Photovoltaic output power prediction method and system, electronic equipment and medium

Abstract

The invention provides a photovoltaic output power prediction method and system, electronic equipment and a medium. The photovoltaic output power prediction method comprises the following steps: acquiring a first current data set, a second current data set and a third current data set; inputting the first current data set into a physical model to output first predicted photovoltaic output power; inputting the second current data set into a compensation model to output a compensation value of the first predicted photovoltaic output power; inputting the third current data set into the data model to output a second predicted photovoltaic output power; and performing weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine a prediction result of the photovoltaic output power. According to the invention, the photovoltaic output power data model is combined with the physical model, and the prediction result of the physical model is compensated according to the actual photovoltaic environment condition, so that the prediction precision of the photovoltaic output power is improved, and a photovoltaic power generation system is helped to carry out refined management and control.

Description

Technical Field

[0001] The present disclosure relates to the field of industrial intelligence technology, and in particular to a photovoltaic output power prediction method, system, electronic equipment and medium. Background Art

[0002] Electric energy production is a fundamental factor in current development. However, the growing demand for energy has had a negative impact on the environment, as fossil fuels are the main source of energy. The consequences of this situation include natural disasters that have a profound impact on humanity. Therefore, it is essential to use alternative, environmentally friendly sources of energy. Photovoltaic solar energy (PV) is one of the most attractive and promising renewable energy sources, as there are many places with high sunshine intensity and low installation costs for PV, making it one of the most attractive and promising renewable energy sources.

[0003] However, photovoltaic solar power generation is greatly affected by factors such as weather and environment, which makes its stability and the prediction accuracy of its power generation through traditional mechanism models poor. In recent years, with the development and popularization of artificial intelligence and machine learning technology, many scholars and researchers have tried to apply data-driven algorithm models to the field of photovoltaic solar energy and have achieved certain results. The core idea of ​​these methods is to collect a large amount of photovoltaic environmental data, record the power generation as data labels, and combine machine learning methods including multi-layer perceptron and random forest to train models for short-term prediction of photovoltaic power generation (i.e. photovoltaic output power).

[0004] However, there are several problems with predicting photovoltaic output power based on data-driven algorithm models: First, pure data models have low reliability and interpretability due to their black box characteristics and limitations such as the range of training samples of the data model itself. Second, photovoltaic power generation is affected by many factors including radiation illumination, temperature, wind speed, air pressure, humidity, etc., and high data synchronization is required when modeling. Third, since the efficiency of photovoltaic power generation is often affected by factors such as material aging and environmental changes, the constructed models usually become invalid after a period of time and are difficult to use continuously. Fourth, variables such as radiation illumination, temperature, wind speed, air pressure, and humidity are affected by the outdoor environment, and there are many missing values ​​and outliers, which will directly affect the photovoltaic power prediction model.

[0005] In general, both physical models and machine learning models can be used to predict photovoltaic output power. The physical model is more reliable because it combines actual physical properties and calibration results, but it fails to consider environmental factors, resulting in errors in the prediction results of photovoltaic output power. The machine learning model uses pure data mapping to make predictions. By incorporating environmental factors into the model input, the prediction accuracy can be improved for working conditions within the training sample set, but the reliability is low, especially when the predicted working conditions exceed the sample set, the prediction results will be extremely unstable. Summary of the invention

[0006] The technical problem to be solved by the present disclosure is to overcome the defects of inaccurate or unreliable photovoltaic output power prediction results in the prior art, and to provide a photovoltaic output power prediction method, system, electronic device and medium.

[0007] The present invention solves the above technical problems through the following technical solutions:

[0008] The present disclosure provides a photovoltaic output power prediction method, the photovoltaic output power prediction method comprising:

[0009] Acquire a first current data set, a second current data set, and a third current data set; wherein the first current data set includes a calibration data type and an operating condition data type, the second current data set includes an operating condition data type and an environmental data type, and the third current data set includes a calibration data type, an operating condition data type, and an environmental data type;

[0010] inputting the first current data set into a physical model to output a first predicted photovoltaic output power;

[0011] inputting the second current data set into a compensation model to output a compensation value of the first predicted photovoltaic output power;

[0012] Inputting the third current data set into a data model to output a second predicted photovoltaic output power;

[0013] A weighted sum is performed on the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power to determine a prediction result of the photovoltaic output power.

