Photovoltaic output data prediction method and system for time sequence production simulation

Abstract

The invention discloses a photovoltaic output data prediction method and system for time sequence production simulation. The method comprises the steps of performing detrending processing and deseasonal processing on historical photovoltaic output data; training a photovoltaic output data prediction model based on the historical photovoltaic output data after the detrending processing and the seasonal processing; and performing photovoltaic output data prediction by using the trained photovoltaic output data prediction model. According to the method, the influence of long-term trend offset and periodic seasonal offset on photovoltaic output in photovoltaic output is considered, and detrending processing and deseasonal processing are performed on historical photovoltaic output data, so that the randomness and volatility of photovoltaic output are simulated and predicted more accurately, the photovoltaic output (sequence) tends to be stable, and the accuracy of photovoltaic output is improved. Therefore, the accuracy and reliability of new energy power generation prediction and consumption capability evaluation are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic output prediction, and in particular relates to a photovoltaic output data prediction method and system for time series production simulation. Background Art

[0002] As an important renewable energy power generation method, the output power of photovoltaic power generation is affected by many factors, such as sunlight intensity, weather conditions, sunshine time, etc. When conducting time-series production simulation, technicians need to take these factors into account to accurately reflect the characteristics and output fluctuations of photovoltaic power generation. In order to reflect the randomness and volatility of photovoltaic power generation, randomized model inputs, consideration of intraday and seasonal changes, and introduction of weather forecasts can be used. Time-series production simulation of photovoltaic output is a complex process that requires the comprehensive use of meteorological, statistical and engineering technical knowledge to better manage and plan energy.

[0003] Current photovoltaic output data forecasting methods often ignore the long-term trend offset and periodic seasonal offset in photovoltaic power generation. When using general statistical methods or machine learning methods to forecast photovoltaic output data, if this long-term trend offset is not considered, it is impossible to accurately capture the long-term changes in photovoltaic power generation, resulting in the systematic deviation of the predicted value from the actual value. At the same time, if the periodic seasonal offset is not considered, the traditional method may not be able to correctly identify the periodic changes in the data, thus affecting the prediction accuracy of a specific season or time point.

[0004] Therefore, how to propose a photovoltaic output data prediction method and system for time-series production simulation that takes into account the medium- and long-term trend offset and periodic seasonal offset of photovoltaic power generation to improve the prediction accuracy of photovoltaic output data is an urgent problem to be solved by technical personnel in this field. Summary of the invention

[0005] In order to solve the problems existing in the prior art, the present invention proposes a photovoltaic output data prediction method and system for time-series production simulation.

[0006] The technical solution of the present invention is as follows:

[0007] A photovoltaic output data prediction method for time series production simulation, comprising:

[0008] Detrending and deseasonalizing historical PV output data;

[0009] Based on the historical PV output data after detrending and deseasonalizing, the PV output data prediction model is trained;

[0010] The trained photovoltaic output data prediction model is used to predict photovoltaic output data.

[0011] Furthermore, the method for obtaining the historical photovoltaic output data is as follows:

[0012] A historical photovoltaic output sequence is collected, and the historical photovoltaic output sequence is preprocessed with respect to missing values ​​and abnormal values, and the preprocessing result is used as the historical photovoltaic output data.

[0013] Furthermore, the specific steps of detrending and deseasonalizing the historical photovoltaic output data include:

[0014] Detrending the historical photovoltaic output data, determining and removing trend components;

[0015] The de-trended historical PV output data is de-seasonalized to determine and remove the seasonal components.

[0016] Furthermore, the specific steps of the detrending process include:

[0017] Based on the moving average method, the trend component in the historical photovoltaic output data is determined; wherein the expression of the moving average method is:

[0018]

[0019] Where, T t is the trend component in the historical photovoltaic output data; N is the window size, N = 2k + 1, k is the window boundary; y t+i is the latest observation value, i is the data subscript in the window; subscript t is time, t=1,2,3…,T, T is the total time;

[0020] The expression of the historical photovoltaic output data after detrending is:

[0021] y′ t =y t -T t

[0022] In the formula, y' t is the historical photovoltaic output data after detrending; t It is the historical photovoltaic output data.

