Adjustable capacity analysis method and system for source load prediction

Abstract

The invention discloses an adjustable capacity analysis method and system for source load prediction, and relates to the field of electric power Internet of Things, and the method comprises the steps: obtaining output time sequence data and load time sequence data of distributed energy based on a monitoring system and a data collection platform, and obtaining an initial data set; taking the initial data set as an input layer, establishing a CNN-LSTM combined matrix model and setting model parameters; based on the initial data set, constructing an input sample matrix, and training a CNN-LSTM combined matrix model; based on the input sample matrix, outputting the adjustable capacity predicted by the model through a full connection layer of the CNN-LSTM combination matrix model; according to the method, complex modes and nonlinear relations can be identified, and time sequence characteristics and change trends in data can be effectively captured; the adaptability and robustness of the model are improved, so that the model can quickly respond to the change of load and distributed energy demand; the proposed CNN-LSTM combined matrix model architecture can automatically learn features from original data, and the demand for complex data preprocessing is reduced.

Description

Technical Field

[0001] The present invention relates to the field of electric power Internet of Things, and in particular to an adjustable capacity analysis method and system for source-load prediction. Background Art

[0002] In traditional energy systems, such as planning and optimization of power generation and transmission networks, physical modeling calculations are widely used; they can predict and optimize system performance based on reliable physical theories and principles, but their modeling process may have high complexity and limitations when faced with complex system behaviors or systems that are not fully understood. Data-driven methods rely on historical data and advanced data analysis techniques to process and analyze large amounts of real-time and historical data through machine learning, artificial intelligence, and statistical analysis to discover patterns and associations in system behavior. This approach is particularly suitable for complex systems such as virtual power plants and electric vehicle charging stations, where system behavior may be difficult to accurately describe using traditional physical models. The advantage of data-driven methods lies in their strong predictive power and flexible adaptability, which can update and optimize system strategies in real time to cope with dynamic changes. However, it has a high reliance on the quality and quantity of data, and the interpretability and transparency of the model are low.

[0003] Traditional statistical forecasting models show low forecasting accuracy when dealing with the high nonlinearity and uncertainty of renewable energy output. These models are usually unable to accurately capture complex temporal dependencies, which affects the stability and reliability of the system. In addition, existing methods have high computational complexity and slow data processing speed when processing large-scale data, which is particularly prominent in scenarios that require real-time adjustment. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is: how to improve the prediction accuracy of the adjustable capacity in the energy system.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides an adjustable capacity analysis method for source-load prediction, comprising:

[0008] Based on the monitoring system and data acquisition platform, the output time series data and load time series data of distributed energy are obtained to obtain the initial data set;

[0009] Use the initial data set as the input layer, build a CNN-LSTM combined matrix model and set the model parameters;

[0010] Based on the initial data set, construct the input sample matrix and train the CNN-LSTM combined matrix model;

[0011] Adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model based on the input sample matrix.

[0012] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0013] The establishment of the CNN-LSTM combined matrix model includes:

[0014] Setting a convolutional layer: The convolutional neural network maps the initial data set through weight sharing and local connection methods to extract relevant features. The relevant features include implicit features or implicit mapping relationships of the source-load data set for the adjustable capacity extracted by the neural network.

[0015] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0016] The constructing of the input sample matrix comprises:

[0017] The first 70% of the initial data set is used as the training set, and the last 30% is used as the test set. 70% of the initial data set is used as the training data set; the training data set is used to form a 2×2 input sample matrix to form the LSTM layer.

[0018] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0019] The training of the CNN-LSTM combined matrix model includes:

[0020] The prediction result is calculated through forward propagation, the error between the prediction result and the actual value is calculated using the loss function, and the model parameters are updated through the back propagation algorithm. The model is iteratively trained until the loss function converges or the preset number of training rounds is reached.

[0021] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0022] The forward propagation includes inputting the constructed input sample matrix into the CNN-LSTM combined matrix model; extracting features through the convolution layer, processing time series data through the LSTM layer, and integrating features through the fully connected layer; and calculating the prediction results;

[0023] The error between the predicted result and the actual value is calculated using the loss function.

[0024] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0025] The back propagation includes updating the model parameters through the back propagation algorithm, calculating the gradient and updating the weights and biases using the optimization algorithm;

[0026] Introduce the chain rule to calculate and update parameters layer by layer;

[0027] Repeat forward propagation, loss calculation, and back propagation for iterative training until the loss function converges or the preset number of training rounds is reached.

