Photovoltaic power generation power prediction method based on LSTM and EBRB

Abstract

The invention relates to a photovoltaic power generation power prediction method based on LSTM and EBRB, and belongs to the field of photovoltaic power generation power prediction. The method comprises the following steps: step 1, extracting a photovoltaic power generation power influence factor based on feature engineering; step 2, constructing a prediction model based on LSTM; step 3, constructing an inference model based on EBRB; step 4, constructing a Stacking integration model based on the LSTM and the EBRB; and 5, inputting the data into the Stacking integrated model to obtain a photovoltaic power generation power prediction result. The method can effectively improve the prediction precision of the photovoltaic power generation power, and guarantees the operation stability of a power grid. According to the method, the LSTM and the EBRB are combined, the Stacking integrated model based on the LSTM and the EBRB is constructed, the interpretability and the prediction accuracy of the model are effectively improved, and the research provides support for further application and development of the LSTM and the EBRB in the field of photovoltaic power generation power prediction.

Description

Technical Field

[0001] The present invention belongs to the field of photovoltaic power generation prediction, and specifically relates to a photovoltaic power generation power prediction method based on LSTM and EBRB. Background Art

[0002] With the global energy transition and the rapid development of renewable energy, photovoltaic power generation, as a key form of clean energy, has seen significant growth in both installed capacity and power generation. However, photovoltaic power generation is characterized by randomness and volatility, primarily influenced by factors such as sunlight intensity, weather conditions, and seasonal variations. This poses challenges to grid scheduling and operation, and also impacts the economic benefits and market competitiveness of photovoltaic power generation. Therefore, photovoltaic power prediction technology is crucial for improving photovoltaic conversion efficiency and optimizing capacity allocation. Traditional photovoltaic power prediction methods primarily rely on physical models and statistical methods, such as electronic component model prediction and time series methods. These methods may underperform newer algorithms such as deep learning when dealing with complex nonlinear relationships and large amounts of data. In recent years, prediction methods based on artificial intelligence (AI) have gradually emerged, with deep learning, in particular, playing a significant role in photovoltaic power prediction. Deep learning leverages its self-learning capabilities to construct complex mappings from multiple inputs to single or multiple outputs, enabling accurate prediction of photovoltaic power. However, deep learning models are often viewed as black-box models, making their internal mechanisms and decision-making processes difficult to explain and lacking interpretability. Summary of the Invention

[0003] The purpose of the present invention is to solve the problems raised by the existing background technology and provide a photovoltaic power generation power prediction method based on LSTM and EBRB.

[0004] To achieve the above purpose, the technical solution of the present invention is: a photovoltaic power generation prediction method based on LSTM and EBRB, comprising

[0005] Step 1: Extract factors influencing photovoltaic power generation based on feature engineering: Use the extreme gradient boosting algorithm to extract key factors affecting photovoltaic power generation and construct a set of key features with high relevance and representativeness;

[0006] Step 2: Build a prediction model based on LSTM: Using the extracted key features, train an LSTM model to capture the nonlinearity and long-term dependencies in the PV power time series;

[0007] Step 3: Build an inference model based on EBRB: Build an EBRB model based on the extracted key features and their distribution, assess the uncertainty and complexity of the data, and generate reliable prediction outputs;

[0008] Step 4: Build a stacking ensemble model based on LSTM and EBRB: Use the constructed LSTM model and EBRB model as base learners, the linear regression model as the meta-learner, and then combine the outputs of the base learners to train the meta-learner.

[0009] Step 5: Input the data into the Stacking integrated model to obtain the photovoltaic power generation power prediction results.

[0010] In one embodiment of the present invention, in step one, the photovoltaic power generation influencing factors are extracted based on the extreme gradient boosting algorithm, the feature importance score provided by the XGBoost model is used to evaluate the contribution of each feature to the photovoltaic power generation power prediction, and the features are sorted by importance, and the top five key features are selected as the independent variables of the model.