[0023] Furthermore, the specific steps of the deseasonalization process include:

[0024] Based on the seasonal decomposition method, the seasonal components in the detrended historical photovoltaic output data are obtained and removed. The expression is:

[0025] ”'

[0026] y t =y t -St

[0027] In the formula, y' t ' is the historical photovoltaic output data after de-seasonalization; y' t is the historical photovoltaic output data after detrending, y' t =S t +R t , R t is the random part of the historical photovoltaic output data after detrending; S t A seasonal ingredient.

[0028] Furthermore, the seasonal decomposition method is an STL decomposition method.

[0029] Furthermore, the specific steps of training the photovoltaic output data prediction model based on the historical photovoltaic output data after detrending and deseasonalizing include:

[0030] The historical PV output data after detrending and deseasonalizing are divided into training set and test set, and then normalized;

[0031]

[0032] In the formula, is the historical photovoltaic output data after normalization; y' t ' is the de-seasonalized historical PV output data; μ and σ are the mean and standard deviation of the de-trended and de-seasonalized data, respectively;

[0033] Based on the long short-term memory network model, a photovoltaic output data prediction model is constructed and the photovoltaic output data prediction model is trained;

[0034]

[0035] In the formula, It is the historical photovoltaic output data after normalization; are respectively the normalized historical photovoltaic output data sequences tn, t-n+1, ​​t-1; LSTM(*) is the long short-term memory network model;

[0036] Use the trained photovoltaic output data prediction model to make predictions and denormalize the prediction results:

[0037]

[0038] In the formula, To de-seasonalize historical PV output data;

[0039] Use evaluation indicators such as mean square error (MSE) to evaluate the prediction effect of the photovoltaic output data prediction model:

[0040]

[0041] Where N is the data length; is the predicted de-seasonal historical photovoltaic output data; y' t ' +i is the corresponding de-seasonal historical photovoltaic output data.

[0042] A photovoltaic output data prediction system for time-series production simulation, comprising a data processing module, a model training module and a prediction module;

[0043] The data processing module is used to perform detrending and deseasonalizing processing on the historical photovoltaic output data;

[0044] The model training module is used to train the photovoltaic output data prediction model based on the historical photovoltaic output data after detrending and deseasonalizing;

[0045] The prediction module is used to predict photovoltaic output data using the trained photovoltaic output data prediction model.

[0046] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute any of the methods described above.

[0047] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of any of the above methods when executed by a processor.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] The present invention proposes a photovoltaic output data prediction method and system for time series production simulation, which performs detrending and deseasonalizing processing on historical photovoltaic output data, and then trains a photovoltaic output data prediction model based on the detrending and deseasonalizing historical photovoltaic output data; finally, the trained photovoltaic output data prediction model is used to predict photovoltaic output data. The method takes into account the impact of the medium- and long-term trend deviation and periodic seasonal deviation of photovoltaic output on photovoltaic output, and performs detrending and deseasonalizing processing on historical photovoltaic output data, thereby more accurately simulating and predicting the randomness and volatility of photovoltaic output, making the photovoltaic output (sequence) tend to be stable, thereby improving the accuracy and reliability of new energy power generation prediction and consumption capacity assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Flow chart of a photovoltaic output data prediction method for time-series production simulation in an embodiment. DETAILED DESCRIPTION

[0051] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0052] Embodiment 1:

[0053] A photovoltaic output data prediction method for time series production simulation of the present invention comprises:

[0054] S1. Detrending and deseasonalizing historical photovoltaic output data;

[0055] S2. Training a photovoltaic output data prediction model based on the historical photovoltaic output data after detrending and deseasonalizing;

[0056] S3. Use the trained photovoltaic output data prediction model to predict photovoltaic output data.

[0057] Embodiment 2:

[0058] This embodiment is further designed on the basis of the first embodiment in that: the method for obtaining the historical photovoltaic output data in this embodiment is as follows:

[0059] The historical photovoltaic output series is collected, and the historical photovoltaic output series is preprocessed with respect to missing values ​​and abnormal values, and the preprocessing results are used as the historical photovoltaic output data.