[0028] As a preferred solution of the adjustable capacity analysis method for source load prediction, wherein:

[0029] The adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model includes:

[0030] The fully connected layer is responsible for integrating the features processed by the convolutional layer, pooling layer and LSTM layer, and gathering comprehensive information from different features; the fully connected layer maps the high-dimensional feature space to the required low-dimensional space through linear transformation to achieve dimensionality degradation. The final output is the final output of the entire model, generating a vector representing the adjustable capacity to meet specific prediction needs; the tail of the CNN-LSTM combined matrix model is two parallel LSTM units, which correspond to the output of two parallel one-dimensional values, that is, a two-dimensional row vector; the fully connected layer degrades the dimensionality of the two-dimensional row vector output by the parallel LSTM units, and outputs the adjustable capacity predicted by the model.

[0031] In a second aspect, an embodiment of the present invention provides an adjustable capacity analysis system for source-load prediction, including:

[0032] An initial data set acquisition module is used to acquire the output time series data and load time series data of distributed energy based on the monitoring system and the data acquisition platform to obtain the initial data set;

[0033] The model initialization module is used to use the initial data set as the input layer, establish the CNN-LSTM combination matrix model and set the model parameters;

[0034] The training module is used to construct an input sample matrix based on the initial data set and train the CNN-LSTM combined matrix model;

[0035] The prediction module is used to predict the adjustable capacity based on the input sample matrix and the output model of the fully connected layer of the CNN-LSTM combined matrix model.

[0036] In a third aspect, an embodiment of the present invention provides a computing device, including:

[0037] Memory and processor;

[0038] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the adjustable capacity analysis method for source load prediction as described in any embodiment of the present invention.

[0039] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the adjustable capacity analysis method for source-load prediction.

[0040] The beneficial effects of the present invention are as follows: the present invention constructs a CNN-LSTM combined matrix model architecture, which combines the feature extraction capability of CNN and the time series modeling capability of LSTM, can identify complex patterns and nonlinear relationships, and effectively capture the time series features and change trends in the data; using source-load data as model input improves the adaptability and robustness of the model, which enables the model to quickly respond to changes in load and distributed energy demand; the proposed CNN-LSTM combined matrix model architecture can automatically learn features from raw data, reducing the need for complex data preprocessing. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0042] Figure 1 It is an overall flow chart of the adjustable capacity analysis method for source-load prediction according to the first embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of a CNN-LSTM combined matrix model of an adjustable capacity analysis method for source-load prediction according to the first embodiment of the present invention;

[0044] Figure 3 It is a schematic diagram of the structure of a convolutional neural network unit of the adjustable capacity analysis method for source load prediction described in the first embodiment of the present invention;

[0045] Figure 4 It is a schematic diagram of the fully connected layer structure of the adjustable capacity analysis method for source load prediction described in the first embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0049] Example 1

[0050] Reference Figure 1-Figure 4 Table 1 is a first embodiment of the present invention, which provides an adjustable capacity analysis method for source-load prediction, including:

[0051] S1: Based on the monitoring system and data acquisition platform, the output time series data and load time series data of distributed energy are obtained to obtain the initial data set;

[0052] S2: Use the initial data set as the input layer, build a CNN-LSTM combined matrix model and set model parameters;

[0053] S3: Based on the initial data set, construct the input sample matrix and train the CNN-LSTM combined matrix model;

[0054] S4: Adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model based on the input sample matrix.

[0055] For example, the initial data set is shown in Table 1:

[0056] Table 1 Source-load time series data set

[0057]

[0058] In the embodiment of the present application, establishing the CNN-LSTM combined matrix model in the above step S2 includes:

[0059] like Figure 2 As shown, a convolutional layer is set: the convolutional neural network maps the initial data set through weight sharing and local connection methods to extract relevant features, and the relevant features include implicit features or implicit mapping relationships of the source load data set for the adjustable capacity extracted by the neural network.

[0060] It should be noted that examples of relevant features are as follows: the higher the load, the lower the corresponding adjustable capacity; it is not a feature column in the data set; setting the convolution layer not only reduces the parameter dimension during model training, but also increases the calculation speed. The basic principle of convolutional neural network (CNN) is to extract data characteristics by constructing multiple filters. By performing layer-by-layer convolution and pooling operations on the input data, CNN can extract the topology and features between the data. Finally, these features are merged in the fully connected layer and the regression problem is solved through the activation function. Its structure is as follows Figure 3 shown.

[0061] In the embodiment of the present application, constructing the input sample matrix in the above step S3 includes:

[0062] The first 70% of the initial data set is used as the training set, and the last 30% is used as the test set. 70% of the initial data set is used as the training data set; the training data set is used to form a 2×2 input sample matrix to form the LSTM layer;

[0063] It should be noted that the LSTM layer is composed of 2×2 LSTM units to form an LSTM matrix, that is, the load characteristic items and the output characteristic items each correspond to a group of LSTM units with a sliding window of 2, and then the two types of characteristic items form a 2×2 LSTM matrix.