[0011] In one embodiment of the present invention, in step 2, an LSTM model is trained to capture the nonlinearity and long-term dependencies in the photovoltaic power generation time series, that is: in the process of constructing the LSTM model, an input layer that matches the shape of the training data is first set, an LSTM layer is added to capture the long-term dependencies of the photovoltaic power generation data, and overfitting is reduced through a Dropout layer. Finally, a fully connected layer is used as the output layer to predict the target variable, and an Adam optimizer and a mean square error loss function are configured to compile the model.

[0012] In one embodiment of the present invention, in step 3, an EBRB model set is constructed, an evaluation model is constructed using EBRB, and ER rules are used as the inference engine. The modeling process is as follows:

[0013] (1) EBRB-type knowledge base description

[0014] The EBRB model is constructed by IF-THEN rules;

[0015] (2) EBRB model reasoning process

[0016] For the EBRB model, the ER rule is used as the inference engine, and the photovoltaic power generation power prediction value is obtained based on the EBRB model reasoning.

[0017] In one embodiment of the present invention, in the IF-THEN rule, the kth extended confidence rule is defined as follows:

[0018]

[0019] Where Aki represents the candidate value of the i-th premise attribute in the k-th extended confidence rule; T k represents the number of premise attributes in the kth extended confidence rule; β n,k Denotes the attribute D assigned to the result in the k-th extended confidence rule nConfidence, if ΣNn=1β n,k <1, the information contained in the kth extended confidence rule is said to be complete, otherwise the information contained in the kth extended confidence rule is incomplete; δ k,i (i=1,…,T k ) represents the attribute weight of the i-th premise attribute in the k-th extended confidence rule; θ k represents the rule weight of the kth extended confidence rule.

[0020] In one embodiment of the present invention, the EBRB model reasoning process is as follows:

[0021] Step 1: Initialize the values ​​of key parameters in the EBRB model: use expert knowledge to give all candidate values ​​in all premise attributes, the utility values ​​of all evaluation levels of the result attributes, and the attribute weights of all premise attributes;

[0022] Step 2: Calculate individual matching degree: Assume there are T groups of data, expressed as <x i ,y i >(i=1,…,T), and its candidate value set is represented by A i ={A i,j ; j=1,…,J i}, that is, all evaluation levels of various indicators that affect photovoltaic power generation; among them, x i represents the input value vector of the i-th premise attribute, i.e., the data of various indicators that affect photovoltaic power generation; i represents the i-th output value, i.e., photovoltaic power generation. Then, the individual matching degree of each candidate value is calculated based on the rule-based information conversion technology. The corresponding formula is as follows:

[0023]

[0024] After calculating the individual matching degrees of all candidate values, the individual matching degrees are expressed through a distributed framework; the corresponding formula is as follows:

[0025] S(x i )={(A ij , α ij );i=1,2,…,T;j=1,2,…,J i}

[0026] Among them, α i,j Indicates the jth candidate value A in the i-th premise attribute i,j Individual matching degree;

[0027] Step 3. Calculate the activation weight: Integrate the rule weight, attribute weight, and individual matching degree to obtain the activation weight of the kth extended confidence rule:

[0028]

[0029] Among them, αk i represents the individual matching degree of the i-th premise attribute candidate value in the k-th extended confidence rule; when ω k When >0, it means that the kth extended confidence rule is activated;

[0030] Step 4: Apply ER rules to synthesize confidence: All activated rules are synthesized at once through the analytical formula in the ER rules. The corresponding formula for the synthesized new confidence is as follows:

[0031]

[0032] When the input value is x={x1,…,x M}, the predicted value based on the EBRB model is:

[0033] f(x)={(D i , β i (x)), i=1, 2,...,N}.