[0060] Embodiment three:

[0061] This embodiment is further designed on the basis of the second embodiment in that: in this embodiment, the specific steps of detrending and deseasonalizing the historical photovoltaic output data include:

[0062] Detrending historical PV output data, identifying and removing trend components;

[0063] The de-trended historical PV output data is de-seasonalized to determine and remove the seasonal components.

[0064] Embodiment 4:

[0065] This embodiment is further designed on the basis of the third embodiment in that the specific steps of detrending processing in this embodiment include:

[0066] Based on the moving average method, the trend component in the historical photovoltaic output data is determined; (the local weighted regression method (LOWESS) can also be used to determine the trend component.) Among them, the expression of the moving average method is:

[0067]

[0068] Where, T t is the trend component in the historical photovoltaic output data; N is the window size, N = 2k + 1, k is the window boundary; y t+i is the latest observation value, i is the data subscript in the window; subscript t is time, t=1,2,3…,T, T is the total time;

[0069] The expression of the historical photovoltaic output data after detrending is:

[0070] y′ t =y t -T t

[0071] In the formula, y' t is the historical photovoltaic output data after detrending; t It is the historical photovoltaic output data.

[0072] Embodiment five:

[0073] This embodiment is further designed on the basis of the fourth embodiment in that the specific steps of deseasonalization processing in this embodiment include:

[0074] Based on the seasonal decomposition method, the seasonal components in the detrended historical photovoltaic output data are obtained and removed. The expression is:

[0075] y″ t =y′ t -S t

[0076] In the formula, y' t ' is the historical photovoltaic output data after de-seasonalization; y' t is the historical photovoltaic output data after detrending, y' t =S t +R t , R t is the random part of the historical photovoltaic output data after detrending; S t A seasonal ingredient.

[0077] Embodiment six:

[0078] This embodiment is further designed on the basis of the fifth embodiment in that: the seasonal decomposition method in this embodiment is the STL decomposition method.

[0079] Embodiment seven:

[0080] This embodiment is further designed on the basis of the sixth embodiment in that: in this embodiment, based on the historical photovoltaic output data after detrending and deseasonalizing, the specific steps of training the photovoltaic output data prediction model include:

[0081] The historical PV output data after detrending and deseasonalizing are divided into training set and test set, and then normalized;

[0082]

[0083] In the formula, is the historical photovoltaic output data after normalization; y' t ' is the de-seasonalized historical PV output data; μ and σ are the mean and standard deviation of the de-trended and de-seasonalized data, respectively;

[0084] Based on the long short-term memory network model (LSTM), a photovoltaic output data prediction model is constructed and the photovoltaic output data prediction model is trained; (the photovoltaic output data prediction model can also be a recurrent neural network model (RNN), etc.)

[0085]

[0086] In the formula, It is the historical photovoltaic output data after normalization; are respectively the normalized historical photovoltaic output data sequences tn, t-n+1, ​​t-1; LSTM(*) is the long short-term memory network model;

[0087] Use the trained photovoltaic output data prediction model to make predictions and denormalize the prediction results:

[0088]

[0089] In the formula, To de-seasonalize historical PV output data;

[0090] Use evaluation indicators such as mean square error (MSE) to evaluate the prediction effect of the photovoltaic output data prediction model:

[0091]

[0092] Where N is the data length; is the predicted de-seasonal historical photovoltaic output data; y' t ' +i is the corresponding de-seasonal historical photovoltaic output data.

[0093] Embodiment eight:

[0094] A photovoltaic output data prediction system for time-series production simulation of the present invention comprises a data processing module, a model training module and a prediction module;

[0095] Data processing module, used to de-trend and de-seasonalize historical photovoltaic output data;

[0096] A model training module is used to train a photovoltaic output data prediction model based on historical photovoltaic output data after detrending and deseasonalizing;

[0097] The prediction module is used to predict photovoltaic output data using the trained photovoltaic output data prediction model.