[0064] In the embodiment of the present application, the training of the CNN-LSTM combined matrix model in the above step S3 includes:

[0065] The prediction result is calculated through forward propagation, the error between the prediction result and the actual value is calculated using the loss function, and the model parameters are updated through the back propagation algorithm. The model is iteratively trained until the loss function converges or the preset number of training rounds is reached.

[0066] Forward propagation: Input the constructed input sample matrix into the CNN-LSTM combined matrix model; extract features through the convolution layer, process time series data through the LSTM layer, and integrate features through the fully connected layer; calculate the prediction results;

[0067] Calculate loss: Use the loss function to calculate the error between the predicted result and the actual value.

[0068] In the embodiment of the present application, the loss function used is mean square error; mean square error (MSE) is a commonly used loss function for regression problems, which has the following advantages:

[0069] Smooth and differentiable: MSE is continuously differentiable, which makes it easy to use optimization algorithms such as gradient descent to update parameters;

[0070] Sensitive to large errors: MSE has a larger penalty for large errors (predictions that deviate far from the actual value), which helps the model correct these errors faster.

[0071] Good mathematical properties: MSE has good mathematical properties, which facilitates theoretical analysis and practical applications.

[0072] In another possible implementation, the root mean square error (RMSE) may be used as the loss function; RMSE is the square root of MSE and has the same unit as the actual value, which is easy to explain.

[0073] In another possible implementation, the mean absolute error (MAE) may be used as the loss function. MAE is insensitive to outliers and is suitable for situations where there is large noise in the data.

[0074] In another possible implementation, Huber Loss may be used as the loss function. Huber Loss is expressed as MSE when the error is small and as MAE when the error is large. It combines the advantages of both and is insensitive to outliers.

[0075] Back propagation: Update model parameters through the back propagation algorithm, calculate gradients and use optimization algorithms to update weights and biases;

[0076] The chain rule is introduced to calculate and update parameters layer by layer.

[0077] In the embodiment of the present application, the optimization algorithm used is stochastic gradient descent SGD; SGD is a commonly used optimization algorithm with the following advantages:

[0078] High computational efficiency: SGD uses only one sample (or a small batch of samples) at a time to calculate gradients and update parameters, resulting in less computation and faster training.

[0079] Small memory footprint: Since only one sample is processed at a time, SGD has low memory requirements and is suitable for processing large-scale data sets.

[0080] Easy to parallelize: SGD can be easily parallelized and is suitable for training on multi-core or distributed environments.

[0081] In another possible implementation, SGD with Momentum may be used. Momentum introduces a momentum term during updating to help accelerate the convergence of SGD in relevant directions and suppress oscillations.

[0082] In another possible implementation, AdaGrad (Adaptive Gradient Algorithm) may also be used. AdaGrad adaptively adjusts the learning rate of each parameter and performs better for sparse features.

[0083] In another possible implementation, Adam (Adaptive Moment Estimation) may also be used. Adam combines the advantages of Momentum and RMSprop, adaptively adjusts the learning rate of each parameter, and introduces a momentum term, which has better convergence performance and stability.

[0084] Repeat forward propagation, loss calculation, and back propagation for iterative training until the loss function converges or the preset number of training rounds is reached.

[0085] Furthermore, after training is completed, you can use the validation set data to evaluate the model performance and observe the performance of the model on unseen data. Based on the validation results, adjust the model parameters and structure to optimize the model performance. You can choose to adjust the hyperparameters or add regularization methods.

[0086] In the embodiment of the present application, the adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model in the above step S4 includes:

[0087] like Figure 4 As shown in the figure, the fully connected layer is responsible for integrating the features processed by the convolutional layer, the pooling layer and the LSTM layer, and aggregating the comprehensive information from different features; the fully connected layer maps the high-dimensional feature space to the required low-dimensional space through linear transformation to achieve dimensionality degradation. The final output is the final output of the entire model, generating a vector representing the adjustable capacity to meet specific prediction needs; the tail of the CNN-LSTM combined matrix model is two parallel LSTM units, which correspond to the output of two parallel one-dimensional values, that is, a two-dimensional row vector; the fully connected layer degrades the dimensionality of the two-dimensional row vector output by the parallel LSTM units, and outputs the adjustable capacity predicted by the model.

[0088] The above is a schematic scheme of the adjustable capacity analysis method for source load prediction of this embodiment. It should be noted that the technical scheme of the adjustable capacity analysis system for source load prediction and the technical scheme of the adjustable capacity analysis method for source load prediction described above belong to the same concept. For details not described in detail in the technical scheme of the adjustable capacity analysis system for source load prediction in this embodiment, please refer to the description of the technical scheme of the adjustable capacity analysis method for source load prediction described above.

