Photovoltaic power prediction method and system based on multi-model fusion, and electronic equipment

Abstract

The invention discloses a photovoltaic power prediction method and system based on multi-model fusion and electronic equipment, and relates to the technical field of photovoltaic power prediction. Each preset power prediction model is utilized to obtain a photovoltaic power prediction result of the preset photovoltaic power station in a second preset time period, and the second preset time period is located after the first preset time period in the time sequence; and fusing all the photovoltaic power prediction results based on the weight allocated to the photovoltaic power prediction result obtained through each preset power prediction model to obtain a final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period. According to the method, more accurate photovoltaic power prediction can be provided for power grid dispatching, and stable operation of a power grid and efficient energy distribution are facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power prediction, and particularly relates to a photovoltaic power prediction method and system based on multi-model fusion and an electronic device. BACKGROUND

[0002] In the current energy field, photovoltaic power prediction is crucial for power grid scheduling. However, existing power prediction models often have many problems. Most models only use a single power prediction model for prediction, which is difficult to comprehensively and accurately reflect the actual operation of the photovoltaic system in complex and variable environments, resulting in inaccurate prediction results. In addition, it may also be disturbed by factors such as input data quality and unreasonable model parameter settings. These factors together make photovoltaic power prediction often unable to achieve ideal accuracy, making it difficult to meet the strict requirements of power grid scheduling for accuracy, and thus affecting the stable operation of the power grid and the effective allocation of energy. SUMMARY

[0003] The technical problem to be solved by the present application is the deficiency of the prior art. Specifically, a photovoltaic power prediction method and system based on multi-model fusion and an electronic device are provided, as follows: 1) In a first aspect, the present application provides a photovoltaic power prediction method based on multi-model fusion, and the specific technical solution is as follows: based on the photovoltaic power correlation data of a preset photovoltaic power station in a first preset time period, and using each preset power prediction model to obtain the photovoltaic power prediction result of the preset photovoltaic power station in a second preset time period, the second preset time period is located after the first preset time period in terms of time sequence. Based on the weight assigned to the photovoltaic power prediction result obtained by each preset power prediction model, all photovoltaic power prediction results are fused to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

[0004] The photovoltaic power prediction method based on multi-model fusion provided by the present application has the following beneficial effects: The existing photovoltaic power prediction model mostly uses a single model for prediction, which is affected by data quality, environmental factors, etc., and is difficult to accurately predict, resulting in inaccurate results and failing to meet the requirements of power grid dispatching. However, the scheme is based on the photovoltaic power correlation data of the preset photovoltaic power station in the first preset time period, obtains the photovoltaic power prediction results in the second preset time period through multiple preset power prediction models, and then fuses the prediction results based on the weights of the prediction results, effectively improving the prediction accuracy and meeting the requirements of power grid dispatching. The multi-model fusion can comprehensively utilize the advantages of each model, reduce the error of a single model, and improve the reliability. At the same time, the weight distribution can be dynamically adjusted according to the model performance and data characteristics, enhancing the prediction flexibility and adaptability, and better coping with complex environments and data changes. The scheme provides more accurate photovoltaic power prediction for power grid dispatching, and helps the stable operation of power grid and the efficient distribution of energy.

[0005] Based on the above scheme, the photovoltaic power prediction method based on multi-model fusion of the present application can be further improved as follows.

[0006] Further, it further comprises: According to the used years of the photovoltaic components of the preset photovoltaic power station and the aging curve corresponding to the photovoltaic components, the aging rate is calculated; and the aging rate is used to correct the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

[0007] The beneficial effects of the above further scheme are: By considering the used years of the photovoltaic components and the corresponding aging curve to calculate the aging rate, the performance degradation of the photovoltaic components in the actual operation process can be more accurately reflected. By introducing the aging rate into the correction link of the final photovoltaic power prediction result, the error caused by the traditional prediction model not fully considering the component aging factor can be effectively made up, thereby improving the accuracy and reliability of the prediction result. This correction method helps to improve the accuracy and flexibility of power grid dispatching. Since power grid dispatching needs to be reasonably arranged based on the prediction result of photovoltaic power, more accurate prediction helps to better balance the supply and demand relationship of power grid, improve the overall operation efficiency, reduce the problems such as unstable power supply caused by prediction deviation, and provide strong support for the stable operation of power grid. Considering that the aging of photovoltaic components is an inevitable process, the correction method based on the aging curve can make the photovoltaic power prediction more consistent with the actual operation law, and provide more valuable reference basis for the long-term optimization operation and maintenance of photovoltaic power stations.

[0008] Further, the mathematical expression of the aging curve is: Wherein, x represents: used years, and a represents: aging rate.

[0009] The beneficial effect of the above further scheme is that the degradation characteristics of the photovoltaic module at different service life stages are more accurately characterized in the form of a piecewise function, making the calculation of the degradation rate more targeted and scientific. Applying the degradation curve to the correction of the photovoltaic power prediction result can more accurately reflect the performance degradation of the photovoltaic module in actual operation, further improving the accuracy of photovoltaic power prediction, better meeting the needs of power grid dispatching, and providing a more reliable reference for the long-term optimization and maintenance of photovoltaic power stations.

[0010] Further, the obtaining process of each preset power prediction model comprises: When the number of preset power prediction models is 3, the time series Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are trained respectively to obtain 3 preset power prediction models.

[0011] The beneficial effect of the above further scheme is that the three models have their own advantages. The time series Informer model performs well in processing long time series data, and can effectively capture the long-term trend and seasonal changes of photovoltaic power. The recurrent neural network LSTM model is good at processing sequence data with time dependence, and can better understand the short-term fluctuations and dynamic characteristics of photovoltaic power. The support vector regression SVR model has unique advantages in processing nonlinear relationships, and can model complex nonlinear patterns in photovoltaic power data. By combining the three models, their respective characteristics can be fully utilized to achieve complementary advantages, thereby improving the accuracy of the prediction results. Secondly, the prediction results based on the three models are fused, which can effectively reduce the risk of overfitting or underfitting of a single model, and improve the stability and reliability of the prediction results. Finally, this multi-model fusion method can better adapt to different scenarios and data characteristics, providing a more comprehensive and accurate solution for photovoltaic power prediction, further meeting the requirements of power grid dispatching for photovoltaic power prediction accuracy, and helping to improve the stability and economy of power grid operation.