[0098] Embodiment nine:

[0099] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to call and run the computer program stored in the memory to execute any of the methods described above.

[0100] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0101] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A photovoltaic output data prediction method for time series production simulation, characterized in that: include: Detrending and deseasonalizing historical PV output data; Based on the historical PV output data after detrending and deseasonalizing, the PV output data prediction model is trained; The trained photovoltaic output data prediction model is used to predict photovoltaic output data.

2. The photovoltaic output data prediction method for time series production simulation according to claim 1, characterized in that: The method for obtaining the historical photovoltaic output data is as follows: A historical photovoltaic output sequence is collected, and the historical photovoltaic output sequence is preprocessed with respect to missing values ​​and abnormal values, and the preprocessing result is used as the historical photovoltaic output data.

3. The photovoltaic output data prediction method for time series production simulation according to claim 1, characterized in that: The specific steps for detrending and deseasonalizing historical PV output data include: Detrending the historical photovoltaic output data, determining and removing trend components; The de-trended historical PV output data is de-seasonalized to determine and remove the seasonal components.

4. The photovoltaic output data prediction method for time series production simulation according to claim 3 is characterized in that: The specific steps of the detrending process include: Based on the moving average method, the trend component in the historical photovoltaic output data is determined; wherein the expression of the moving average method is: Where, T t is the trend component in the historical photovoltaic output data; N is the window size, N = 2k + 1, k is the window boundary; y t+i is the latest observation value, i is the data subscript in the window; subscript t is time, t=1,2,3…,T, T is the total time; The expression of the historical photovoltaic output data after detrending is: and' t =and t -T t In the formula, y' t is the historical photovoltaic output data after detrending; t It is the historical photovoltaic output data.

5. The photovoltaic output data prediction method for time series production simulation according to claim 4, characterized in that: The specific steps of the deseasonalization process include: Based on the seasonal decomposition method, the seasonal components in the detrended historical photovoltaic output data are obtained and removed. The expression is: and" t and' t -S t In the formula, y” t is the historical photovoltaic output data after de-seasonalization; y' t is the historical photovoltaic output data after detrending, y' t =S t +R t , R t is the random part of the historical photovoltaic output data after detrending; S t A seasonal ingredient.

6. The photovoltaic output data prediction method for time series production simulation according to claim 5, characterized in that: The seasonal decomposition method is the STL decomposition method.

7. The photovoltaic output data prediction method for time series production simulation according to claim 5, characterized in that: The specific steps of training the photovoltaic output data prediction model based on the historical photovoltaic output data after detrending and deseasonalizing include: The historical PV output data after detrending and deseasonalizing are divided into training set and test set, and then normalized; In the formula, is the normalized historical photovoltaic output data; y” t is the de-seasonalized historical PV output data; μ and σ are the mean and standard deviation of the de-trended and de-seasonalized data, respectively; Based on the long short-term memory network model, a photovoltaic output data prediction model is constructed and the photovoltaic output data prediction model is trained; In the formula, It is the historical photovoltaic output data after normalization; are respectively the normalized historical photovoltaic output data sequences tn, t-n+1, ​​t-1; LSTM(*) is the long short-term memory network model; Use the trained photovoltaic output data prediction model to make predictions and denormalize the prediction results: In the formula, To de-seasonalize historical PV output data; Use evaluation indicators such as mean square error (MSE) to evaluate the prediction effect of the photovoltaic output data prediction model: Where N is the data length; is the predicted de-seasonal historical photovoltaic output data; y” t+i is the corresponding de-seasonal historical photovoltaic output data.

8. A photovoltaic output data prediction system for time series production simulation, characterized in that: Includes data processing module, model training module and prediction module; The data processing module is used to perform detrending and deseasonalizing processing on the historical photovoltaic output data; The model training module is used to train the photovoltaic output data prediction model based on the historical photovoltaic output data after detrending and deseasonalizing; The prediction module is used to predict photovoltaic output data using the trained photovoltaic output data prediction model.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is configured to call and run the computer program stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method as claimed in any one of claims 1 to 7 are implemented.

Images