New energy generating capacity prediction method and system based on improved LSTM model

Abstract

The invention discloses a new energy generating capacity prediction method and system based on an improved LSTM model, and relates to the technical field of new energy generating capacity prediction, and the method comprises the steps: obtaining historical power generation data and multi-dimensional meteorological raster data of a new energy power station, and carrying out the preprocessing and feature extraction, and obtaining the preprocessed data; the preprocessed data serve as input, the trained improved LSTM model is used for predicting the generating capacity of the new energy power station, and a generating capacity prediction value is obtained; wherein the improved LSTM model comprises a periodic memory gating unit in which a periodic weight item is introduced; in the training process of the improved LSTM model, a memory attenuation compensation mechanism is introduced, and a cross validation and grid search method is adopted to carry out architecture and hyper-parameter optimization. According to the method, the technical problems of insufficient prediction precision and poor model stability when the traditional prediction method is used for processing the problems of long-term trend, seasonal change, complex meteorological influence and the like are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy power generation prediction, and specifically provides a new energy power generation prediction method and system based on an improved LSTM model. Background Technique

[0002] In recent years, with the large-scale access of new energy such as wind power and photovoltaic power, the new energy power generation data has shown strong non-linear, periodic and time-varying characteristics. Traditional prediction methods (including regression models, support vector machines, and conventional LSTM networks, etc.) generally face the problems of insufficient prediction accuracy and poor model stability when dealing with problems such as long-term trends, seasonal changes, and complex meteorological impacts. In particular, conventional LSTM performs poorly in capturing long-term dependencies, periodicity, and multi-scale information, resulting in difficulty in meeting the high-precision prediction requirements in the new energy power system dispatching and planning. Summary of the Invention

[0003] The purpose of the present invention is to provide a new energy power generation prediction method and system based on an improved LSTM model to solve at least one of the above technical problems.

[0004] In a first aspect, an embodiment of the present invention provides a new energy power generation prediction method based on an improved LSTM model, which is applied to a new energy power station; the method includes: obtaining historical power generation data and multi-dimensional meteorological grid data of the new energy power station; preprocessing and feature extraction of the historical power generation data and the multi-dimensional meteorological grid data to obtain preprocessed data; using the preprocessed data as input, and using the trained improved LSTM model to predict the power generation of the new energy power station to obtain a power generation prediction value; wherein, the improved LSTM model includes an input gate, a forget gate, and an output gate, and the input gate, the forget gate, and the output gate are all periodic memory gate control units introduced with periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

[0005] Further, the historical power generation data includes historical power generation time series data of the new energy power station; the multi-dimensional meteorological grid data includes wind speed, temperature, humidity, and irradiance.

[0006] Further, preprocess and extract features from the historical power generation data and the multi-dimensional meteorological grid data to obtain the preprocessed data, including: performing data cleaning and normalization on the historical power generation data and the multi-dimensional meteorological grid data to obtain the normalized historical power generation data and the normalized multi-dimensional meteorological grid data respectively; extracting local spatial features from the normalized multi-dimensional meteorological grid data based on a spatio-temporal convolutional network to obtain multi-dimensional meteorological spatial feature data; fusing the multi-dimensional meteorological spatial feature data with the normalized historical power generation data to obtain the fused data; performing dimensionality reduction and screening on the fused data based on the principal component analysis method to obtain the preprocessed data.

[0007] Further, the structural formula of the input gate in the improved LSTM model includes:

[0008] In the formula: represents the activation value of the input gate at time; represents the Sigmoid activation function; represents the weight matrix of the input gate; represents the hidden state at time; represents the input data at time; represents the bias of the input gate; represents the periodic weight term introduced in the input gate; the structural formula of the forget gate in the improved LSTM model includes:

[0009] In the formula: represents the activation value of the forget gate at time; represents the weight matrix of the forget gate; represents the bias of the forget gate; represents the periodic weight term introduced in the forget gate; represents the memory decay compensation term; the structural formula of the output gate in the improved LSTM model includes:

[0010] In the formula: represents the activation value of the output gate at time; represents the weight matrix of the output gate; represents the bias of the output gate; represents the periodic weight term introduced in the output gate.