[0012] Further, based on the weights assigned to the photovoltaic power prediction results obtained by each preset power prediction model, all photovoltaic power prediction results are fused to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, comprising: The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is calculated by the weight fusion formula, and the weight fusion formula is: y = a x model_informer + b x model_lstm + c x model_svr wherein y represents the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, a represents the weight assigned to the photovoltaic power prediction result obtained through the first preset power prediction model, b represents the weight assigned to the photovoltaic power prediction result obtained through the second preset power prediction model, c represents the weight assigned to the photovoltaic power prediction result obtained through the third preset power prediction model, a+b+c=1, the first preset power prediction model is a preset power prediction model obtained after training a time series Informer model, the second preset power prediction model is a preset power prediction model obtained after training a recurrent neural network LSTM model, the third preset power prediction model is a preset power prediction model obtained after training a support vector regression SVR model, model_informer represents the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the first preset power prediction model; model_lstm represents the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the second preset power prediction model; and model_svr represents the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the third preset power prediction model.

[0013] The above further scheme has the beneficial effect that the prediction results of different models are integrated through the weight fusion formula, the advantages of the time series Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are fully utilized, the limitations of inaccurate single model prediction are overcome, and the accuracy of photovoltaic power prediction is improved. Reasonable distribution of weights can highlight the better model and make the prediction result more reliable. This fusion method also enhances the generalization ability of the model, so that it can better adapt to complex and variable photovoltaic power prediction scenarios, meet the requirements of high accuracy and stability of prediction results for power grid dispatching, and help the stable operation of power grid and efficient management of energy.

[0014] 2) In a second aspect, the present application also provides a photovoltaic power prediction system based on multi-model fusion, and the specific technical solutions are as follows: It comprises a photovoltaic power prediction module and a weight fusion module. The photovoltaic power prediction module is used to obtain the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period based on the photovoltaic power correlation data of the preset photovoltaic power station in the first preset time period and by using each preset power prediction model, and the second preset time period is located after the first preset time period in time sequence. The weight fusion module is configured to fuse all the photovoltaic power prediction results based on the weights assigned to the photovoltaic power prediction results obtained by each preset power prediction model to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

[0015] Based on the above scheme, the photovoltaic power prediction system based on multi-model fusion can be further improved as follows.

[0016] Further, the system further comprises an aging rate calculation module and a correction module. The aging rate calculation module is configured to calculate the aging rate according to the used years of the photovoltaic module of the preset photovoltaic power station and the aging curve corresponding to the photovoltaic module. The correction module is configured to correct the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period by using the aging rate.

[0017] Further, the mathematical expression of the aging curve is as follows: Wherein, x represents the used years, and a represents the aging rate.

[0018] Further, the system further comprises a model acquisition module, and the model acquisition module is configured to: When the number of preset power prediction models is 3, the time sequence Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are trained respectively to obtain three preset power prediction models.

[0019] Further, the weight fusion module is specifically configured to: The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is calculated by using the weight fusion formula, and the weight fusion formula is as follows: y=a*model_informer+b*model_lstm+c*model_svr wherein, y represents: a final photovoltaic power prediction result of a preset photovoltaic power station in a second preset time period, a represents: a weight allocated to a photovoltaic power prediction result obtained through a first preset power prediction model, b represents: a weight allocated to a photovoltaic power prediction result obtained through a second preset power prediction model, c represents: a weight allocated to a photovoltaic power prediction result obtained through a third preset power prediction model, a+b+c=1, the first preset power prediction model is: a preset power prediction model obtained after training a time series Informer model, the second preset power prediction model is: a preset power prediction model obtained after training a recurrent neural network LSTM model, the third preset power prediction model is: a preset power prediction model obtained after training a support vector regression SVR model, model_informer represents: a photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the first preset power prediction model; model_lstm represents: a photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the second preset power prediction model; model_svr represents: a photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the third preset power prediction model.

[0020] 3) In a third aspect, the present application also provides an electronic device, the electronic device comprising a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the electronic device to implement any of the above photovoltaic power prediction methods based on multi-model fusion.

[0021] 4) In a fourth aspect, the present application also provides a computer readable storage medium, the computer readable storage medium storing a computer program, the computer program being executed by a processor to implement any of the above photovoltaic power prediction methods based on multi-model fusion.

[0022] It should be noted that the technical solutions of the second to fourth aspects of the present application and the corresponding possible implementation manners have the beneficial effects described above for the first aspect and the corresponding possible implementation manners, which will not be described here again. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the description of the embodiments of the present application will be briefly introduced as follows: Figure 1 A flowchart of a photovoltaic power prediction method based on multi-model fusion according to an embodiment of the present application; Figure 2A structure schematic diagram of a photovoltaic power prediction system based on multi-model fusion according to an embodiment of the present application; Figure 3 A structure schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The principles and features of the present application are described below, and the examples are only used to explain the present application, and are not used to limit the scope of the present application.

[0025] The technical solutions of the present application and how the technical solutions solve the above technical problems are described in detail below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes can not be described again in some embodiments. The embodiments of the present application will be described below with reference to the drawings.

[0026] As shown in the drawings, Figure 1 A photovoltaic power prediction method based on multi-model fusion according to an embodiment of the present application includes the following steps: S1, based on the photovoltaic power correlation data of the preset photovoltaic power station in the first preset time period, and using each preset power prediction model to obtain the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the second preset time period is after the first preset time period in time sequence; Wherein, the second preset time period is after the first preset time period, specifically: the starting time of the second preset time period is exactly the end time of the first preset time period, that is, the two time periods are closely connected, and there is no time interval between them, or the starting time of the second preset time period is not the end time of the first preset time period, that is, there may be a time interval between the two time periods.

[0027] Wherein, the length of the first preset time period and the second preset time period can be the same or different, which can be determined according to the specific application scene and demand.

[0028] S2, based on the weight allocated to the photovoltaic power prediction result obtained by each preset power prediction model, all photovoltaic power prediction results are fused to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

[0029] Optionally, in the above technical solution, it further includes: S3, according to the used years of the photovoltaic module of the preset photovoltaic power station and the aging curve corresponding to the photovoltaic module, the aging rate is calculated; Wherein, the mathematical expression of the aging curve is: Wherein, x represents: the number of years of use, a represents: the aging rate.

[0030] S4, using the aging rate to correct the final photovoltaic power prediction result in the second preset time period to obtain a corrected final photovoltaic power prediction result.