[0014] Preferably, the step of performing weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine the prediction result of the photovoltaic output power comprises:

[0015] Obtain historical actual photovoltaic output power;

[0016] Acquire a first historical data set, a second historical data set, and a third historical data set;

[0017] The first historical data set and the first current data set contain the same data type, the second historical data set and the second current data set contain the same data type, and the third historical data set and the third current data set contain the same data type;

[0018] Inputting the first historical data set into a physical model to output a first historical predicted photovoltaic output power;

[0019] inputting the second current data set into a compensation model to output a historical compensation value of the first historical predicted photovoltaic output power;

[0020] Inputting the third current data set into a data module to output a second historical predicted photovoltaic output power;

[0021] Performing a multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value, and the second historical predicted photovoltaic output power to respectively determine a first weighting coefficient, a second weighting coefficient, and a third weighting coefficient of the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power;

[0022] The prediction result of the photovoltaic output power is determined according to the sum of the first product, the second product and the third product; wherein the first product is the product of the first predicted photovoltaic output power and the first weighting coefficient, the second product is the product of the compensation value and the second weighting coefficient, and the third product is the product of the second predicted photovoltaic output power and the third weighting coefficient.

[0023] Preferably, the step of performing a multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value, and the second historical predicted photovoltaic output power to respectively determine a first weighting coefficient, a second weighting coefficient, and a third weighting coefficient of the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power includes:

[0024] Determine the following objective function:

[0025] f(a,b,c)=|La*pst1-b*pst2-c*pst3|;

[0026] Wherein, f(a, b, c) is used to characterize the objective function, L is used to characterize the historical photovoltaic actual output power, pst1 is used to characterize the first historical predicted photovoltaic output power, pst2 is used to characterize the historical compensation value, pst3 is used to characterize the second historical predicted photovoltaic output power, a is used to characterize the first weighting coefficient of the first predicted photovoltaic output power, b is used to characterize the second weighting coefficient of the compensation value, and c is used to characterize the third weighting coefficient of the second predicted photovoltaic output power;

[0027] A multivariate linear regression calculation is performed on the objective function to determine a, b, and c corresponding to the minimum value of f(a, b, c).

[0028] Preferably, the step of obtaining the first current data set, the second current data set and the third current data set comprises:

[0029] Interpolation and filling are performed on the time-related data in the first current data set, the second current data set, and the third current data set.

[0030] Preferably, the photovoltaic output power prediction method further includes:

[0031] Get the current actual photovoltaic output power;

[0032] The data model is updated according to the third current data set and the current actual photovoltaic output power.

[0033] Preferably, the physical model includes a first formula and a second formula:

[0034] The first formula is as follows:

[0035]

[0036] Among them, T C Used to characterize the current temperature of the photovoltaic panel, T a It is used to characterize the current ambient temperature, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, and NOCT is used to characterize the standard operating temperature;

[0037] The second formula is as follows:

[0038]

[0039] Among them, P PV For characterizing the first predicted photovoltaic output power, P STC It is used to characterize the power output value under standard test conditions, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, G STC It is used to characterize the incident irradiance of photovoltaic panels under standard test conditions, and γ is used to characterize the power temperature coefficient, TC Used to characterize the current temperature of the photovoltaic panel, T STC Used to characterize the temperature of photovoltaic panels under standard test conditions;

[0040] and / or,

[0041] The compensation model includes at least one of a gated recurrent unit neural network, a convolutional neural network, and a long short-term memory neural network;

[0042] and / or,

[0043] The data model includes a random forest data model.

[0044] The present disclosure also provides a photovoltaic output power prediction system, the photovoltaic output power prediction system comprising:

[0045] an acquisition module, configured to acquire a first current data set, a second current data set, and a third current data set; wherein the first current data set includes a calibration data type and an operating condition data type, the second current data set includes an operating condition data type and an environmental data type, and the third current data set includes a calibration data type, an operating condition data type, and an environmental data type;

[0046] A physical model module, configured to input the first current data set into a physical model to output a first predicted photovoltaic output power;

[0047] a compensation model module, configured to input the second current data set into a compensation model to output a compensation value of the first predicted photovoltaic output power;

[0048] A data model module, configured to input the third current data set into a data model to output a second predicted photovoltaic output power;

[0049] The prediction module is used to perform weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine a prediction result of the photovoltaic output power.

[0050] The present disclosure also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and used to run on the processor, and the processor implements the above-mentioned photovoltaic output power prediction method when executing the computer program.

[0051] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned photovoltaic output power prediction method is implemented.

[0052] The present disclosure also provides a computer program product, including a computer program, which implements the above-mentioned photovoltaic output power prediction method when executed by a processor.

[0053] On the basis of being in accordance with the common sense in the art, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present disclosure.