[0011] Furthermore, the cell state update formula of the improved LSTM model includes:

[0012] In the formula: represents the cell state at time step; represents the cell state at time step t-1; represents the element-wise product; represents the hyperbolic tangent activation function; represents the weight matrix used to calculate the candidate cell state; represents the corresponding bias; the hidden state update formula of the improved LSTM model includes:

[0013] In the formula: represents the hidden state at time step.

[0014] Furthermore, the memory decay compensation term is adaptively adjusted according to the second-order derivative information of the loss function of the improved LSTM model with respect to the time step during training; the calculation formula of the memory decay compensation term includes:

[0015] In the formula: represents the memory decay compensation term at time step; is the adjustment coefficient; represents the loss function adopted by the improved LSTM model during training.

[0016] Furthermore, the periodic weight term includes:

[0017] Wherein, represents the periodic weight term corresponding to time step; is the learnable scaling parameter used to adjust the amplitude of the periodic weight during the learning process; are the Fourier cosine coefficient and sine coefficient corresponding to the th periodic component respectively; represents the fundamental frequency of the nth periodic component; is the index of the periodic component; represents the total number of considered periodic components.

[0018] Second aspect, an embodiment of the present invention further provides a new energy power generation prediction system based on an improved LSTM model, which is applied to a new energy power station; it includes: an acquisition module, a preprocessing module, and a prediction module; wherein, the acquisition module is used to acquire historical power generation data and multi-dimensional meteorological grid data of the new energy power station; the preprocessing module is used to preprocess and extract features from the historical power generation data and the multi-dimensional meteorological grid data to obtain the preprocessed data; the prediction module is used to use the preprocessed data as input, and use the trained improved LSTM model to predict the power generation of the new energy power station to obtain a power generation prediction value; wherein, the improved LSTM model includes an input gate, a forget gate, and an output gate, and the input gate, the forget gate, and the output gate are all periodic memory gating units introduced with periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

[0019] Third aspect, an embodiment of the present invention further provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, it implements the method provided by the embodiment of the present invention.

[0020] Fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, and the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method provided by the embodiment of the present invention is implemented.

[0021] The present invention provides a new energy power generation prediction method and system based on an improved LSTM model. By introducing periodic weights and a memory decay compensation mechanism into the traditional LSTM model, the periodicity and long-term dependence information in historical power generation and meteorological data are fully captured, thereby improving the medium- and long-term prediction accuracy of new energy power generation. At the same time, through the spatial feature extraction of multi-dimensional meteorological data and the selection of key features, redundant information is effectively reduced, the training efficiency and generalization ability of the model are improved, and the technical problems of insufficient prediction accuracy and poor model stability existing in traditional prediction methods when dealing with long-term trends, seasonal changes, and complex meteorological impacts are alleviated. Description of the Drawings

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific embodiments or the description of the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0023] Figure 1 Flow chart of a new - energy power generation prediction method based on an improved LSTM model provided by an embodiment of the present invention; Figure 2 Schematic diagram of a new - energy power generation prediction system based on an improved LSTM model provided by an embodiment of the present invention. Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0025] Embodiment 1

[0026] Figure 1 is a flow chart of a new - energy power generation prediction method based on an improved LSTM model provided by an embodiment of the present invention. This method is applied to a new - energy power station. As Figure 1 shown, the method specifically includes the following steps: Step S102, obtain the historical power generation data and multi - dimensional meteorological grid data of the new - energy power station.

[0027] Specifically, the historical power generation data includes the historical power generation time - series data of the new - energy power station; for example, the monthly or annual power generation data of the new - energy power station in the past 1 to 3 years, including specific timestamps and corresponding power generation values.

[0028] The multi - dimensional meteorological grid data includes wind speed, temperature, humidity, and irradiance. Among them, the multi - dimensional meteorological grid data is meteorological data corresponding to the historical power generation time - series data. The meteorological data is provided in the form of grids, covering the spatial range of the area where the new - energy power station is located, and having the same time resolution (such as monthly or annual) as the historical power generation time - series data.