[0031] Wherein, the number of preset power prediction models can be flexibly set according to actual conditions, and is trained for different models. Specifically, when the number of preset power prediction models is 3, the time series Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are trained respectively to obtain 3 preset power prediction models; when the number of preset power prediction models is 6, the time series Informer model, the recurrent neural network LSTM model and the support vector regression SVR model, the recurrent neural network GRU model, the random forest model and the multilayer perception machine model (MLP) can be trained respectively, the recurrent neural network GRU model is specially used to process sequence data, solves the gradient disappearance problem in the long sequence training process, and is suitable for photovoltaic power time series prediction. The random forest model can process multi-dimensional data features, and has good fitting effect on complex nonlinear relationship in photovoltaic power prediction. The multilayer perception machine model can learn the input-output mapping relationship, and can be used as a basic neural network model for photovoltaic power prediction.

[0032] Take "training the time series Informer model, the recurrent neural network LSTM model and the support vector regression SVR model respectively to obtain 3 preset power prediction models" as an example for description, specifically: 1) Construct a data set: According to the preset sampling frequency, the photovoltaic power correlation data is extracted from the historical meteorological data and photovoltaic power data of the photovoltaic power station, the photovoltaic power correlation data includes: irradiance, wind speed, wind direction, temperature, pressure, humidity, instantaneous power (i.e. actual photovoltaic power) at multiple collection times, the preset sampling frequency is: 15 minutes / time, that is, irradiance, wind speed, wind direction, temperature, pressure, humidity, instantaneous power are extracted from the historical meteorological data and photovoltaic power data of the photovoltaic power station every 15 minutes, irradiance, wind speed, wind direction, temperature, pressure, humidity, instantaneous power at each collection time form an original data sample, each original data sample is preprocessed to obtain multiple data samples, the preprocessing process is as follows: ① delete the original data samples with all the photovoltaic power correlation data being 0 or empty, and remove the repeated original data samples; ② perform missing value processing on the original data samples with negative values in the irradiance and instantaneous power, specifically, under normal circumstances, the irradiance is the energy intensity of the solar radiation to the surface of the photovoltaic device, and the instantaneous power is the electric power generated by the photovoltaic device at a certain moment, which should be non-negative in theory. If negative values appear, it is usually caused by unreasonable data due to sensor failure, data acquisition error or data transmission error, which cannot reflect the real physical state, and therefore missing value processing is performed.

[0033] ③ perform missing value processing on the records with abnormally large or small values, specifically, the reasonable fluctuation range of the irradiance and instantaneous power can be determined according to the geographical environment, equipment capacity and historical data of the photovoltaic power station, etc. If the data exceeds the upper limit or is lower than the lower limit of the normal range, it may be caused by abnormal sensor performance, equipment failure or data processing error. Such data is not reliable, and if used for analysis and decision-making, it will be misleading, and therefore missing value processing is performed.

[0034] Among them, the missing value processing method includes deleting records containing missing values, filling missing values with statistical quantities such as mean or median, and filling using interpolation method, etc.

[0035] Each data sample is standardized, and combined with the collection time, the standardized data sample is obtained, and all the standardized data samples form a data set.

[0036] Among them, the Z-Score standardization method can be used to standardize the irradiance, wind speed, wind direction, temperature, pressure, humidity and instantaneous power, and the expression of the Z-Score standardization method is: Among them, μ is the mean of the total data, i.e. the mean of the irradiance, wind speed, temperature, pressure, humidity or instantaneous power in all data samples, σ is the standard deviation of the total data, i.e. the standard deviation of the irradiance, wind speed, temperature, pressure, humidity or instantaneous power in all data samples, x' is the observation value of the individual, i.e. the value of the irradiance, wind speed, temperature, pressure, humidity or instantaneous power in each data sample. z represents: standardized irradiance, standardized wind speed, standardized temperature, standardized pressure, standardized humidity or standardized instantaneous power.

[0037] Among them, the wind direction is usually represented by an angle, ranging from 0° to 360°, and the process of standardizing the wind direction is as follows: ① use the trigonometric function conversion method to standardize the wind direction, specifically: The wind direction angle θ is converted into two components, namely the sine and cosine components. This method can convert the angle of the wind direction into a coordinate point in two-dimensional space, so that the model can capture the periodic characteristics of the wind direction. For example, the wind direction of 0° and 360° corresponds to the coordinate point (1, 0) after conversion, and 180° corresponds to the coordinate point (-1, 0).

[0038] ②Divide the wind direction into multiple intervals (for example, 8 intervals), each interval corresponds to a specific direction (such as northeast, east, southeast, etc.), and then convert the wind direction into the corresponding direction code, for example: 0°-45°: Northeast (code 0), 45°-90°: East (code 1), this method converts the continuous wind direction angle into discrete direction categories, simplifying the complexity of the data.

[0039] ③Use the Z-Score standardization method to standardize the angle of the wind direction.

[0040] The data format of the standardized data sample is: [collection time, standardized irradiance, standardized wind speed, standardized wind direction, standardized temperature, standardized pressure, standardized humidity, standardized instantaneous power], and the data format of the standardized data sample can also be set according to the actual situation, which will not be elaborated here.

[0041] 2) Model training: ①Build a time series Informer model, train the time series Informer model based on the data set, and obtain the first preset power prediction model. Among them, the time series Informer model regards power prediction as a periodic time series. Photovoltaic power prediction and time have a close relationship, and photovoltaic power generation power changes periodically with time (solar irradiance is received in the morning to start power generation, the power generation power is the largest at noon, and gradually decreases to 0 in the evening, and there is no power generation at night). Fully exploring time information can improve the accuracy of power prediction. The time series Informer model is based on the input of time data, which represents the time data after position coding, reduces the computational complexity through the multi-head self-attention and self-attention distillation mechanism of the encoder, so that the time series Informer model can learn the time dependence of long time series data, and finally form the intermediate representation form of data information through the interaction of multi-head attention and features. Then input into the decoder of the time series Informer model, the decoder fuses the meteorological data corresponding to the power to be predicted, the decoder of the time series Informer model adopts the occlusion attention mechanism, and finally directly outputs the photovoltaic power prediction result through the full connection layer of the decoder. The objective function of the time series Informer model is the mean square error function, which is defined as: wherein MSE represents a mean square error function, l ij and respectively represent a true value (instantaneous power, that is, actual photovoltaic power) and a predicted value (predicted value of instantaneous power) of the jth acquisition moment in the ith input sequence (standardized data sample), L y is the length of the sequence to be tested (input sequence); the training of the time series Informer model is realized by minimizing MSE.