[0089] The adjustable capacity analysis system for source-load prediction in this embodiment includes:

[0090] An initial data set acquisition module is used to acquire the output time series data and load time series data of distributed energy based on the monitoring system and the data acquisition platform to obtain the initial data set;

[0091] The model initialization module is used to use the initial data set as the input layer, establish the CNN-LSTM combination matrix model and set the model parameters;

[0092] The training module is used to construct an input sample matrix based on the initial data set and train the CNN-LSTM combined matrix model;

[0093] The prediction module is used to predict the adjustable capacity based on the input sample matrix and the output model of the fully connected layer of the CNN-LSTM combined matrix model.

[0094] This embodiment further provides a computing device, which is applicable to the adjustable capacity analysis method for source-load prediction, and includes:

[0095] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the adjustable capacity analysis method for source load prediction proposed in the above embodiment.

[0096] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the adjustable capacity analysis method for source load prediction proposed in the above embodiment is implemented.

[0097] The storage medium proposed in this embodiment and the adjustable capacity analysis method for source load prediction proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An adjustable capacity analysis method for source load prediction, characterized in that: include: Based on the monitoring system and data acquisition platform, the output time series data and load time series data of distributed energy are obtained to obtain the initial data set; Use the initial data set as the input layer, build a CNN-LSTM combined matrix model and set the model parameters; Based on the initial data set, construct the input sample matrix and train the CNN-LSTM combined matrix model; Adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model based on the input sample matrix.

2. The adjustable capacity analysis method for source load prediction according to claim 1, characterized in that: The establishment of the CNN-LSTM combined matrix model includes: Setting a convolutional layer: The convolutional neural network maps the initial data set through weight sharing and local connection methods to extract relevant features. The relevant features include implicit features or implicit mapping relationships of the source-load data set for the adjustable capacity extracted by the neural network.

3. The adjustable capacity analysis method for source load prediction according to claim 2, characterized in that: The constructing of the input sample matrix comprises: The first 70% of the initial data set is used as the training set, and the last 30% is used as the test set. 70% of the initial data set is used as the training data set; the training data set is used to form a 2×2 input sample matrix to form the LSTM layer.

4. The adjustable capacity analysis method for source load prediction according to claim 3, characterized in that: The training of the CNN-LSTM combined matrix model includes: The prediction result is calculated through forward propagation, the error between the prediction result and the actual value is calculated using the loss function, and the model parameters are updated through the back propagation algorithm. The model is iteratively trained until the loss function converges or the preset number of training rounds is reached.

5. The adjustable capacity analysis method for source load prediction according to claim 4, characterized in that: The forward propagation includes inputting the constructed input sample matrix into the CNN-LSTM combined matrix model; extracting features through the convolution layer, processing time series data through the LSTM layer, and integrating features through the fully connected layer; Calculate the prediction results; The loss function is used to calculate the error between the predicted result and the actual value.

6. The adjustable capacity analysis method for source load prediction according to claim 5, characterized in that: The back propagation includes updating the model parameters through the back propagation algorithm, calculating the gradient and updating the weights and biases using the optimization algorithm; Introduce the chain rule to calculate and update parameters layer by layer; Repeat forward propagation, calculate loss, and back propagation for iterative training until the loss function converges or the preset number of training rounds is reached.

7. The adjustable capacity analysis method for source load prediction according to claim 6, characterized in that: The adjustable capacity predicted by the fully connected layer output model of the CNN-LSTM combined matrix model includes: The fully connected layer is responsible for integrating the features processed by the convolutional layer, pooling layer and LSTM layer, and gathering comprehensive information from different features; the fully connected layer maps the high-dimensional feature space to the required low-dimensional space through linear transformation to achieve dimensionality degradation. The final output is the final output of the entire model, generating a vector representing the adjustable capacity to meet specific prediction needs; the tail of the CNN-LSTM combined matrix model is two parallel LSTM units, which correspond to the output of two parallel one-dimensional values, that is, a two-dimensional row vector; the fully connected layer degrades the dimensionality of the two-dimensional row vector output by the parallel LSTM units, and outputs the adjustable capacity predicted by the model.

8. A system using the adjustable capacity analysis method for source load prediction as claimed in any one of claims 1 to 7, characterized in that: include: An initial data set acquisition module is used to acquire the output time series data and load time series data of distributed energy based on the monitoring system and the data acquisition platform to obtain the initial data set; The model initialization module is used to use the initial data set as the input layer, establish the CNN-LSTM combination matrix model and set the model parameters; The training module is used to construct an input sample matrix based on the initial data set and train the CNN-LSTM combined matrix model; The prediction module is used to predict the adjustable capacity based on the input sample matrix and the output model of the fully connected layer of the CNN-LSTM combined matrix model.

9. A computing device comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the adjustable capacity analysis method for source load prediction according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the adjustable capacity analysis method for source load prediction according to any one of claims 1 to 7.

Images