[0029] Optionally, the historical power generation time - series data can be obtained from the power grid dispatching system or the monitoring system of the new - energy power station; the meteorological data can be obtained from the meteorological data platform provided by the meteorological department or a meteorological data service company.

[0030] Step S104, pre - process and extract features from the historical power generation data and multi - dimensional meteorological grid data to obtain the pre - processed data. Specifically, it includes the following steps: Step S1041, perform data cleaning and normalization processing on the historical power generation data and multi - dimensional meteorological grid data respectively to obtain the normalized historical power generation data and the normalized multi - dimensional meteorological grid data.

[0031] Specifically, data cleaning includes removing outliers and missing values from the data; among them, for missing values, interpolation methods (such as linear interpolation or time series interpolation) can be used for filling; for outliers, they can be removed or corrected by setting a reasonable threshold range.

[0032] Specifically, in order to ensure the timestamp consistency and integrity of the data, the embodiments of the present invention also perform alignment processing on the data, so that the historical power generation data and the multi-dimensional meteorological grid data correspond one by one in time.

[0033] Specifically, the normalization processing formula includes:

[0034] Among them, represents the original data value; represents the data value after normalization; represents in the data set the minimum value; represents in the data set the maximum value. The data range after normalization is [0, 1], which helps to improve the training efficiency and stability of the model.

[0035] Step S1042, based on the spatio-temporal convolutional network, perform local spatial feature extraction on the multi-dimensional meteorological grid data after normalization to obtain multi-dimensional meteorological spatial feature data.

[0036] Specifically, input the meteorological grid data into the spatio-temporal convolutional network (STCNN), and perform feature extraction on the data through the convolutional layer and the pooling layer to extract a feature map containing local spatial information.

[0037] Step S1043, fuse the multi-dimensional meteorological spatial feature data with the historical power generation data after normalization to obtain the fused data.

[0038] Optionally, the fusion method can adopt simple splicing or weighted summation and other methods, which specifically depend on the characteristics of the data and the requirements of the model, and the present invention does not make specific limitations.

[0039] Step S1044, based on the principal component analysis method, perform dimensionality reduction and screening on the fused data to obtain the preprocessed data.

[0040] Specifically, for the principal component analysis (PCA) method, by calculating the eigenvalues and eigenvectors, select the first k principal components as the input features of the model, where the value of k can be determined according to the cumulative contribution rate (for example, taking 95%). The dimensionality-reduced feature data not only retains the main information of the original data but also reduces redundant information, improving the training efficiency and prediction effect of the model.

[0041] Step S106: Using the preprocessed data as input, the trained improved LSTM model is used to predict the power generation of the new energy power station, and the power generation prediction value is obtained.

[0042] Among them, the improved LSTM model includes an input gate, a forget gate, and an output gate. The input gate, forget gate, and output gate are all periodic memory gating units introducing periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced, and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

[0043] Specifically, the periodic weight term is obtained by performing frequency decomposition on the historical state sequence using Fourier basis functions, which is used to capture annual, quarterly, and monthly periodic characteristics, and at the same time, the periodic components are weighted and fused through learnable parameters.

[0044] In the embodiment of the present invention, by introducing a memory decay compensation mechanism during the training process of the improved LSTM model, the forget gate coefficient is dynamically adjusted according to the second derivative information of the loss function to compensate for the decay of long-term dependence information during cross-cycle transmission.

[0045] In the embodiment of the present invention, cross-validation and grid search methods are used to optimize the architecture and hyperparameters of each layer of the improved LSTM model, ensuring that the model can accurately fit historical data and has good generalization ability.

[0046] Specifically, during the training process of the improved LSTM model, cross-validation and grid search methods are used to determine the weights, biases, periodic weight term parameters, and memory decay compensation coefficients of each layer, so as to ensure high accuracy and robustness of the model in the medium- and long-term prediction stage while fitting historical data.