[0042] ② constructing a recurrent neural network LSTM model, training the recurrent neural network LSTM model based on the data set to obtain a second preset power prediction model, specifically: data feature extraction is performed on the data set, and then the extracted features are input to the shared neurons of the recurrent neural network LSTM model for training. A shared learning layer is established, and the output of the recurrent neural network LSTM model is input to the shared learning layer for training, so as to extract the periodicity of the generated power for power prediction.

[0043] Among them, the data feature extraction on the data set refers to the process of selecting a feature combination with important value for the prediction target (instantaneous power) from the original data or deriving a new feature. Specifically, it includes: screening features highly related to instantaneous power from original data samples, such as irradiance, temperature, etc.; time features can also be extracted, such as time periods corresponding to acquisition moments, seasons, because different time periods have different solar radiation intensities, which have different effects on instantaneous power. In addition, meteorological features such as comprehensive wind speed and wind direction can be fused to form more comprehensive meteorological indicators. Through feature extraction, the data dimension can be reduced, redundant information can be removed, the features input to the shared neurons of the recurrent neural network LSTM model are more representative, and the model training efficiency and prediction accuracy are improved.

[0044] In another embodiment, the recurrent neural network LSTM model is trained using the data set, specifically: first, the data set is divided into multiple time sequence segments, each segment containing a plurality of standardized data samples. Each time step data includes the acquisition moment, the standardized irradiance, the wind speed, the wind direction, the temperature, the pressure, the humidity and the instantaneous power. Then, input these time sequence segments as input and instantaneous power as output label into the LSTM model. The LSTM model processes the input data of each time step through its internal recurrent neurons, capturing the time dependence and feature patterns in the data. During the training process, the model continuously adjusts the weights to minimize the error between the predicted value and the actual value. Through multiple iterations of training, the LSTM model can learn the hidden rules in the data, thereby achieving accurate prediction of instantaneous power.

[0045] ③ Construct a support vector regression (SVR) model, train the SVR model based on the dataset, and obtain a third preset power prediction model. Specifically: The support vector regression (SVR) model is trained based on the dataset. Specifically, the dataset is first divided into a training set, a validation set, and a test set. Then, a suitable kernel function (such as a linear kernel, a polynomial kernel, a radial basis function kernel, etc.) is determined, because the kernel function determines the mapping of data in high-dimensional space and affects the model's fitting ability for nonlinear relationships. Key hyperparameters are set, including the regularization parameter (used to control the model complexity and prevent overfitting), the kernel function parameter (such as the width parameter in the radial basis function kernel), etc. The training set is used to train the SVR model, and the loss function is optimized to make the model's predicted values on the training set as close as possible to the actual values. At the same time, the validation set is used to verify the model and adjust the hyperparameters to find the optimal parameter combination that maximizes the model's performance. During the training process, the model is iteratively optimized to learn the nonlinear relationships and internal rules in the data. Finally, the test set is used to evaluate the trained SVR model and test its generalization ability on unseen data, ensuring that the model can accurately predict the instantaneous power of photovoltaic equipment in actual applications, providing strong support for the operation monitoring and power prediction of photovoltaic power stations, and improving the management efficiency and power generation benefits of the power station.

[0046] Optionally, in S1, all photovoltaic power prediction results are fused based on the weights assigned to the photovoltaic power prediction results obtained by each preset power prediction model to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, including: The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is calculated by the weight fusion formula, which is: y = a x model_informer + b x model_lstm + c x model_svr Wherein, y represents: the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, a represents: the weight allocated to the photovoltaic power prediction result obtained by the first preset power prediction model, b represents: the weight allocated to the photovoltaic power prediction result obtained by the second preset power prediction model, c represents: the weight allocated to the photovoltaic power prediction result obtained by the third preset power prediction model, a+b+c=1, the first preset power prediction model is: a preset power prediction model obtained after training a time series Informer model, the second preset power prediction model is: a preset power prediction model obtained after training a recurrent neural network LSTM model, the third preset power prediction model is: a preset power prediction model obtained after training a support vector regression SVR model, model_informer represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the first preset power prediction model; model_lstm represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the second preset power prediction model; model_svr represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the third preset power prediction model.

[0047] Wherein, the implementation process of determining the values of a, b and c is as follows: S10, obtaining historical actual power data of the preset photovoltaic power station in a first preset time period; S11, inputting historical input data (photovoltaic power related data) in the first preset time period into the first preset power prediction model, the second preset power prediction model and the third preset power prediction model respectively, which are pre-trained, to obtain historical power prediction results of each model in the first preset time period; S12, calculating the prediction error indicators between the historical power prediction results and the historical actual power data for the first preset power prediction model, the second preset power prediction model and the third preset power prediction model respectively; S13, calculating the relative prediction accuracy corresponding to each model based on the prediction error indicators of each model; S14, determining and normalizing the corresponding weight coefficients according to the relative prediction accuracy of the first preset power prediction model, the second preset power prediction model and the third preset power prediction model respectively, that is, obtaining the values of a, b and c.

[0048] In S10 to S14, the historical prediction accuracy of each model is evaluated by quantification, the optimal weight is dynamically allocated, the accuracy of the fused prediction result is significantly improved, specifically, the limitations of a single model are overcome, the advantages of the fusion model are fused, and the prediction error is reduced; the weight is dynamically adjusted based on the historical data to adapt to different weather and seasonal changes; the implementation process is simple, the calculation complexity is reduced; and finally the stability and reliability of the photovoltaic power prediction result are enhanced to provide more reliable decision basis for power grid dispatching.

[0049] Optionally, in the above technical solution, after obtaining the modified final photovoltaic power prediction result, it further comprises: S5, planning the power generation plan of the preset photovoltaic power station in advance according to the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period. Moreover, according to the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the charging and discharging strategy of the energy storage device can be optimized, and the energy utilization efficiency can be improved. At the same time, reliable basis is provided for power grid dispatching, ensuring the balance between supply and demand of power grid, and enhancing the stability of power grid.