[0034] In one embodiment of the present invention, step 4 is specifically implemented as follows:

[0035] (1) Data division

[0036] Divide the time series data into training set and test set in chronological order; let T be the total number of time periods, T train is the number of time periods in the training set, T test is the number of time periods of the test set, then the training set and test set are expressed as:

[0037] D train ={X1,X2,...,X T_train}

[0038] D test ={X T_train+1 ,X T_train+2 ,...,X T}

[0039] (2) Base learner training

[0040] For each base learner M i Train on the training set; let M i is the i-th base learner, θ i Its parameters; the training process is expressed as:

[0041]

[0042] Where L is the loss function, X t is the data at time t in the training set, Y tis the true label at time t;

[0043] (3) Cross-validation prediction

[0044] For each base learner M i , using time series cross validation, D train Divide into K folds, D train (k) is the kth fold; for each fold k, the training set and the validation set are divided, where the validation set data volume is 10% of the original training set; using D train (k) training set training M i , and in D train (k) is predicted on the validation set, and K prediction results are obtained, which are recorded as Y' t (k);

[0045] (4) Combination learner training

[0046] Use the predicted values ​​of the base learners on the validation set as input features to train the combined learner; let Y' t (i) is the predicted value of the i-th base learner at time t, then the new feature vector Z t =[Y' t (1),Y' t (2),...,Y' t (N)], where N is the number of base learners; the combination learner M comb The parameter θ comb Train by:

[0047]

[0048] Using the combined learner M comb Make predictions on the test set and get Y' test .

[0049] In one embodiment of the present invention, in step five, the Stacking integrated model is a photovoltaic power prediction model with higher prediction accuracy and stability constructed by combining the LSTM model and the EBRB model. By inputting data into the model, the photovoltaic power prediction value can be obtained.

[0050] The present invention also provides a photovoltaic power generation prediction system based on LSTM and EBRB, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0051] The present invention also provides a computer-readable storage medium on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0052] Compared with the existing technology, the present invention has the following beneficial effects: the method of the present invention combines LSTM with EBRB to construct a Stacking integration model based on LSTM and EBRB, which effectively improves the interpretability and prediction accuracy of the model. This research provides support for the further application and development of LSTM and EBRB in the field of photovoltaic power generation prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a step diagram of the present invention;

[0054] Figure 2 Schematic diagram of the LSTM model of the present invention;

[0055] Figure 3 This is the EBRB model construction process of the present invention;

[0056] Figure 4 Schematic diagram of the Stacking integration model of the present invention. DETAILED DESCRIPTION

[0057] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0058] The present invention provides a photovoltaic power generation power prediction method based on LSTM and EBRB, including

[0059] Step 1: Extract factors influencing photovoltaic power generation based on feature engineering: Use the extreme gradient boosting algorithm to extract key factors affecting photovoltaic power generation and construct a set of key features with high relevance and representativeness;

[0060] Step 2: Build a prediction model based on LSTM: Using the extracted key features, train an LSTM model to capture the nonlinearity and long-term dependencies in the PV power time series;

[0061] Step 3: Build an inference model based on EBRB: Build an EBRB model based on the extracted key features and their distribution, assess the uncertainty and complexity of the data, and generate reliable prediction outputs;

[0062] Step 4: Build a stacking ensemble model based on LSTM and EBRB: Use the constructed LSTM model and EBRB model as base learners, the linear regression model as the meta-learner, and then combine the outputs of the base learners to train the meta-learner.

[0063] Step 5: Input the data into the Stacking integrated model to obtain the photovoltaic power generation power prediction results.

[0064] The following is a specific implementation process of the present invention.

[0065] See also Figures 1 to 4 The present invention provides a photovoltaic power generation prediction model based on LSTM and EBRB, the steps are as follows:

[0066] Step 1: Extract factors influencing photovoltaic power generation based on feature engineering: Use the extreme gradient boosting algorithm to extract key factors affecting photovoltaic power generation and construct a highly relevant and representative feature set;

[0067] Step 2: Build a prediction model based on LSTM: Using the extracted key features, train an LSTM model to capture the nonlinearity and long-term dependencies in the PV power time series.

[0068] Step 3: Build an inference model based on EBRB: Build EBRB based on the extracted features and their distribution, evaluate the uncertainty and complexity of the data, and generate reliable prediction outputs;

[0069] Step 4: Build a stacking ensemble model based on LSTM and EBRB: Use the constructed LSTM model and EBRB model as base learners, the linear regression model as the meta-learner, and then combine the outputs of the base learners to train the meta-learner.

