New energy power prediction method and system based on cluster division, equipment and medium
Through the new energy power prediction method based on cluster division, the problems of mutual influence between power stations and large-scale data processing are solved, and accurate prediction and efficient prediction efficiency of new energy power are achieved.
Patent Information
- Application Number
- CN202411695678.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-25
- Publication Date
- 2025-04-25
AI Technical Summary
The existing power plant power prediction methods ignore the mutual influence between power plants, resulting in inaccurate predictions, and individual predictions of each power plant are inefficient and difficult to cope with large-scale data processing.
A new energy power prediction method based on cluster division is adopted, and the historical data of the power station is obtained for preprocessing and feature selection. A cluster division algorithm is used to cluster the power stations in the target area, and a reference power station is determined, and a prediction model is built to output the power prediction curve.
Accurate prediction of new energy power is achieved, prediction efficiency is improved, large-scale data can be processed more effectively, and data from multiple power stations are analyzed to improve prediction accuracy.
Smart Images
Figure CN119884994A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a new energy power station and relates to a technical field, and in particular to a new energy power prediction method and system, equipment and medium based on cluster division. Background Art
[0002] In traditional technology, power plant power prediction methods are mainly divided into statistical methods and physical methods, and each method has its own advantages and disadvantages. The physical method does not require a large amount of measurement data, but requires an accurate mathematical description of the physical properties of the atmosphere and the characteristics of the power plant. These equations are difficult to solve and require a large amount of calculation. The statistical method does not require solving physical equations and has a fast calculation speed, but requires a large amount of historical data. At the same time, most of the current prediction manufacturers do not provide cluster power predictions. Some manufacturers calculate cluster power predictions by simply accumulating power plant power predictions, which has low accuracy and cannot meet the actual needs of guiding scheduling decisions.
[0003] like Figure 2 As shown in the figure, due to the lack of cluster prediction, the current power plant power prediction has the following technical problems:
[0004] Ignoring the mutual influence between power plants: There may be mutual dependence or influence between the power outputs of power plants, and these interactions may not be captured, resulting in inaccurate forecasts.
[0005] Inefficiency: Forecasting each plant individually requires more time and resources.
[0006] Difficulty in coping with large-scale data processing: When a large number of power plants and complex data are involved, forecasting each plant individually may become impractical.
[0007] Therefore, how to achieve accurate prediction of the total power of new energy in a larger range to avoid the above problems caused by single-point prediction has become an urgent problem to be solved by technical personnel in this field. Summary of the invention
[0008] The purpose of the present invention is to provide a new energy power prediction method and system, device and medium based on cluster division to solve the above problems in the prior art.
[0009] The embodiments of the present invention are implemented by the following technical solutions:
[0010] In a first aspect, the present invention provides a new energy power prediction method based on cluster division, comprising:
[0011] Obtain historical data of the power station and pre-process the historical data;
[0012] Obtain preprocessed historical data and perform feature selection;
[0013] Use a clustering algorithm to cluster the power stations in the target area and determine the benchmark power stations;
[0014] By acquiring the data of the benchmark power station to establish training samples, build a prediction model, train the prediction model through the training samples, and output the optimal prediction model;
[0015] The prediction model includes a cluster power prediction model and an electric energy data model, and outputs a power prediction curve based on the cluster power prediction model and the electric energy data model.
[0016] Preferably, the pretreatment comprises:
[0017] Identify outliers in the data and remove them;
[0018] Get the average of the two data points adjacent to the outlier on the left and right, and insert the average value into the empty space of the removed outlier.
[0019] Preferably, the clustering of power stations in the target area using a clustering algorithm includes:
[0020] Determine the number of clusters to be divided, randomly select K data points as the initial cluster centers, and for each power station, calculate its distance to each cluster center and assign it to the cluster with the nearest cluster center;
[0021] Calculate the average power output of each cluster's plants to update the location of the cluster center;
[0022] The steps of assigning data points and updating cluster centers are repeated until the positions of cluster centers no longer change significantly, or a predetermined number of iterations is reached.
[0023] Preferably, the clustering of power stations in the target area using a clustering algorithm includes:
[0024] Calculate the distance between each power station and construct a distance matrix;
[0025] Create the initial cluster: each power station acts as a separate cluster;
[0026] Merge the most similar clusters: According to the distance matrix, merge the two clusters with the closest distance to form a new cluster;
[0027] Update the distance matrix: calculate the distance between the new cluster and other clusters and update the distance matrix;
[0028] The steps of merging the most similar clusters and updating the distance matrix are repeated until all power stations are merged into one large cluster or a predetermined number of clusters is reached.
[0029] Preferably, the step of acquiring the data of a reference power station to establish a training sample includes:
[0030] Obtain data of a reference power station, and obtain relationship coefficients of a number of data through the data of the reference power station and the output power of the power station;
[0031] And set the threshold of the relationship coefficient to retain the information characteristics of the data corresponding to the relationship coefficient;
[0032] Select the information features of the cluster power plant power at the predicted moment as the reference sequence, and the information feature vectors of the cluster power plant power at the corresponding moment in the historical samples of the previous n days as the comparison sequence;
[0033] Select samples with high sample data integrity and accuracy, calculate the correlation between the comparison sequence and the reference sequence, sort the correlations from large to small, and select samples with a large correlation with the predicted moment as the final training samples.