[0050] Among them, according to the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the power generation plan of the preset photovoltaic power station is planned in advance, and the specific implementation process is as follows: S501, according to the dispatching instruction of the preset photovoltaic power station, the operation constraint condition of the preset photovoltaic power station and the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the planned output target value of the preset photovoltaic power station in each dispatching period in the second preset time period is determined, specifically: The output direction and range are determined by the dispatching instruction, the output is limited by combining the operation constraint condition, and the safety and reliability are ensured. Then, according to the photovoltaic power prediction result, the output of each dispatching period is reasonably distributed, and the planned output target value meeting all conditions is determined through optimization algorithm and the like.

[0051] Among them, the operation constraint condition mainly includes: the maximum output and minimum output limit of the photovoltaic power station equipment, such as the rated power of the inverter; the voltage stability requirement of the power station to avoid voltage out-of-limit caused by sudden change of output; and the access power change rate limit of the power grid to the power station to prevent excessive power fluctuation from affecting the stability of the power grid. These conditions ensure the safe, stable and efficient operation of the photovoltaic power station.

[0052] S502, based on the planned output target value, the active power and reactive power of the preset photovoltaic power station in the second preset time period are generated, specifically: The overall range of active power is determined according to the planned output target value. On this basis, the reactive power regulation capability of the photovoltaic power station inverter and other equipment is comprehensively considered, and the active and reactive power combination that meets the safe operation of the photovoltaic power station itself and the demand of the power grid is calculated through an optimization algorithm. At the same time, the distribution of reactive power also needs to be dynamically adjusted in combination with the voltage stability requirements of the power grid, so as to finally generate active and reactive power that meets the requirements.

[0053] S503, send the control instruction corresponding to the active power to the automatic generation control (AGC) system of the preset photovoltaic power station and control it, and send the control instruction corresponding to the reactive power to the automatic voltage control (AVC) system of the preset photovoltaic power station and control it.

[0054] S504, according to the corrected final photovoltaic power prediction result and the planned output target value, calculate the expected power generation and expected deviation range of the preset photovoltaic power station in the second preset time period; Among them, the time integral of the corrected final photovoltaic power prediction result in the second preset time period is carried out, that is, the predicted power of each dispatching period is multiplied by the time length of the period, and then the results of all periods are added to obtain the expected power generation.

[0055] Among them, the difference between the corrected predicted power and the planned output target value in each dispatching period is first calculated, and then the statistical indicators such as standard deviation of these differences in the second preset time period are calculated to determine the expected deviation range.

[0056] S505, based on the expected power generation and the expected deviation range, formulate the charging and discharging plan of the energy storage system or the standby power calling strategy of the preset photovoltaic power station.

[0057] Among them, the process of formulating the charging and discharging plan of the energy storage system of the preset photovoltaic power station is: comparing the cumulative power of the expected power generation and the planned output target value. If the photovoltaic power generation is greater than the planned target value, the excess power is used to charge the energy storage system, and the charging power is determined according to the difference and the characteristics of the energy storage system; otherwise, the energy storage system is discharged to supplement the insufficient part of the power generation, and the discharge power is distributed as needed.

[0058] Among them, the process of formulating the standby power calling strategy of the preset photovoltaic power station is: according to the expected deviation range, the fluctuation risk of power generation is evaluated. When the deviation is large, it may cause the power generation to be unable to meet the planned output, so the standby power is called in advance to fill the gap, and the calling power value is determined based on the upper and lower limits of the deviation range, so as to ensure the stable operation of the preset photovoltaic power station in the second preset time period and to meet the planned output target value as much as possible.

[0059] In S501-S505, the energy storage / backup strategy is formulated in combination with the predicted deviation, the light abandonment rate and the backup cost are reduced, the expected power generation and the deviation range are quantified, and the ability of the power grid to cope with fluctuations is enhanced. Finally, the grid connection compliance and economic benefits of the photovoltaic power station are improved, and high proportion of new energy consumption is supported.

[0060] In the method, the charging and discharging strategy of the energy storage device is optimized according to the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, and the specific implementation process is as follows: S506, obtaining the modified final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period and the price curve of the region where the photovoltaic power station grid connection point is located; S507, real-time monitoring the current state of charge (SOC) and charging and discharging power capability of the energy storage system; S508, constructing an optimization model with the maximum expected net income of the energy storage system in the second preset time period or the minimum expected total cost as the objective function, wherein the objective function at least considers the following factors: ① the value of the final photovoltaic power prediction result in each period of the second preset time period; ② the value of the price curve in each period of the second preset time period; ③ the charging and discharging efficiency, self-discharge rate, charging and discharging power limit and SOC safe operation range constraint of the energy storage system; ④ the SOC constraint of the energy storage system at the initial and ending time of the second preset time period.

[0061] S509, solving the optimization model to obtain the optimal charging and discharging power instruction sequence of the energy storage system in each period of the second preset time period; S510, generating the charging and discharging control instruction of the energy storage system in the second preset time period according to the optimal charging and discharging power instruction sequence and issuing for execution; S511, setting the dynamic backup capacity of the energy storage system or adjusting the robustness parameters of the charging and discharging control instruction based on the expected deviation range of the final photovoltaic power prediction result.

[0062] In S506-S511, the price peak-valley difference is actively hedged to improve the whole life cycle income of the energy storage; the SOC range and power limit are strictly constrained to avoid overcharging / overdischarging of the equipment; the backup capacity is dynamically reserved based on the prediction deviation to reduce the risk of strategy failure caused by prediction error. The economic value and system adjustment flexibility of the energy storage in the photovoltaic power station are significantly improved.

[0063] S6, according to the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the grid-connected power is adjusted in advance to avoid the impact of power fluctuation on the power grid. At the same time, according to the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the reactive power compensation device is reasonably configured to stabilize the grid voltage and ensure the safe operation of the power grid.

[0064] Among them, according to the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, the grid-connected power is adjusted in advance, and the specific implementation process is as follows: S601, based on the final photovoltaic power prediction result, the expected grid-connected power of the preset photovoltaic power station in each period of the second preset time period is predicted.

[0065] Since the grid-connected power of the photovoltaic power station will be affected by factors such as self-equipment loss and conversion efficiency, the predicted photovoltaic power needs to be corrected accordingly to calculate the actual grid-connected power. According to the corrected photovoltaic power prediction value, combined with the equipment parameters and operating characteristics of the photovoltaic power station, such as the conversion efficiency of the inverter, the expected grid-connected power of the preset photovoltaic power station in each period of the second preset time period is calculated. The formula can be expressed as: expected grid-connected power = corrected photovoltaic power prediction value x equipment conversion efficiency coefficient.