[0054] The positive and progressive effects of this disclosure are:

[0055] The present invention combines the photovoltaic output power data model with the physical model and compensates the prediction results of the physical model according to the actual photovoltaic environment conditions, thereby improving the prediction accuracy of the photovoltaic output power and helping to carry out refined management and control of the photovoltaic power generation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 A flowchart of a photovoltaic output power prediction method provided in Embodiment 1 of the present disclosure;

[0057] Figure 2 A schematic diagram of a model structure of a specific example of a photovoltaic output power prediction method provided in Embodiment 1 of the present disclosure;

[0058] Figure 3 A structural diagram of a photovoltaic output power prediction system provided in Example 2 of the present disclosure;

[0059] Figure 4 A schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present disclosure. DETAILED DESCRIPTION

[0061] The present disclosure is further described below by way of examples, but the present disclosure is not limited to the scope of the examples.

[0062] Prefixes such as "first" and "second" are used in the embodiments of the present disclosure only to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers to distinguish description objects in the embodiments of the present disclosure does not constitute a limitation on the described objects. For the statement of the described objects, please refer to the description in the context of the claims or embodiments, and no unnecessary limitation should be constituted due to the use of such prefixes. In addition, in the description of the present embodiment, unless otherwise specified, the meaning of "plurality" is two or more.

[0063] In the embodiments of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0064] Example 1

[0065] This embodiment is an application method of artificial intelligence technology and machine learning algorithm in industrial data analysis scenarios. In view of the problem that the output power in the photovoltaic power generation process is easily affected by environmental factors, this embodiment provides a photovoltaic output power prediction method, see Figure 1 , the prediction methods of photovoltaic output power include:

[0066] S1. Obtain a first current data set, a second current data set, and a third current data set.

[0067] Among them, the first current data set includes calibration data types and operating condition data types, the second current data set includes operating condition data types and environmental data types, and the third current data set includes calibration data types, operating condition data types and environmental data types.

[0068] Specifically: Environmental data types include wind speed, air pressure, humidity, ambient temperature and other data types; operating condition data types include radiant illumination, actual output power and other data types; calibration data types include power output value of photovoltaic under standard test conditions, incident irradiance under standard test conditions, power temperature coefficient, temperature of photovoltaic panels under standard test conditions, standard operating temperature and other data types.

[0069] S2. Inputting a first current data set into a physical model to output a first predicted photovoltaic output power.

[0070] In an optional embodiment, the physical model includes a first formula and a second formula:

[0071] The first formula is as follows:

[0072]

[0073] Among them, T C Used to characterize the current temperature of the photovoltaic panel, T a It is used to characterize the current ambient temperature, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, and NOCT is used to characterize the standard operating temperature.

[0074] The second formula is as follows:

[0075]

[0076] Among them, P PV For characterizing the first predicted photovoltaic output power, P STC It is used to characterize the power output value under standard test conditions, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, G STC It is used to characterize the incident irradiance of photovoltaic panels under standard test conditions, and γ is used to characterize the power temperature coefficient, T C Used to characterize the current temperature of the photovoltaic panel, T STCUsed to characterize the temperature of the photovoltaic panel under standard test conditions.

[0077] S3. Input the second current data set into the compensation model to output the compensation value of the first predicted photovoltaic output power.

[0078] In an alternative embodiment, the compensation model includes at least one of a gated recurrent unit neural network (GRU), a convolutional neural network CNN, and a long short-term memory neural network LSTM.

[0079] During the process of training the compensation model, select the corresponding input dimension n according to the number of feature columns of the first sample set SA (including the second historical data), and set the number of hidden layer neuron nodes to three times n. The label column of the first sample set is the difference between the historical actual photovoltaic output power and the theoretical value calculated by the physical model. Use data types such as wind speed, air pressure, humidity, ambient temperature, and irradiance in the first sample set as the input of the compensation model, and use the difference between the historical actual photovoltaic output power and the theoretical value calculated by the physical model as the output for model training to obtain the compensation model.

[0080] S4. Input the third current data set into the data model to output the second predicted photovoltaic output power.

[0081] In an alternative embodiment, the data model includes a random forest data model.

[0082] During the process of training the data model, for constructing the second sample set SA', first set the time step as t (1 < t <= 3), and then extend and reconstruct the data (such as wind speed, air pressure, humidity, ambient temperature, and irradiance) within the time window of the first sample set SA according to the length of the time step t to form the second sample set SA'. The label column of the second sample set SA' is the historical actual photovoltaic output power.

[0083] Select the corresponding input dimension n' according to the number of feature columns of the second sample set SA'. Set the number of trees (n_estimators) as n' * 5, the maximum depth (max_depth) as n' * 3, and the evaluation criterion (criterion) as MSE (mean squared error).

[0084] Use data types such as wind speed, air pressure, humidity, ambient temperature, and irradiance in the second sample set SA' as the model input, and use the historical actual photovoltaic output power as the model output for model training to obtain the data model.

[0085] S5. Perform weighted summation on the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power to determine the prediction result of the photovoltaic output power.

