schematic diagram of the structure of the power prediction model in the prior art.

[0026] Figure 2 It is a structural diagram of the power prediction model in an embodiment of the present invention.

[0027] Figure 3 This is a flow chart of an embodiment of the power prediction method of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0029] like Figure 1 As shown in the figure, the basic framework of the medium- and short-term photovoltaic power prediction model is to predict power based on meteorological forecast data ( Figure 1 ), most power prediction models are machine learning models and deep learning models, such as SVR, xgboost, lightGBM, LSTM (long short-term memory neural network model), TCN (one-dimensional convolutional neural network model), etc. The loss function of the optimization model is the mean square error (MSE loss), and the specific formula is shown in formula (1):

[0030]

[0031] Where n is the number of samples, y i is the actual generated power, The model predicts the generated power. Using the mean square error loss function directly reflects the error between the model's predicted power and the actual power generated. Therefore, conventional power prediction models all use the MSE loss function.

[0032] Based on this, in addition to specific power generation values, qualitative information about actual weather conditions can also be obtained, such as sunny, overcast, cloudy, light rain, heavy rain, and snowfall. Photovoltaic power generation is affected by weather, with high efficiency on sunny days and low efficiency on rainy days. If snow covers the panels, power generation can even drop to zero. Therefore, accurately predicting weather types can improve the model's power prediction performance.

[0033] The present invention is based on combining actual weather types with power forecast models, and the specific way of combining is by adding additional loss function terms.

[0034] like Figure 3 As shown, the photovoltaic power generation system power prediction method combined with qualitative weather indicators provided by the embodiment of the present invention includes the following steps:

[0035] Construct a short- to medium-term photovoltaic power prediction model; the input of the short- to medium-term photovoltaic power prediction model is weather forecast data, and the output is power data;

[0036] Build a weather type model, obtain real weather forecast data, and input the real weather forecast data into the weather type model to obtain the weather type;

[0037] A loss function is constructed based on weather types to optimize the training process of the medium- and short-term photovoltaic power prediction model to obtain the final power prediction model.

[0038] The power of the photovoltaic power generation system is predicted based on the final power prediction model.

[0039] Specifically, if Figure 2 As shown, the short- to medium-term photovoltaic power prediction model ( Figure 2 The input of Model A in the example is weather forecast data, and the output is power data for 96 times a day (with a resolution of 15 minutes). Model A is a deep learning model capable of processing sequential data, including but not limited to LSTM (Long Short-Term Memory) and TCN (Time Convolutional Network).

[0040] The weather type model (Model B) takes weather forecast data as input and outputs the weather type for the corresponding date. It has C nodes, corresponding to weather types such as sunny, cloudy, partly cloudy, light rain, heavy rain, and snow. Model B can be any deep learning model, including but not limited to a fully connected neural network (NN), a long short-term memory network (LSTN), a temporal convolutional network (TCN), or a graph neural network.

[0041] Figure 2 In the model framework presented, the number of output nodes is C + 96, corresponding to 96 power levels and C weather types. Therefore, when preparing training data, in addition to preparing actual power generation data as annotations, it is also necessary to collect the actual weather types for a single day.

[0042] Specifically, the specific formula for constructing the loss function according to the weather type is:

[0043]

[0044] Where n is the number of days covered by the training data, y i,k is the actual power generation at the kth moment on the i-th day, is the model's predicted value of power generation at the kth moment on the i-th day; C is the number of weather types, q i,j is the value of the jth element of the true label corresponding to the i-th day (q i,j is the one-hot encoded vector, for the true category, q i,j is 1, and for other categories, q i,j is 0), p i,j is the value of the jth element of the predicted probability vector for the i-th day; λ is the weight.

[0045] By constructing an additional loss function using formula (2), the actual daily weather type is used to correct the weather forecast data. If the input data has multiple weather sources, introducing the actual weather type can also guide the model to select the optimal weather source.