[0066] S602, according to the real-time dispatching instruction or grid-connected power limit constraint issued by the power grid dispatching mechanism, determine the allowed grid-connected power upper limit of the preset photovoltaic power station in the corresponding period; S603, compare the expected grid-connected power with the allowed grid-connected power upper limit: ① If the expected grid-connected power is lower than or equal to the allowed grid-connected power upper limit, generate a grid-connected power set value following the expected grid-connected power; ② If the expected grid-connected power is higher than the allowed grid-connected power upper limit, generate a grid-connected power set value with the allowed grid-connected power upper limit as the target; S604, according to the grid-connected power set value, combined with the real-time output of the preset photovoltaic power station, the expected deviation range of the final photovoltaic power prediction result and the power frequency / voltage fluctuation signal, and dynamically calculate the active power and reactive power of the photovoltaic inverter cluster, and control the photovoltaic inverter cluster according to the calculated active power and reactive power.

[0067] The calculation process of the active power is as follows: the grid-connected power set value is acquired, and the output of the preset photovoltaic power station is monitored in real time to acquire real-time output data; the expected deviation range of the final photovoltaic power prediction result is acquired, and the possible fluctuation interval of the output is analyzed. The grid frequency fluctuation signal is focused on. When the grid frequency is higher than the normal range, the active power output of the inverter cluster is appropriately reduced; otherwise, it is appropriately increased. According to the grid-connected power set value, the real-time output deviation and the grid frequency demand, the active power target value meeting the requirements of the grid is dynamically adjusted and calculated.

[0068] The calculation process of the reactive power is as follows: the grid voltage fluctuation is monitored in real time. If the grid voltage rises, the reactive power output needs to be reduced; if the grid voltage decreases, the reactive power output needs to be increased to maintain the stability of the grid voltage. Within the range of the reactive power regulation capacity, the grid voltage fluctuation degree and the reactive power regulation demand are combined to dynamically calculate the reactive power target value of the inverter cluster in each period.

[0069] S605, the deviation between the actual grid-connected power and the grid-connected power set value is monitored in real time. If the deviation continuously exceeds the preset threshold and exceeds the set time length, the dynamic re-adjustment of the grid-connected power set value or the alarm is triggered.

[0070] Through S601 to S605, the scheduling instructions and the power limit value can be preferentially met to ensure the safe and stable operation of the grid; the real-time output, the frequency / voltage signal are combined to improve the instruction response speed and accuracy; the deviation is continuously monitored to trigger re-adjustment to reduce manual intervention; the new energy consumption and the grid constraint are effectively balanced to improve the grid-friendly nature of the preset photovoltaic power station.

[0071] The reactive power compensation device is reasonably configured according to the corrected final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period to stabilize the grid voltage, and the specific implementation process is as follows: S606, based on the corrected final photovoltaic power prediction result, the topology structure, the line parameters and the load characteristics of the preset photovoltaic power station are combined to perform grid flow calculation to predict the voltage fluctuation trend and the voltage out-of-limit risk of the grid-connected point and the associated key nodes of the preset photovoltaic power station in each period in the second preset time period, specifically: According to the modified final photovoltaic power prediction result, the power output conditions of the preset photovoltaic power station in each period in the second preset time period are obtained, and the power data is taken as an input source of power flow calculation of the power grid. Then, combined with the topological structure of the preset photovoltaic power station, the connection mode and path of the preset photovoltaic power station and the power grid are determined, and the positions of the grid-connected point and the associated key nodes are determined, which is helpful to accurately divide the power flow direction and distribution area in the power grid. By using the line parameters, including the resistance, reactance and other characteristics of the line, the distribution of power in the power grid under different time periods is calculated by using a power flow calculation model, such as Newton-Raphson method, especially the influence of power injection and outflow of the grid-connected point and the key nodes on the voltage. At the same time, considering the load characteristics, the adjusting effect of photovoltaic power output change on the node voltage under different load levels is analyzed, and the voltage fluctuation trend of each period is predicted. According to the comparison between the calculated voltage value and the voltage safe operation range specified by the system, the possibility of voltage over-limit of each node is evaluated, and the time period and node with over-limit risk are identified, which provides a basis for subsequent voltage control and optimization.

[0072] S607, according to the predicted voltage fluctuation trend and voltage over-limit risk, determining the total demand for reactive power required for the preset photovoltaic power station to maintain the voltage stability of the power grid in the second preset time period and the dynamic change characteristics thereof; S608, taking minimizing the comprehensive operation cost of the reactive power compensation device, maximizing the voltage stability margin or minimizing the voltage deviation as an objective function, a reactive power compensation configuration optimization model is constructed; wherein the optimization model considers the following constraints: ① the total demand for reactive power and the dynamic change characteristics thereof; ② the capacity, response speed, adjustment range and operation frequency limit of the existing reactive power compensation devices (including photovoltaic inverter reactive power, SVC, SVG, capacitor bank, etc.) of the preset photovoltaic power station; ③ the safe operation constraints of the power grid, including the upper and lower limits of the voltage of each node and the power transmission limit of the key line; S609, solving the reactive power compensation configuration optimization model to obtain the optimal switching state instruction or optimal reactive power output instruction sequence of the existing reactive power compensation devices in each period in the second preset time period; S610, the optimal switching state instruction or optimal reactive power output instruction sequence is sent to the corresponding reactive power compensation device controller for execution; S611, based on the expected deviation range of the final photovoltaic power prediction result, the voltage control dead zone or the reserved reactive power margin of the operation of the reactive power compensation device is dynamically adjusted.

[0073] From S606 to S611, over-limit risks are predicted through power flow calculations, and reactive power compensation commands are generated in advance; the operation of inverters, SVG, and capacitor banks is optimized in a coordinated manner to reduce compensation losses; and control dead zones and reserved reactive power margins are dynamically adjusted to cope with prediction uncertainties. This significantly improves grid voltage stability, extends the life of reactive power equipment, and reduces operation and maintenance costs.