[0086] This embodiment combines the photovoltaic output power data model with the physical model and compensates the prediction result of the physical model according to the actual photovoltaic environment conditions, thereby improving the prediction accuracy of the photovoltaic output power and helping to carry out refined management and control of the photovoltaic power generation system.

[0087] In an optional embodiment, step S5 includes:

[0088] S51. Obtain historical actual photovoltaic output power.

[0089] S52: Obtain a first historical data set, a second historical data set, and a third historical data set.

[0090] The first historical data set contains the same data type as the first current data set, the second historical data set contains the same data type as the second current data set, and the third historical data set contains the same data type as the third current data set. That is, the first historical data set includes calibration data types and operating condition data types, the second historical data set includes operating condition data types and environmental data types, and the third historical data set includes calibration data types, operating condition data types, and environmental data types.

[0091] Specifically: Environmental data types include wind speed, air pressure, humidity, ambient temperature and other data types; operating condition data types include radiant illumination, actual output power and other data types; calibration data types include power output value of photovoltaic under standard test conditions, incident irradiance under standard test conditions, power temperature coefficient, temperature of photovoltaic panels under standard test conditions, standard operating temperature and other data types.

[0092] S53: Input the first historical data set into the physical model to output a first historical predicted photovoltaic output power.

[0093] S54: Input the second current data set into the compensation model to output a historical compensation value of the first historical predicted photovoltaic output power.

[0094] S55 , inputting the third current data set into the data module to output a second historical predicted photovoltaic output power.

[0095] S56. Perform a multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value, and the second historical predicted photovoltaic output power to respectively determine the first weighting coefficient, the second weighting coefficient, and the third weighting coefficient of the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power.

[0096] In an optional implementation, step S56 includes:

[0097] S561, determine the following objective function:

[0098] f(a,b,c)=|L–a*pst1-b*pst2–c*pst3|;

[0099] Among them, f(a, b, c) is used to characterize the objective function, L is used to characterize the historical photovoltaic actual output power, pst1 is used to characterize the first historical predicted photovoltaic output power, pst2 is used to characterize the historical compensation value, pst3 is used to characterize the second historical predicted photovoltaic output power, a is used to characterize the first weighting coefficient of the first predicted photovoltaic output power, b is used to characterize the second weighting coefficient of the compensation value, and c is used to characterize the third weighting coefficient of the second predicted photovoltaic output power.

[0100] S562. Perform a multivariate linear regression calculation on the objective function to determine a, b, and c corresponding to when f(a, b, c) is minimum.

[0101] S57. Determine a prediction result of the photovoltaic output power according to the sum of the first product, the second product and the third product.

[0102] The first product is the product of the first predicted photovoltaic output power and the first weighting coefficient, the second product is the product of the compensation value and the second weighting coefficient, and the third product is the product of the second predicted photovoltaic output power and the third weighting coefficient.

[0103] In this embodiment, the first weighting coefficient of the first predicted photovoltaic output power, the second weighting coefficient of the compensation value, and the third weighting coefficient of the second predicted photovoltaic output power are determined by performing a multivariate linear regression calculation on the objective function, etc. Then, the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power are weighted and summed to determine the final result of the photovoltaic output power, thereby improving the prediction accuracy and reducing the impact of the environment on the predicted photovoltaic output power.

[0104] In an optional embodiment, step S1 further includes:

[0105] Interpolation and filling are performed on the time-related data in the first current data set, the second current data set, and the third current data set.

[0106] Among them, time-related data include radiation illumination, wind speed, air pressure, humidity, ambient temperature and other data.

[0107] Specific interpolation filling formulas are as follows:

[0108]

[0109] Among them, x t is the missing variable value at time t, x t-1 is the missing variable value at time t-1, x t+1is the missing variable value at time t+1, T t is the specific timestamp at time t, T t-1 is the specific timestamp at time t-1, T t+1 It is the specific timestamp of time t+1.

[0110] In an optional embodiment, the photovoltaic output power prediction method further includes:

[0111] S6. Obtain the current actual photovoltaic output power.

[0112] S7. Update the data model according to the third current data set and the current actual photovoltaic output power.

[0113] In this embodiment, incremental learning can be performed by re-inputting the currently collected first current data set, second current data set, third current data set and current photovoltaic actual output power into the data model, in order to obtain a data model with higher photovoltaic output power prediction accuracy, thereby further improving the accuracy of photovoltaic output power prediction results.

[0114] In addition, during the data training process, because multiple types of environmental data that change over time (for example: irradiance, temperature, wind speed, air pressure, humidity, etc.) need to be collected separately by multiple sensors, the timestamps of the collected signals are highly asynchronous, and when performing multi-data source fusion modeling, the data synchronization requirements are high. The traditional synchronization method that only aligns the timestamps will cause great interference to the model and affect the model accuracy. Therefore, this embodiment also addresses the problem of highly asynchronous timestamps of the collected signals. In the process of constructing the photovoltaic output power prediction data model and compensation model, the dynamic data window is divided or merged according to the frequency of each data, solving the problem of multi-source data synchronization.