[0047] Specifically, in the embodiment of the present invention, the structural formula of the input gate in the improved LSTM model includes:

[0048] In the formula: represents the activation value of the input gate at time; represents the Sigmoid activation function; represents the weight matrix of the input gate; represents the hidden state at time; represents the input data at time; represents the bias of the input gate; represents the periodic weight term introduced in the input gate; represents the time variable or time index; The structural formula for improving the forget gate in the LSTM model includes:

[0049] In the formula: represents the activation value of the forget gate at time; represents the weight matrix of the forget gate; represents the bias of the forget gate; represents the periodic weight term introduced in the forget gate, which is calculated through Fourier basis functions, and its composition method is the same as that of the periodic weight term in the input gate; represents the memory decay compensation term, which is adaptively adjusted according to the second derivative information of the loss function with respect to time during the model training process, and is used to compensate for the decay of long-term dependence information; The structural formula for improving the output gate in the LSTM model includes:

[0050] In the formula: represents the activation value of the output gate at time; represents the weight matrix of the output gate; represents the bias of the output gate; represents the periodic weight term introduced in the output gate.

[0051] Specifically, the periodic weight term includes:

[0052] Among them, represents the corresponding periodic weight term at time; is a learnable scaling parameter used to adjust the amplitude of the periodic weight during the learning process; are the Fourier cosine coefficient and sine coefficient corresponding to the th periodic component respectively; represents the fundamental frequency of the nth periodic component (corresponding to periods such as annual, quarterly, monthly, etc.); is the index of the periodic component, ; represents the total number of periodic components considered.

[0053] For example, the periodic weight term of the input gate is:

[0054] In the formula, represents the periodic weight term introduced in the input gate; represents the learnable scaling parameter used to control the amplitude of the periodic term in the input gate.

[0055] The periodic weight term of the output gate is as follows:

[0056] In the formula, represents the periodic weight term introduced in the output gate; represents the learnable scaling parameter that controls the amplitude of the output gate control period term; are respectively the Fourier cosine and sine coefficients corresponding to the th periodic component in the output gate.

[0057] For example, considering the annual , quarterly , monthly these 3 periodic components , through the learnable parameter weighted fusion of each periodic component is performed to effectively capture annual, quarterly, and monthly periodic characteristics.

[0058] Specifically, the memory decay compensation term is adaptively adjusted according to the second-order derivative information of the loss function with respect to the time step during the training process of the improved LSTM model; the calculation formula of the memory decay compensation term includes:

[0059] In the formula: represents the memory decay compensation term at the moment; is the adjustment coefficient used to control the compensation amplitude; represents the loss function adopted during the training process of the improved LSTM model; represents the loss function with respect to time second-order derivative, reflecting the curvature of the loss change with time, and is used to dynamically optimize the adjustment of the forget gate.

[0060] Specifically, at the beginning of training, if the curvature of the loss function changes rapidly with time (i.e., the absolute value of the second-order derivative is large), it indicates that the model prediction error changes rapidly. At this time, appropriately increase the value to strengthen the compensation for long-term dependence information; as the training progresses, if the loss function tends to be stable and the absolute value of the second-order derivative becomes smaller, then correspondingly reduce the value to avoid overcompensation. Through this adaptive adjustment, the attenuation of long-term dependence information during cross-period transmission is compensated.

[0061] Specifically, in the embodiments of the present invention, the cell state update formula of the improved LSTM model includes:

[0062] In the formula: represents the cell state at time; represents the cell state at time t-1; represents the element-wise product, i.e., the Hadamard product; represents the hyperbolic tangent activation function; represents the weight matrix used to calculate the candidate cell state; represents the corresponding bias; i t used to control the proportion of new information added.

[0063] Specifically, the hidden state update formula of the improved LSTM model includes:

[0064] In the formula: represents the hidden state at used to modulate the result after the non-linear transformation of the cell state.

[0065] The method provided by the embodiments of the present invention further includes training the improved LSTM model to obtain a trained improved LSTM model. Specifically, the preprocessed data is divided into a training set, a validation set, and a test set, with proportions of 70%, 15%, and 15% respectively. The training set is used for model training, the validation set is used for hyperparameter optimization, and the test set is used to evaluate the final performance of the model.