[0074] S7. Based on the final photovoltaic power prediction results of the preset photovoltaic power station within the second preset time period, evaluate the equipment operation status of the preset photovoltaic power station, specifically: S701. Real-time acquisition of operation monitoring data of each device in the preset photovoltaic power station within the second preset time period. The operation monitoring data includes: current, voltage and temperature data of photovoltaic string level; input / output power, conversion efficiency and heat dissipation status data of inverter; electrical parameter data of combiner box, transformer and grid connection point. S702. Based on the deviation between the final photovoltaic power prediction result and the actual total output during the corresponding period, and combined with the operation monitoring data, calculate the performance deviation index of each photovoltaic module, including: string output deviation rate and dispersion; deviation between the actual conversion efficiency and the rated efficiency of the inverter; and abnormal temperature rise index of electrical equipment. S703. Correlation analysis is performed between performance deviation indicators and preset health thresholds, historical data from the same period, and equipment operating years to diagnose potential fault types and performance degradation trends of the equipment. S704. Based on the expected deviation range of the final photovoltaic power prediction results, dynamically correct the evaluation sensitivity threshold of the performance deviation index and generate a graded early warning signal for equipment health status. S705, outputs an assessment report including equipment anomaly location, degree of deterioration, and maintenance recommendations.

[0075] From S701 to S705, potential faulty equipment is located using indicators such as dispersion and deviation; thresholds are adaptively adjusted based on the predicted deviation range to reduce false alarm rate; anomaly location and maintenance suggestions are output to reduce unplanned downtime; and asset utilization and power generation revenue are improved.

[0076] In the above embodiments, although the steps are numbered S1, S2, etc., they are only specific embodiments given by the present invention. Those skilled in the art can adjust the execution order of S1, S2, etc. according to the actual situation, which is also within the protection scope of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0077] like Figure 2 As shown, a photovoltaic power prediction system 200 based on multi-model fusion according to an embodiment of the present invention includes: a photovoltaic power prediction module 201 and a weight fusion module 202; The photovoltaic power prediction module 201 is configured to: based on photovoltaic power correlation data of a preset photovoltaic power station in a first preset time period, and by using each preset power prediction model, obtain a photovoltaic power prediction result of the preset photovoltaic power station in a second preset time period, and the second preset time period is located after the first preset time period in time sequence; The weight fusion module 202 is configured to: based on weights assigned to the photovoltaic power prediction results obtained by each preset power prediction model, fuse all the photovoltaic power prediction results to obtain a final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

[0078] Optionally, in the above technical solution, the aging rate calculation module and the correction module are further included. The aging rate calculation module is configured to: calculate an aging rate according to a used year number of a photovoltaic component of the preset photovoltaic power station and an aging curve corresponding to the photovoltaic component. The correction module is configured to: correct the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period by using the aging rate.

[0079] Optionally, in the above technical solution, a mathematical expression of the aging curve is: Wherein, X represents the used year number, and a represents the aging rate.

[0080] Optionally, in the above technical solution, the model acquisition module is further included, and the model acquisition module is configured to: When the number of the preset power prediction models is 3, the time sequence Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are trained respectively to obtain the three preset power prediction models.

[0081] Optionally, in the above technical solution, the weight fusion module 202 is specifically configured to: The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is calculated by using a weight fusion formula, and the weight fusion formula is: y=a x model_informer+b x model_lstm+c x model_svr Wherein, y represents: the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, a represents: the weight allocated to the photovoltaic power prediction result obtained through the first preset power prediction model, b represents: the weight allocated to the photovoltaic power prediction result obtained through the second preset power prediction model, c represents: the weight allocated to the photovoltaic power prediction result obtained through the third preset power prediction model, a+b+c=1, the first preset power prediction model is: a preset power prediction model obtained after training a time series Informer model, the second preset power prediction model is: a preset power prediction model obtained after training a recurrent neural network LSTM model, the third preset power prediction model is: a preset power prediction model obtained after training a support vector regression SVR model, model_informer represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the first preset power prediction model; model_lstm represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the second preset power prediction model; model_svr represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained through the third preset power prediction model.

[0082] It should be noted that the beneficial effects of the photovoltaic power prediction system 200 based on multi-model fusion provided by the above embodiment are the same as those of the photovoltaic power prediction method based on multi-model fusion, and will not be repeated here. In addition, when the system provided by the above embodiment implements its functions, only the division of the above functional modules is exemplified, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the system is divided into different functional modules according to actual conditions to complete all or part of the above described functions. In addition, the system and method embodiments provided by the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0083] Wherein, the photovoltaic power prediction system based on multi-model fusion of the present application can be a computer program (including program code) running in a computer device, for example, the photovoltaic power prediction system based on multi-model fusion of the present application is an application software, which can be used to execute the corresponding steps in the photovoltaic power prediction method based on multi-model fusion of the present application.

[0084] In some embodiments, the multi-model fusion based photovoltaic power prediction system of the present application can be implemented in a combination of software and hardware. For example, the multi-model fusion based photovoltaic power prediction system of the present application can be a hardware decoding processor programmed to perform the multi-model fusion based photovoltaic power prediction method of the present application. For example, the hardware decoding processor can be one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic elements.

[0085] The modules described in the embodiments of the present application can be implemented in the form of software or hardware. In some cases, the names of the modules do not limit the modules themselves.

[0086] An electronic device according to an embodiment of the present application includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above multi-model fusion based photovoltaic power prediction methods. That is, an electronic device according to an embodiment of the present application can include, but is not limited to, a processor and a memory. The memory is configured to store a computer program. The processor is configured to execute the multi-model fusion based photovoltaic power prediction method according to any of the embodiments of the present application by invoking the computer program.

[0087] In an optional embodiment, an electronic device is provided, as shown in Figure 3 As shown in Figure 3 The electronic device 4000 shown in the optional embodiment includes a processor 4001 and a memory 4003. The processor 4001 and the memory 4003 are connected, for example, via a bus 4002. Optionally, the electronic device 4000 can also include a transceiver 4004, which can be used for data exchange between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not limit the embodiments of the present application.

[0088] The processor 4001 can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It can implement or execute various exemplary logical blocks, modules and circuits described in connection with the present disclosure. The processor 4001 can also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0089] The bus 4002 can include a path for transmitting information between the above-mentioned components. The bus 4002 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, or the like. The bus 4002 can be divided into an address bus, a data bus, a control bus, and the like. For convenience of representation, Figure 3 The bus 4002 is represented by only one thick line, but it does not mean that there is only one bus or only one type of bus.