[0046] This invention builds a weather type model that categorizes meteorological data into different types, such as sunny, cloudy, and rainy, providing more refined input features for the power prediction model. This classification method enables the model to better understand the impact of different weather conditions on power output. For example, sunny days have higher sunlight intensity, which improves photovoltaic power generation efficiency, while rainy days may lead to a decrease in power generation efficiency. By identifying these weather types, the model can adjust its prediction strategy based on different meteorological conditions, thereby improving prediction accuracy.

[0047] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0048] based on Figure 2 The framework shown here trains a power prediction model for a single PV plant. The data used to train the model is assumed to cover the period from January 1, 2023, to May 30, 2024. The meteorological data input to the model comes from two sources, containing the meteorological features predicted irradiance (R), temperature (T), humidity (H), air pressure (P), and wind speed (WS). The data has a 15-minute resolution. Therefore, there are a total of 10 features input to the model: R1, T1, H1, P1, WS1, R2, T2, H2, P2, and WS2, with the subscripts corresponding to the meteorological sources. In addition to the meteorological data, actual power data from January 1, 2023, to May 30, 2024, also with a 15-minute resolution, is also required. Furthermore, actual weather type data is required for each day from January 1, 2023, to May 30, 2024. The actual weather type and actual power data are used to train the model based on the loss function constraints. Of course, if there are multiple meteorological sources, introducing real weather types can also guide the model to select the optimal meteorological source.

[0049] The present invention also discloses a computer program product, comprising a computer program, which executes the steps of the above method when executed by a processor.

[0050] The present invention further discloses a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method described above are executed.

[0051] The present invention also discloses a photovoltaic power generation system power prediction system combined with qualitative weather indicators, including an interconnected memory and a processor, wherein the memory stores a computer program, and when the computer program is run by the processor, the steps of the above method are executed.

[0052] The products, media and systems of the present invention correspond to the above-mentioned methods and also have the advantages described above.

[0053] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned method embodiment. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0054] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A photovoltaic power generation system power prediction method combined with qualitative weather indicators, characterized in that: Including steps: Constructing a short- to medium-term photovoltaic power prediction model; the input of the short- to medium-term photovoltaic power prediction model is weather forecast data, and the output is power data; Obtain real weather forecast data and obtain weather type based on the real weather forecast data; A loss function is constructed based on weather types to optimize the training process of the medium- and short-term photovoltaic power prediction model to obtain the final power prediction model. The power of the photovoltaic power generation system is predicted based on the final power prediction model.

2. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 1 is characterized in that: The weather types include sunny, cloudy, overcast, light rain, heavy rain and snowfall.

3. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 1 is characterized in that: Construct a loss function based on the weather type, specifically: Where n is the number of days covered by the training data, y i,k is the actual power generation at the kth moment on the i-th day, is the model's predicted value of power generation at the kth moment on the i-th day; C is the number of weather types, q i,j is the value of the jth element of the true label corresponding to the i-th day, p i,j is the value of the jth element of the predicted probability vector for the i-th day; λ is the weight; and m is the number of predicted values ​​of generated power in a day.

4. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 1, 2 or 3, characterized in that: The weather forecast data includes forecast radiation, temperature, humidity, air pressure and wind speed.

5. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 1, 2 or 3, characterized in that: The medium- and short-term photovoltaic power prediction model is a deep learning model that can process sequence data, including long short-term memory networks or time domain convolutional networks.

6. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 1, 2 or 3, characterized in that: Construct a weather type model. The input of the weather type model is real weather forecast data, and the output is weather type.

7. The photovoltaic power generation system power prediction method combined with qualitative weather indicators according to claim 6, characterized in that: The weather type model is a deep learning model, including a fully connected neural network, a long short-term memory network, a time-domain convolutional network, or a graph neural network model.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

10. A photovoltaic power generation system power prediction system combined with qualitative weather indicators, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

Images