[0115] For example, assuming that the acquisition frequencies of sensor signals s1, s2, ... sn (sensor signal sn corresponds to a time-related environmental data type) are f1, f2, ... fn respectively, select the maximum acquisition frequency f* and construct a time window with a width of 2 / f*.

[0116] The sensor signal corresponding to f* is used as the time window benchmark and 2 / f* is used as the width to dynamically divide the time window. Each data is sliced ​​into m time windows. Each time window is a sample, and a total of m non-overlapping samples are obtained.

[0117] The mean or maximum value of the data values ​​of the environmental data types such as irradiance, wind speed, air pressure, humidity, and ambient temperature in each sample is retained to reduce the dimensionality of the sample. The sample set is converted into the first sample set SA of m*n dimensions.

[0118] The actual photovoltaic output power corresponding to the timestamp of the first sample set SA is obtained, and the average value L of the actual photovoltaic output power is used as the label column of the first sample set SA.

[0119] In this embodiment, by dynamically dividing or merging data windows, the synchronization of multi-source data is improved when performing multi-data source fusion modeling.

[0120] A specific example is introduced below to describe in detail the photovoltaic output power prediction method of this embodiment.

[0121] The first step is to obtain current and historical data sets of the environment, operating conditions, and calibration related to the PV output power.

[0122] The environmental data types include wind speed, air pressure, humidity, ambient temperature, etc.; the operating condition data types include radiant illumination, actual output power, etc.; the test calibration data include the power output value of photovoltaic under standard test conditions, the incident irradiance under standard test conditions, the power temperature coefficient, the temperature of the photovoltaic panel under standard test conditions, the standard operating temperature, etc.

[0123] The second step is to preprocess key data.

[0124] Among them, the key data is related to the time dimension. It is necessary to interpolate and fill in the missing values ​​of irradiance, wind speed, air pressure, humidity, and ambient temperature in turn according to the time dimension.

[0125] The third step is to divide the dynamic data window according to the frequency of each historical data.

[0126] In this example, the minimum acquisition frequency of all sensor signals is 10 minutes, so the maximum acquisition frequency f* = 1 / 600th of a second (s -1 ). A time window with a width of 2 / f*=20 minutes is constructed, that is, every 20 minutes is taken as a sample and merged, and 5862 samples are obtained.

[0127] The fourth step is to use the average value L of the actual photovoltaic output power as the label column of the first sample set SA.

[0128] The fifth step is to perform output power prediction calculation based on the physical model to obtain the sample theoretical output value T (i.e., the first historical predicted photovoltaic output power).

[0129] The sixth step is to calculate the theoretical compensation amount (ie, LT) and use it as the label column CL (ie, historical compensation value) of the compensation model.

[0130] In the seventh step, 80% of the samples (4690 samples) were randomly selected from the 5862 samples as training samples and input into the gated recurrent unit neural network (GRU), convolutional neural network CNN, and long short-term memory neural network LSTM. That is, the label column CL of the compensation model was trained, and the number of neuron nodes was set to 15 for training to obtain the compensation model C. The remaining 20% ​​of the samples (1172 samples) not used in the training were tested, and the test results shown in Table 1 were obtained. From the results, it can be seen that the prediction accuracy of photovoltaic output power can be greatly improved after using the machine learning model to compensate the physical model, and the effect of using GRU is the best.

[0131]

[0132]

[0133] Step 8: Build data samples for the data model.

[0134] The reconstructed sample set with a time step of 2 is formed to form 2931 samples (each sample consists of two timestamps), and the samples are input into the data model - random forest model. The number of trees (n_emtimators) is set to 50, the maximum depth (max_depth) is set to 30, and the evaluation criterion (criterion) is set to the training model of MSE (mean square error), and the data model M for photovoltaic output power prediction is obtained.

[0135] In the ninth step, predictions are made on the historical data set based on the physical model, the physical prediction compensation model C, and the data model M, and multiple linear regression is performed to calculate the first weighting coefficient a=7.252, the second weighting coefficient b=2.715, and the third weighting coefficient c=0.092.

[0136] In the tenth step, a weighted sum is performed on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to obtain a short-term prediction result of the photovoltaic output power, and the current value is re-input into the data model for iterative optimization.