[0066] Specifically, in the embodiments of the present invention, the specific steps for architecture and hyperparameter optimization using cross-validation and grid search methods are as follows: Set a series of candidate hyperparameter values, including the number of neurons in the LSTM layer, learning rate, batch size, number of training epochs, etc.; use the cross-validation method (such as k-fold cross-validation) to evaluate the model, divide the training set into k subsets, where k-1 subsets are used for training and 1 subset is used for validation, repeat k times, and take the average validation error as the evaluation index of the model performance; according to the cross-validation results, use the grid search method to search for the optimal hyperparameter combination among the candidate hyperparameter values to ensure that the model can accurately fit the historical data and has good generalization ability.

[0067] Train the improved LSTM model using the optimized hyperparameters, use the mean squared error (MSE) as the loss function, and update the model parameters through the backpropagation algorithm until the model converges.

[0068] Use the trained improved LSTM model to predict the new energy power generation at future times (such as monthly or annually), and generate the predicted values of the new energy power generation at future times. The prediction results should include specific timestamps and corresponding predicted power generation values.

[0069] As can be seen from the above description, the embodiment of the present invention provides a new energy power generation prediction method based on an improved LSTM model. By introducing a periodic weight and a memory decay compensation mechanism into the traditional LSTM model, the periodicity and long-term dependence information in historical power generation and meteorological data are fully captured, thereby improving the medium- and long-term prediction accuracy of new energy power generation. At the same time, through the extraction of spatial features and the selection of key features of multi-dimensional meteorological data, redundant information is effectively reduced, the training efficiency and generalization ability of the model are improved, and the technical problems of insufficient prediction accuracy and poor model stability in traditional prediction methods when dealing with long-term trends, seasonal changes, and complex meteorological impacts are alleviated.

[0070] Embodiment 2

[0071] Figure 2 It is a schematic diagram of a new energy power generation prediction system provided according to an embodiment of the present invention. The system is applied to a new energy power station. As Figure 2 shown, it includes: an acquisition module 10, a preprocessing module 20, and a prediction module 30.

[0072] Specifically, the acquisition module 10 is used to acquire historical power generation data and multi-dimensional meteorological grid data of the new energy power station.

[0073] The preprocessing module 20 is used to preprocess and extract features from the historical power generation data and multi-dimensional meteorological grid data to obtain the preprocessed data.

[0074] Specifically, the preprocessing module 20 is further used for: Performing data cleaning and normalization processing on the historical power generation data and multi-dimensional meteorological grid data to obtain the normalized historical power generation data and the normalized multi-dimensional meteorological grid data respectively; Performing local spatial feature extraction on the normalized multi-dimensional meteorological grid data based on a spatio-temporal convolutional network to obtain multi-dimensional meteorological spatial feature data; Fusing the multi-dimensional meteorological spatial feature data with the normalized historical power generation data to obtain the fused data; Performing dimensionality reduction and screening on the fused data based on the principal component analysis method to obtain the preprocessed data.

[0075] The prediction module 30 is used to take the preprocessed data as input and use the trained improved LSTM model to predict the power generation of the new energy power station to obtain the power generation prediction value.

[0076] Among them, the improved LSTM model includes an input gate, a forget gate, and an output gate, and the input gate, the forget gate, and the output gate are all periodic memory gating units introducing periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

[0077] Specifically, as Figure 2 shown, the system provided by the embodiment of the present invention further includes a training module 40, which is used to train the improved LSTM model to obtain a trained improved LSTM model.

[0078] The present invention also provides an electronic device, including: a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the method provided by the embodiment of the present invention is implemented.

[0079] The present invention also provides a computer-readable storage medium, and the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by the processor, the method provided by the embodiment of the present invention is implemented.

[0080] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0081] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A new energy power generation prediction method based on an improved LSTM model, characterized in that, Applied to a new energy power station; including: Obtain the historical power generation data and multi-dimensional meteorological grid data of the new energy power station; Preprocess and extract features from the historical power generation data and the multi-dimensional meteorological grid data to obtain the preprocessed data; Using the preprocessed data as input, utilize the trained improved LSTM model to predict the power generation of the new energy power station to obtain the power generation prediction value; Among them, the improved LSTM model includes an input gate, a forget gate, and an output gate, and the input gate, the forget gate, and the output gate are all periodic memory gating units introducing periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

2. The method according to claim 1, wherein: The historical power generation data includes the historical power generation time series data of the new energy power station; The multi-dimensional meteorological grid data includes wind speed, temperature, humidity, and irradiance.