[0090] The memory 4003 can be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0091] The memory 4003 is configured to store application code (computer program) for implementing the scheme of the present application, and the processor 4001 is configured to control the execution. The processor 4001 is configured to execute the application code stored in the memory 4003 to implement the content shown in the foregoing method embodiments.

[0092] The electronic device can also be a terminal device, and the terminal device can be any device that can install an application, including at least one of a smartphone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, a smart television, and a smart vehicle device.

[0093] It should be noted that, Figure 3 The electronic device shown is only an example and should not limit the functions and use range of the embodiments of the present application.

[0094] The computer readable storage medium of the embodiment of the present application, the computer readable storage medium has a computer program stored thereon, and the computer program is executed by the processor to implement any one of the above-mentioned photovoltaic power prediction methods based on multi-model fusion.

[0095] Optionally, the computer readable storage medium can be a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a read-only compact disc (Compact Disc Read-Only Memory, CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0096] In the exemplary embodiments, a computer program product or computer program is also provided, which includes computer instructions stored in a computer readable storage medium. The processor of the electronic device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the electronic device to perform any one of the above-mentioned photovoltaic power prediction methods based on multi-model fusion.

[0097] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0098] It should be understood that the flowchart and block diagrams in the drawings illustrate the architecture, functionality, and operation of possible implementations of various embodiments of the present application. In this regard, each block in the flowchart and block diagrams can represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations thereof, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or combinations of hardware and software.

[0099] The computer readable storage medium of embodiments of the present application can be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the computer readable storage medium can include, but are not limited to, the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, the computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0100] The computer readable storage medium described above bears one or more programs, when the one or more programs are executed by the electronic device, the electronic device executes the method shown in the above embodiment.

[0101] The above description is merely exemplary of the application and the application principles of the technology used. Those skilled in the art should understand that the disclosed range of the application is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combinations of the above technical features or their equivalent features without departing from the disclosed concept. For example, the above features are replaced with the technical features disclosed in the application (but not limited to) having similar functions to form technical solutions.

[0102] It should be noted that the terms "first", "second" and the like in the specification and claims of the present application are used to distinguish similar objects, and represent a specific order or sequence. The order of use of similar objects can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described.

[0103] Those skilled in the art know that the application can be implemented as a system, a method or a computer program product, so the application can be specifically implemented as follows: it can be a complete hardware, a complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, which is generally referred to as "circuit", "module" or "system" in this paper. In addition, in some embodiments, the application can also be implemented as a computer program product in one or more computer readable media, which contains computer readable program code.

[0104] Although the embodiments of the application have been shown and described above, it should be understood that the above embodiments are exemplary and cannot be understood as limiting the application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the application.

Claims

1. A photovoltaic power prediction method based on multi-model fusion, characterized in that, Comprise: Based on the preset photovoltaic power station in the first preset time period of photovoltaic power correlation data, and using each preset power prediction model respectively get the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, in time sequence, the second preset time period is located after the first preset time period; Based on the weight allocated to the photovoltaic power prediction result obtained by each preset power prediction model, all photovoltaic power prediction results are fused to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

2. The photovoltaic power prediction method based on multi-model fusion according to claim 1, characterized in that, Also include: According to the used years of photovoltaic components of the preset photovoltaic power station and the aging curve corresponding to the photovoltaic components, the aging rate is calculated; The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is corrected by using the aging rate.

3. The photovoltaic power prediction method based on multi-model fusion according to claim 2, characterized in that, The mathematical expression of the aging curve is: Wherein, x represents: used years, alpha represents: aging rate.

4. The photovoltaic power prediction method based on multi-model fusion according to any one of claims 1 to 3, characterized in that, The acquisition process of each preset power prediction model includes: When the number of the preset power prediction model is 3, the time sequence Informer model, the recurrent neural network LSTM model and the support vector regression SVR model are trained respectively to obtain 3 preset power prediction models.

5. The photovoltaic power prediction method based on multi-model fusion according to claim 4, characterized in that, Based on the weight allocated to the photovoltaic power prediction result obtained by each preset power prediction model, all photovoltaic power prediction results are fused to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, including: The final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period is calculated by the weight fusion formula, and the weight fusion formula is: y=a*model_informer+b*model_lstm+c*model_svr Wherein, y represents: the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period, a represents: the weight allocated to the photovoltaic power prediction result obtained by the first preset power prediction model, b represents: the weight allocated to the photovoltaic power prediction result obtained by the second preset power prediction model, c represents: the weight allocated to the photovoltaic power prediction result obtained by the third preset power prediction model, a+b+c=1, the first preset power prediction model is: a preset power prediction model obtained after training the time series Informer model, the second preset power prediction model is: a preset power prediction model obtained after training the recurrent neural network LSTM model, and the third preset power prediction model is: a preset power prediction model obtained after training the support vector regression SVR model; model_informer represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the first preset power prediction model; model_lstm represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the second preset power prediction model; and model_svr represents: the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period obtained by the third preset power prediction model.

6. A photovoltaic power prediction system based on multi-model fusion, characterized in that, It comprises: a photovoltaic power prediction module and a weight fusion module; The photovoltaic power prediction module is used to obtain the photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period based on the photovoltaic power correlation data of the preset photovoltaic power station in the first preset time period and by using each preset power prediction model in time sequence; The weight fusion module is used to fuse all the photovoltaic power prediction results based on the weight allocated to the photovoltaic power prediction result obtained by each preset power prediction model to obtain the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period.

7. The photovoltaic power prediction system based on multi-model fusion according to claim 6, characterized in that, It further comprises an aging rate calculation module and a correction module; The aging rate calculation module is used to calculate the aging rate according to the service years of the photovoltaic components of the preset photovoltaic power station and the aging curve corresponding to the photovoltaic components; The correction module is used to correct the final photovoltaic power prediction result of the preset photovoltaic power station in the second preset time period by using the aging rate.

8. The photovoltaic power prediction system based on multi-model fusion according to claim 7, characterized in that, The mathematical expression of the aging curve is: Wherein, x represents: the service years, and a represents: the aging rate.

9. An electronic device, comprising: It comprises a memory, a processor, and a computer program stored on the memory and executable on the processor, and the processor implements the photovoltaic power prediction method based on multi-model fusion according to any one of claims 1 to 5 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer program is stored on the computer readable storage medium, and the computer program is executed by the processor to implement the photovoltaic power prediction method based on multi-model fusion according to any one of claims 1 to 5.

Images