[0137] See also Figure 2This example introduces data from multiple sources (including irradiance, temperature, humidity, etc.) and conducts a comprehensive analysis of photovoltaic output power prediction through multi-source data fusion analysis to increase the prediction accuracy of the prediction model. This example also proposes a method of correction guided by the physical model and assisted by the compensation model. At the same time, by introducing the correction coefficient, the intervention degree of the compensation model can be adjusted according to the actual test verification situation, which improves the operability of the model and optimization. That is, by collecting environmental data, a deep learning compensation model based on the gated recurrent unit (GRU) is constructed for the error between the predicted output power of the physical model and the actual output power, thereby improving the reliability of the overall prediction; and by introducing the data model to comprehensively measure the impact of environmental factors on the predicted value of the physical model, and output the compensation amount to improve the overall stability of the model. In addition, when doing data preprocessing, this example performs interpolation filling based on the time series, corrects outliers and missing values, thereby improving the robustness of the model. In the modeling process, this example also divides and merges samples through dynamic time windows to improve the synchronization of multiple data sources.

[0138] Finally, this example also takes into account the characteristics of physical models that are prone to change over time and performance degradation. By constructing a physical prediction model for photovoltaic output power, a random forest data prediction model, and a GRU physical prediction compensation model, and setting intervention coefficients, the final model prediction results can be adjusted according to the participation of the three over time. At the same time, an incremental learning element is introduced into the random forest data prediction model (data model), that is, the model is updated regularly, so that the random forest data prediction model can be adjusted according to factors such as photovoltaic material aging and environmental changes, so that the data model can be continuously iterated and optimized.

[0139] Example 2

[0140] This embodiment provides a photovoltaic output power prediction system, see Figure 3 , the photovoltaic output power prediction system includes:

[0141] The acquisition module 1 is used to acquire a first current data set, a second current data set and a third current data set. The first current data set includes a calibration data type and an operating condition data type, the second current data set includes an operating condition data type and an environmental data type, and the third current data set includes a calibration data type, an operating condition data type and an environmental data type.

[0142] The physical model module 2 is used to input a first current data set into the physical model to output a first predicted photovoltaic output power.

[0143] In an optional embodiment, the physical model includes a first formula and a second formula:

[0144] The first formula is as follows:

[0145]

[0146] Among them, T C Used to characterize the current temperature of the photovoltaic panel, T a It is used to characterize the current ambient temperature, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, and NOCT is used to characterize the standard operating temperature;

[0147] The second formula is as follows:

[0148]

[0149] Among them, P PV For characterizing the first predicted photovoltaic output power, P STC It is used to characterize the power output value under standard test conditions, G(β, ∝) is used to characterize the current incident irradiance of the photovoltaic panel, G STC It is used to characterize the incident irradiance of photovoltaic panels under standard test conditions, and γ is used to characterize the power temperature coefficient, T C Used to characterize the current temperature of the photovoltaic panel, T STC Used to characterize the temperature of photovoltaic panels under standard test conditions.

[0150] The compensation model module 3 is used to input the second current data set into the compensation model to output a compensation value of the first predicted photovoltaic output power.

[0151] In an optional embodiment, the compensation model includes at least one of a gated recurrent unit neural network, a convolutional neural network, and a long short-term memory neural network;

[0152] The data model module 4 is used to input the third current data set into the data model to output a second predicted photovoltaic output power.

[0153] In an optional embodiment, the data model includes a random forest data model.

[0154] The prediction module 5 is used to perform weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine the prediction result of the photovoltaic output power.

[0155] In an optional implementation, the acquisition module 1 is further used to acquire historical actual photovoltaic output power; and is further used to acquire a first historical data set, a second historical data set, and a third historical data set.

[0156] The first historical data set contains the same data type as the first current data set, the second historical data set contains the same data type as the second current data set, and the third historical data set contains the same data type as the third current data set.

[0157] The physical model module 2 is further used to input the first historical data set into the physical model to output a first historical predicted photovoltaic output power.

[0158] The compensation model module 3 is further used to input the second current data set into the compensation model to output a historical compensation value of the first historical predicted photovoltaic output power.

[0159] The data model module 4 is used to input the third current data set into the data module to output the second historical predicted photovoltaic output power.

[0160] See also Figure 3 , the photovoltaic output power prediction system also includes:

[0161] Calculation module 6 is used to perform multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value and the second historical predicted photovoltaic output power to respectively determine the first weighting coefficient, the second weighting coefficient and the third weighting coefficient of the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power; and is also used to determine the prediction result of the photovoltaic output power based on the sum of the first product, the second product and the third product.

[0162] The first product is the product of the first predicted photovoltaic output power and the first weighting coefficient, the second product is the product of the compensation value and the second weighting coefficient, and the third product is the product of the second predicted photovoltaic output power and the third weighting coefficient.