3. The method according to claim 1, characterized in that: Preprocessing and extracting features from the historical power generation data and the multi-dimensional meteorological grid data to obtain the preprocessed data includes: Perform data cleaning and normalization processing on the historical power generation data and the multi-dimensional meteorological grid data respectively to obtain the normalized historical power generation data and the normalized multi-dimensional meteorological grid data; Based on a spatio-temporal convolutional network, perform local spatial feature extraction on the normalized multi-dimensional meteorological grid data to obtain multi-dimensional meteorological spatial feature data; Fuse the multi-dimensional meteorological spatial feature data with the normalized historical power generation data to obtain the fused data; Based on the principal component analysis method, perform dimensionality reduction and screening on the fused data to obtain the preprocessed data.

4. The method according to claim 1, wherein: The structural formula of the input gate in the improved LSTM model includes: ， Wherein: represents the activation value of the input gate at the moment; represents the Sigmoid activation function; represents the weight matrix of the input gate; represents the hidden state at represents the input data at represents the bias of the input gate; represents the periodic weight term introduced in the input gate; represents the time variable or the time index; The structural formula of the forget gate in the improved LSTM model includes: ， In the formula: represents the activation value of the forgetting gate at moment; represents the weight matrix of the forgetting gate; represents the bias of the forgetting gate; represents the periodic weight term introduced in the forgetting gate; represents the memory decay compensation term; The structural formula of the output gate in the improved LSTM model includes: ， Wherein: represents the activation value of the output gate at the moment; represents the weight matrix of the output gate; represents the bias of the output gate; represents the periodic weight term introduced in the output gate.

5. The method according to claim 4, wherein: The unit state update formula of the improved LSTM model includes: ， In the formula: represents the unit state at moment; represents the unit state at time t-1; represents the element-wise product; represents the hyperbolic tangent activation function; represents the weight matrix for calculating the candidate unit state; represents the corresponding bias; The hidden state update formula of the improved LSTM model includes: ， Wherein: represents the hidden state at the moment.

6. The method according to claim 4, wherein: The memory decay compensation term is adaptively adjusted according to the second-order derivative information of the loss function of the improved LSTM model with respect to the time step during the training process; The calculation formula of the memory decay compensation term includes: ， Wherein: represents the memory decay compensation term at time; is an adjustment coefficient; represents the loss function adopted by the improved LSTM model during training.

7. The method according to claim 1, wherein: The periodic weight term includes: ， Among them, represents the periodic weight term corresponding to the moment; is a learnable scaling parameter used to adjust the amplitude of the periodic weight during the learning process; are the Fourier cosine coefficient and sine coefficient corresponding to the -th periodic component respectively; represents the fundamental frequency of the n-th periodic component; is the index of the periodic component; represents the total number of periodic components considered.

8. A new energy power generation prediction system based on an improved LSTM model, characterized in that, Applied to a new energy power station; including: an acquisition module, a preprocessing module, and a prediction module; where The acquisition module is used to obtain the historical power generation data and multi-dimensional meteorological grid data of the new energy power station; The preprocessing module is used to preprocess and extract features from the historical power generation data and the multi-dimensional meteorological grid data to obtain the preprocessed data; The prediction module is used to use the preprocessed data as input and utilize the trained improved LSTM model to predict the power generation of the new energy power station to obtain the power generation prediction value; Among them, the improved LSTM model includes an input gate, a forget gate, and an output gate, and the input gate, the forget gate, and the output gate are all periodic memory gating units introducing periodic weight terms; during the training process of the improved LSTM model, a memory decay compensation mechanism is introduced and cross-validation and grid search methods are used for architecture and hyperparameter optimization.

9. An electronic device, characterized in that, It includes: A memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the method according to any one of claims 1-7 is implemented.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by the processor, the method according to any one of claims 1-7 is implemented.

Images