[0070] Step 5: Input the data into the Stacking integrated model to obtain the photovoltaic power generation power prediction results.

[0071] In step 1, the influencing factors of photovoltaic power generation are extracted based on the extreme gradient boosting algorithm. The feature importance score provided by the XGBoost model is used to evaluate the contribution of each feature to the photovoltaic power generation prediction. The features are sorted by importance, and the top 5 features are selected as the independent variables of the model.

[0072] In step 2, an LSTM model is trained to capture the nonlinearities and long-term dependencies in the photovoltaic power generation time series. To build the LSTM model, we first set up an input layer that matches the shape of the training data. We then added an LSTM layer to capture the long-term dependencies in the photovoltaic power generation data. A Dropout layer was then used to reduce overfitting. Finally, a fully connected layer was used as the output layer to predict the target variable. The model was then compiled using the Adam optimizer and the mean squared error loss function.

[0073] In step 3, we built an EBRB model set, used EBRB to build an evaluation model, and used the ER algorithm as the inference engine of the model. The modeling process is as follows:

[0074] (1) Description of EBRB Knowledge Base

[0075] The model is constructed by IF-THEN rules, where the k-th extended confidence rule is defined as follows:

[0076]

[0077] Where Aki represents the candidate value of the i-th premise attribute in the k-th extended confidence rule; T k represents the number of premise attributes in the kth extended confidence rule; β n,k Denotes the attribute D assigned to the result in the k-th extended confidence rule n Confidence, if ΣNn=1β n,k <1, the information contained in the kth extended confidence rule is said to be complete, otherwise the information contained in the kth extended confidence rule is incomplete; δ k,i (i=1,…,T k ) represents the attribute weight of the i-th premise attribute in the k-th extended confidence rule; θ k represents the rule weight of the kth extended confidence rule.

[0078] (2) EBRB reasoning process

[0079] For the EBRB model, the ER algorithm is often used as the inference engine. The specific EBRB model inference process is as follows:

[0080] Step 1: Initialize the values ​​of key parameters in EBRB. Use expert knowledge to give all candidate values ​​of all premise attributes, the utility values ​​of all evaluation levels of the result attributes, and the attribute weights of all premise attributes.

[0081] Step 2: Calculate individual matching. Assume there are T groups of data, expressed as <x i ,y i >(i=1,…,T), and its candidate value set is represented by A i ={A i,j ; j=1,…,J i}, that is, all evaluation levels of various indicators that affect photovoltaic power generation; among them, x i represents the input value vector of the i-th premise attribute, i.e., the data of various indicators that affect photovoltaic power generation; i represents the i-th output value, i.e., photovoltaic power generation; then, based on the rule-based information conversion technology, the individual matching degree of each candidate value is calculated. The corresponding formula is as follows:

[0082]

[0083] After calculating the individual matching degrees of all candidate values, the individual matching degrees can be expressed through a more clear and intuitive distributed framework. The corresponding formula is as follows:

[0084] S(x i )={(A ij , α ij );i=1,2,...,T;j=1,2,...,J i}

[0085] Among them, α i,j Indicates the jth candidate value A in the i-th premise attribute i,j individual matching degree.

[0086] Step 3: Calculate the activation weight. Integrate the rule weight, attribute weight, and individual matching degree to obtain the activation weight of the kth extended confidence rule:

[0087]

[0088] Among them, αk i represents the individual matching degree of the i-th premise attribute candidate value in the k-th extended confidence rule. k When >0, it means that the kth extended confidence rule is activated.

[0089] Step 4: Apply the ER algorithm to synthesize confidence. All activation rules are synthesized at once using the analytical formula in the ER algorithm. The corresponding formula for the synthesized new confidence is as follows:

[0090]

[0091] When the input value is x={x1,…,x M}, the predicted value based on EBRB inference is:

[0092] f(x)={(D i , β i (x)), i=1, 2,...,N}.