[0163] In an optional embodiment, the calculation module 6 is further used to determine the following objective function:

[0164] f(a,b,c)=|L–a*pst1-b*pst2–c*pst3|;

[0165] Among them, f(a, b, c) is used to characterize the objective function, L is used to characterize the historical photovoltaic actual output power, pst1 is used to characterize the first historical predicted photovoltaic output power, pst2 is used to characterize the historical compensation value, pst3 is used to characterize the second historical predicted photovoltaic output power, a is used to characterize the first weighting coefficient of the first predicted photovoltaic output power, b is used to characterize the second weighting coefficient of the compensation value, and c is used to characterize the third weighting coefficient of the second predicted photovoltaic output power.

[0166] The calculation module 6 is also used to perform a multivariate linear regression calculation on the target function to determine a, b, and c corresponding to the minimum value of f(a, b, c).

[0167] In an alternative embodiment, see Figure 3 , the photovoltaic output power prediction system also includes:

[0168] The filling module 7 is used to perform interpolation filling on the time-related data in the first current data set, the second current data set and the third current data set.

[0169] In an optional implementation, the acquisition module 1 is used to acquire the current actual photovoltaic output power.

[0170] See also Figure 3 , the photovoltaic output power prediction system also includes:

[0171] The updating module 8 is used to update the data model according to the third current data set and the current actual photovoltaic output power.

[0172] As for the system embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The system embodiment described above is only illustrative, in which the units described as separate components may or may not be physically separated, and the components as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution.

[0173] Example 3

[0174] Figure 4 This is a structural schematic diagram of an electronic device shown in an example embodiment of the present disclosure, the electronic device includes a memory, a processor, and a computer program stored in the memory and used to run on the processor, and when the processor executes the computer program, a method for predicting photovoltaic output power of any of the above embodiments is implemented. Figure 4 The electronic device 40 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0175] like Figure 4 As shown, the electronic device 40 may be in the form of a general-purpose computing device, for example, it may be a server device. The components of the electronic device 40 may include, but are not limited to: at least one processor 41, at least one memory 42, and a bus 43 connecting different system components (including the memory 42 and the processor 41).

[0176] The bus 43 includes a data bus, an address bus, and a control bus.

[0177] The memory 42 may include a volatile memory, such as a random access memory (RAM) 421 and / or a cache memory 422 , and may further include a read-only memory (ROM) 423 .

[0178] The memory 52 may also include a program tool 425 (or utility) having a set (at least one) of program modules 424, such program modules 424 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0179] The processor 41 executes various functional applications and data processing by running the computer program stored in the memory 42, such as the photovoltaic output power prediction method provided in any of the above embodiments.

[0180] The electronic device 40 may also communicate with one or more external devices 44 (e.g., keyboards, pointing devices, etc.). Such communication may be performed via an input / output (I / O) interface 45. Furthermore, the electronic device 40 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 46. As shown, the network adapter 46 communicates with other modules of the electronic device 40 via a bus 43. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 40, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems, etc.

[0181] It should be noted that although several units / modules or sub-units / modules of the electronic device are mentioned in the above detailed description, this division is merely exemplary and not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0182] Example 4

[0183] The embodiments of the present disclosure further provide a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the photovoltaic output power prediction method provided in any of the above embodiments is implemented.

[0184] The readable storage medium may include but is not limited to: a portable disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical storage device, a magnetic storage device or any suitable combination of the above.

[0185] Example 5

[0186] The embodiments of the present disclosure further provide a computer program product, including a computer program, which implements any of the above photovoltaic output power prediction methods when executed by a processor.

[0187] Among them, the program code for executing the computer program product of the present disclosure can be written in any combination of one or more programming languages, and the program code can be executed completely on the user device, partially on the user device, as an independent software package, partially on the user device and partially on a remote device, or completely on the remote device.

[0188] Although the specific embodiments of the present disclosure are described above, those skilled in the art should understand that this is only an example, and the protection scope of the present disclosure is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present disclosure, but these changes and modifications all fall within the protection scope of the present disclosure.

Claims

1. A method for predicting photovoltaic output power, characterized in that: The photovoltaic output power prediction method includes: Acquire a first current data set, a second current data set, and a third current data set; wherein the first current data set includes a calibration data type and an operating condition data type, the second current data set includes an operating condition data type and an environmental data type, and the third current data set includes a calibration data type, an operating condition data type, and an environmental data type; inputting the first current data set into a physical model to output a first predicted photovoltaic output power; inputting the second current data set into a compensation model to output a compensation value of the first predicted photovoltaic output power; Inputting the third current data set into a data model to output a second predicted photovoltaic output power; A weighted sum is performed on the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power to determine a prediction result of the photovoltaic output power.