[0093] In step 4, the integrated model is optimized by combining the characteristics of multiple models. The specific implementation is as follows:

[0094] (1) Data division

[0095] Divide the time series data into training and test sets, usually in chronological order. Let T be the total number of time periods, T train is the number of time periods in the training set, T test is the number of time periods of the test set, then the training set and test set can be expressed as:

[0096] D train ={X1,X2,...,X T_train}

[0097] D test ={X T_train+1 ,X T_train+2 ,...,X T}

[0098] (2) Base learner training

[0099] For each base learner M i Train on the training set. Let M i is the i-th base learner, θ i Its parameters. The training process can be expressed as:

[0100]

[0101] Where L is the loss function and Yt is the true label at time t.

[0102] (3) Cross-validation prediction

[0103] For each base learner M i , using time series cross validation, D train Divide into K folds, D train (k) is the kth fold. For each fold k, the training set and the validation set are divided, where the validation set data volume is 10% of the original training set. train (k) training set training M i , and in D train (k) is predicted on the validation set, and K prediction results are obtained, which are recorded as Y' t (k).

[0104] (4) Combination learner training

[0105] Use the predicted values ​​of the base learners on the validation set as input features to train the combined learner. Let Y' t (i) is the predicted value of the i-th base learner at time t, then the new feature vector Z t =[Y' t (1),Y' t (2),...,Y' t (N)], where N is the number of base learners. Combined learner M comb The parameter θ comb Train by:

[0106]

[0107] Using the combined learner M comb Make predictions on the test set and get Y' test .

[0108] In step five, by combining the LSTM model and the EBRB model, a photovoltaic power generation prediction model with higher prediction accuracy and stability is constructed. By inputting data into the model, the photovoltaic power generation power prediction value can be obtained.

[0109] To verify the effectiveness of the present invention, an experimental verification was conducted using the measured data of the Yulara photovoltaic power station of the Australian Desert Knowledge Solar Energy Center as an example. The main process included the following:

[0110] (1) Extracting factors affecting photovoltaic power generation based on feature engineering

[0111] First, the data was read using the Pandas library in Python. Data was cleaned by filling missing values, replacing outliers, and deleting duplicate records. Next, the extreme gradient boosting algorithm was used to iteratively build a new tree model and rank the importance of the resulting features. The top five influencing factors were selected as independent variables: air temperature, global radiation, air pressure, module temperature, and solar radiation intensity. The corresponding photovoltaic power generation was used as the dependent variable to build the model.

[0112] (2) Constructing an LSTM model for photovoltaic power generation prediction

[0113] Introduce the LSTM model for modeling, set the input layer to match the shape of the training data, add the LSTM layer, Dropout layer, and fully connected layer in sequence, and configure the Adam optimizer and mean squared error loss function to compile the model.

[0114] (3) Constructing an EBRB model for photovoltaic power generation prediction

[0115] Initialize the values ​​of the key parameters in EBRB, and obtain the evaluation levels and utility values ​​of the premise attributes and result attributes, as well as the attribute weights of the premise attributes, as shown in Table 1. Based on the set parameter values, calculate the individual matching degree and activation weight, and synthesize the confidence according to the ER algorithm.

[0116] Table 1 Evaluation levels and utility values ​​of each attribute

[0117]

[0118]

[0119] (4) Building a Stacking integration model based on LSTM and EBRB

[0120] First, the training set and test set are divided into 7:3 ratios, with the first 70% of the data used as the training set and the last 30% of the data as the test set. Second, the training set is divided into 6 parts using the time series cross-validation method. Each division ensures the temporal continuity of the divided data. With 40% as the starting ratio and 10% as the moving step ratio, 6 different training-validation pairs are obtained. The validation set data size of each pair is 10% of the original training set. Then, the constructed LSTM model and EBRB model are used as base learners, and the linear regression model is used as the meta-learner to construct a stacking ensemble model based on LSTM and EBRB.

[0121] (5) Input data into the Stacking ensemble model to obtain prediction results

[0122] The photovoltaic power generation and its meteorological data were input, and the Stacking integrated model was trained and tested. The obtained partial photovoltaic power generation prediction values ​​are shown in Table 2.