2. The photovoltaic output power prediction method according to claim 1, characterized in that: The step of performing weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine the prediction result of the photovoltaic output power comprises: Obtain historical actual photovoltaic output power; Acquire a first historical data set, a second historical data set, and a third historical data set; The first historical data set and the first current data set contain the same data type, the second historical data set and the second current data set contain the same data type, and the third historical data set and the third current data set contain the same data type; Inputting the first historical data set into a physical model to output a first historical predicted photovoltaic output power; inputting the second current data set into a compensation model to output a historical compensation value of the first historical predicted photovoltaic output power; Inputting the third current data set into a data module to output a second historical predicted photovoltaic output power; Performing a multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value, and the second historical predicted photovoltaic output power to respectively determine a first weighting coefficient, a second weighting coefficient, and a third weighting coefficient of the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power; The prediction result of the photovoltaic output power is determined according to the sum of the first product, the second product and the third product; wherein the first product is the product of the first predicted photovoltaic output power and the first weighting coefficient, the second product is the product of the compensation value and the second weighting coefficient, and the third product is the product of the second predicted photovoltaic output power and the third weighting coefficient.

3. The photovoltaic output power prediction method according to claim 2, characterized in that: The step of performing a multivariate linear regression calculation on the historical photovoltaic actual output power, the first historical predicted photovoltaic output power, the historical compensation value, and the second historical predicted photovoltaic output power to respectively determine a first weighting coefficient, a second weighting coefficient, and a third weighting coefficient of the first predicted photovoltaic output power, the compensation value, and the second predicted photovoltaic output power comprises: Determine the following objective function: f(a,b,c)=|L–a*pst1-b*pst2–c*pst3|; Wherein, f(a, b, c) is used to characterize the objective function, L is used to characterize the historical photovoltaic actual output power, pst1 is used to characterize the first historical predicted photovoltaic output power, pst2 is used to characterize the historical compensation value, pst3 is used to characterize the second historical predicted photovoltaic output power, a is used to characterize the first weighting coefficient of the first predicted photovoltaic output power, b is used to characterize the second weighting coefficient of the compensation value, and c is used to characterize the third weighting coefficient of the second predicted photovoltaic output power; A multivariate linear regression calculation is performed on the objective function to determine a, b, and c corresponding to the minimum value of f(a, b, c).

4. The photovoltaic output power prediction method according to claim 1, characterized in that: The step of obtaining the first current data set, the second current data set and the third current data set comprises: Interpolation and filling are performed on the time-related data in the first current data set, the second current data set, and the third current data set.

5. The photovoltaic output power prediction method according to claim 1, characterized in that: The photovoltaic output power prediction method further includes: Get the current actual photovoltaic output power; The data model is updated according to the third current data set and the current actual photovoltaic output power.

6. The photovoltaic output power prediction method according to claim 1, characterized in that: The physical model includes a first formula and a second formula: The first formula is as follows: Among them, T C Used to characterize the current temperature of the photovoltaic panel, T a It is used to characterize the current ambient temperature, G(β,∝) is used to characterize the current incident irradiance of the photovoltaic panel, and NOCT is used to characterize the standard operating temperature; The second formula is as follows: Among them, P PV For characterizing the first predicted photovoltaic output power, P STC It is used to characterize the power output value under standard test conditions, G(β,∝) is used to characterize the current incident irradiance of the photovoltaic panel, G STC It is used to characterize the incident irradiance of photovoltaic panels under standard test conditions, and γ is used to characterize the power temperature coefficient, T C Used to characterize the current temperature of the photovoltaic panel, T STC Used to characterize the temperature of photovoltaic panels under standard test conditions; and / or, The compensation model includes at least one of a gated recurrent unit neural network, a convolutional neural network, and a long short-term memory neural network; and / or, The data model includes a random forest data model.

7. A photovoltaic output power prediction system, characterized in that: The photovoltaic output power prediction system comprises: an acquisition module, configured to acquire a first current data set, a second current data set, and a third current data set; wherein the first current data set includes a calibration data type and an operating condition data type, the second current data set includes an operating condition data type and an environmental data type, and the third current data set includes a calibration data type, an operating condition data type, and an environmental data type; A physical model module, configured to input the first current data set into a physical model to output a first predicted photovoltaic output power; a compensation model module, configured to input the second current data set into a compensation model to output a compensation value of the first predicted photovoltaic output power; A data model module, configured to input the third current data set into a data model to output a second predicted photovoltaic output power; The prediction module is used to perform weighted summation on the first predicted photovoltaic output power, the compensation value and the second predicted photovoltaic output power to determine a prediction result of the photovoltaic output power.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and used to run on the processor, characterized in that: When the processor executes the computer program, the photovoltaic output power prediction method according to any one of claims 1 to 6 is implemented.

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 photovoltaic output power prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the photovoltaic output power prediction method according to any one of claims 1 to 6 is implemented.

Images