[0123] Table 2 Predicted values ​​of some photovoltaic power generation on January 23

[0124]

[0125]

[0126] The predicted values ​​were compared with the true values ​​and the R2 was 0.949, indicating that the prediction effect of this model is good and can be applied to photovoltaic power generation prediction.

[0127] This paper proposes a photovoltaic power generation prediction model based on LSTM and EBRB. The model combines LSTM and EBRB to construct a Stacking integration model based on LSTM and EBRB, which effectively improves the interpretability and prediction accuracy of the model. This study provides support for the further application and development of LSTM and EBRB in the field of photovoltaic power generation prediction.

[0128] The present invention also provides a photovoltaic power generation prediction system based on LSTM and EBRB, including a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0129] The present invention also provides a computer-readable storage medium on which computer program instructions that can be executed by a processor are stored. When the processor executes the computer program instructions, the method steps described above can be implemented.

[0130] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0131] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other manner. Any person skilled in the art may utilize the above-disclosed technical content to modify or modify the present invention into equivalent embodiments. However, any simple modifications, equivalent variations, and modifications to the above embodiments that do not depart from the technical content of the present invention and are based on the technical essence of the present invention remain within the scope of protection of the present invention.

Claims

1. A photovoltaic power generation prediction method based on LSTM and EBRB, characterized in that: include Step 1: Extract the factors affecting photovoltaic power generation based on feature engineering: Use the extreme gradient boosting algorithm to extract the key factors affecting photovoltaic power generation and construct a key feature set with high correlation and representativeness; Step 2: Build a prediction model based on LSTM: Use the extracted key features to train the LSTM model to capture the nonlinearity and long-term dependencies in the PV power time series; Step 3: Build an inference model based on EBRB: Build an EBRB model based on the extracted key features and their distribution, evaluate the uncertainty and complexity of the data, and generate reliable prediction outputs; Step 4: Build a Stacking ensemble model based on LSTM and EBRB: Use the built LSTM model and EBRB model as base learners, use the linear regression model as a meta-learner, and then combine the outputs of the base learners to train the meta-learner. Step 5: Input the data into the Stacking integrated model to obtain the photovoltaic power generation power prediction results.

2. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 1 is characterized in that: In step 1, the influencing factors of photovoltaic power generation are extracted based on the extreme gradient boosting algorithm, and the contribution of each feature to the prediction of photovoltaic power generation is evaluated using the feature importance score provided by the XGBoost model. The features are sorted by importance, and the top five key features are selected as the independent variables of the model.

3. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 1 is characterized in that: In step 2, the LSTM model is trained to capture the nonlinearity and long-term dependencies in the photovoltaic power time series, that is, in the process of building the LSTM model, first set the input layer that matches the shape of the training data, add the LSTM layer to capture the long-term dependencies of the photovoltaic power generation data, and reduce overfitting through the Dropout layer. Finally, the fully connected layer is used as the output layer to predict the target variable, and the Adam optimizer and the mean square error loss function are configured to compile the model.

4. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 1 is characterized in that: In step 3, we build an EBRB model set, use EBRB to build an evaluation model, and use ER rules as the inference engine. The modeling process is as follows: (1) Description of EBRB-type knowledge base The EBRB model is constructed by IF-THEN rules; (2) EBRB model reasoning process For the EBRB model, the ER rule is used as the reasoning engine, and the photovoltaic power generation power prediction value is obtained based on the EBRB model reasoning.

5. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 4 is characterized in that: In the IF-THEN rule, the kth extended confidence rule is defined as follows: then{(D1, β 1，k ),(D2,β 2，k ),…,(D N ,b N，k )} with rule weightθ k and attribute weightδ k，i (i=1,2,…,T k ) Where Ak i represents the candidate value of the i-th premise attribute in the k-th extended confidence rule; T k represents the number of premise attributes in the kth extended confidence rule; β n，k represents the attribute D assigned to the result in the kth extended confidence rule n Confidence, if ∑N n=1β n，k <1, the information contained in the kth extended confidence rule is said to be complete, otherwise the information contained in the kth extended confidence rule is incomplete; k，i (i=1,...,T k ) represents the attribute weight of the i-th premise attribute in the k-th extended confidence rule; θ k represents the rule weight of the kth extended confidence rule.

6. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 5 is characterized in that: The EBRB model reasoning process is as follows: Step 1: Initialize the values ​​of key parameters in the EBRB model: use expert knowledge to give all candidate values ​​in all premise attributes, utility values ​​of all evaluation levels in the result attributes, and attribute weights of all premise attributes; Step 2: Calculate individual matching degree: Assume there are T groups of data, expressed as <x i ,y i >(i=1,...,T), and its candidate value set is denoted as A i ={A i，j ; j = 1, ..., J i }, that is, all evaluation levels of various indicators that affect photovoltaic power generation; among them, x i Represents the input value vector of the i-th premise attribute, that is, the data of various indicators that affect photovoltaic power generation; y i represents the i-th output value, i.e., photovoltaic power generation; then, the individual matching degree of each candidate value is calculated according to the rule-based information conversion technology; the corresponding formula is as follows: After calculating the individual matching degrees of all candidate values, the individual matching degrees are expressed through a distributed framework; the corresponding formula is as follows: S(x i )={(A ij ,α ij );i=1,2,…,T;j=1,2,…,J i } Among them, α i，j Represents the jth candidate value A in the i-th premise attribute i，j Individual matching degree; Step 3: Calculate the activation weight: Integrate the rule weight, attribute weight and individual matching degree to obtain the activation weight of the kth extended confidence rule: Among them, αk i represents the individual matching degree of the i-th premise attribute candidate value in the k-th extended confidence rule; when ω k When >0, it means that the kth extended confidence rule is activated; Step 4: Apply ER rules to synthesize confidence: All activated rules are synthesized at once through the analytical formula in the ER rules. The corresponding formula for the synthesized new confidence is as follows: When the input value is x={x1,...,x M }, the predicted value based on the EBRB model is: f(x)={(D i ,β i (x)),i=1,2,…,N}。 7. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 1 is characterized in that: Step 4 is implemented as follows: (1) Data Division Divide the time series data into training set and test set in chronological order; let T be the total number of time periods, T train is the number of time periods in the training set, T test is the number of time periods of the test set, then the training set and test set are expressed as: D train ={X1, X2, …, X T_train } D test ={X T_train+1 ,X T_train+2 ,...,X T } (2) Base learner training For each base learner M i Train on the training set; let M i is the i-th base learner, θ i For its parameters; The training process is expressed as: Where L is the loss function, X t is the data at time t in the training set, Y t is the true label at time t; (3) Cross-validation prediction For each base learner M i , using time series cross validation, D train Divide into K folds, D train (k) is the kth fold; for each fold k, the training set and the validation set are divided, where the validation set data volume is 10% of the original training set; using D train (k) training set training M i , and in D train (k) is predicted on the validation set, and K prediction results are obtained, denoted as Y' t (k); (4) Combination learner training Use the predicted values ​​of the base learner on the validation set as input features to train the combined learner; let Y' t (i) is the predicted value of the i-th base learner at time t, then the new feature vector Z t =[Y' t (1), Y' t (2), ..., Y' t (N)], where N is the number of base learners; the combination learner M comb The parameter θ comb Train by: Using the combined learner M comb Make predictions on the test set and get Y' test .

8. The photovoltaic power generation prediction method based on LSTM and EBRB according to claim 1 is characterized in that: In step five, the Stacking integrated model is a photovoltaic power prediction model with higher prediction accuracy and stability constructed by combining the LSTM model and the EBRB model.

9. A photovoltaic power generation prediction system based on LSTM and EBRB, characterized in that: The method comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor. When the processor executes the computer program instructions, the method steps as claimed in any one of claims 1 to 8 can be implemented.

10. A computer-readable storage medium having stored thereon computer program instructions that can be executed by a processor, and when the processor executes the computer program instructions, the method steps according to any one of claims 1 to 8 can be implemented.

Images