[0034] Preferably, the constructed prediction model includes:
[0035] After the training samples are standardized or normalized, a data set is constructed and divided into a training set, a validation set, and a test set;
[0036] Input the training set into CNN for training to obtain the initial model;
[0037] The performance of the initial model is evaluated using the validation set, and the final prediction model is obtained after adjusting the parameters based on the evaluation results.
[0038] In a second aspect, the present invention further provides a new energy power prediction system based on cluster division, comprising:
[0039] The data processing module is configured to obtain historical data of the power station and preprocess the historical data; obtain the preprocessed historical data and perform feature selection;
[0040] A partitioning module is configured to use a cluster partitioning algorithm to perform cluster partitioning on the power stations in the target area and determine a benchmark power station;
[0041] The prediction module is configured to establish training samples by acquiring data from a benchmark power station, construct a prediction model, train the prediction model through the training samples, and output an optimal prediction model, wherein the prediction model includes a cluster power prediction model and an electric energy data model, and outputs a power prediction curve based on the cluster power prediction model and the electric energy data model.
[0042] A main control device is connected to the data processing module, the partitioning module and the prediction module, and the main control device is used to execute the above-mentioned new energy power prediction method based on cluster partitioning.
[0043] In a third aspect, the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned new energy power prediction method based on cluster division when executing the computer program.
[0044] In a fourth aspect, the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned new energy power prediction method based on cluster division.
[0045] The technical solution of the embodiment of the present invention has at least the following advantages and beneficial effects:
[0046] The above method provided by the present invention mainly includes preprocessing historical data; obtaining historical data after preprocessing and performing feature selection; clustering the power stations in the target area using a clustering algorithm and determining the benchmark power station; training the prediction model through training samples, outputting the optimal prediction model, and outputting the power prediction curve based on the cluster power prediction model and the electric energy data model. The above method can accurately predict the power of new energy through clustering, thereby realizing accurate power prediction in a larger range, and can more effectively process large-scale data, while analyzing the data of multiple power stations to improve prediction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0048] Figure 1 is a control flow chart of the present invention;
[0049] Figure 2 It is a prior art diagram of the present invention. DETAILED DESCRIPTION
[0050] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.
[0051] The division of modules in this application is a logical division. There may be other division methods when implemented in actual applications. For example, multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.
[0052] The modules or submodules described independently may be physically separated or not: they may be implemented by software or hardware, and some modules or submodules may be implemented by software, and the processor may call the software to implement the functions of these modules or submodules, and other modules or submodules may be implemented by hardware, such as by hardware circuits. In addition, some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.
[0053] Please refer to Figure 1-Figure 2 In a first aspect, the present invention provides a new energy power prediction method based on cluster division, comprising:
[0054] S101: Acquire historical data of the power station and pre-process the historical data;
[0055] Among them, the power station’s historical power output data, weather data, equipment status, etc. need to be cleaned and formatted to ensure data quality;
[0056] For example, identifying outliers in the data and removing them;
[0057] Get the average of the two data points adjacent to the outlier on the left and right, and insert the average value into the empty space of the removed outlier.
[0058] The identification of outliers may be to calculate the Z-score of each data point, and generally, a value with a Z-score greater than 3 or less than -3 is regarded as an outlier.
[0059] IQR (Interquartile Range): Calculate the first quartile (Q1) and the third quartile (Q3) of the data, and then calculate the IQR (Q3-Q1). Generally, values greater than Q3+1.5*IQR or less than Q1-1.5*IQR are considered outliers.
[0060] S102: Obtain preprocessed historical data and perform feature selection;
[0061] Select characteristics that have a significant impact on power output, such as time of day, weather conditions, historical power output, etc.
[0062] S103: clustering the power stations in the target area using a clustering algorithm, and determining a benchmark power station;
[0063] S104: Establishing training samples by acquiring data of a reference power station, constructing a prediction model, training the prediction model by using the training samples, and outputting an optimal prediction model;
[0064] S105: The prediction model includes a cluster power prediction model and an electric energy data model, and a power prediction curve is output based on the cluster power prediction model and the electric energy data model.
[0065] The above method provided by the present invention mainly includes preprocessing historical data; obtaining historical data after preprocessing and performing feature selection; clustering the power stations in the target area using a clustering algorithm and determining the benchmark power station; training the prediction model through training samples, outputting the optimal prediction model, and outputting the power prediction curve based on the cluster power prediction model and the electric energy data model. The above method can accurately predict the power of new energy through clustering, thereby realizing accurate power prediction in a larger range, and can more effectively process large-scale data, while analyzing the data of multiple power stations to improve prediction efficiency.
[0066] In an exemplary embodiment of the present invention, the clustering of power stations in the target area using a clustering algorithm includes:
[0067] S201: Determine the number of clusters to be divided, randomly select K data points as initial cluster centers, calculate the distance between each power station and each cluster center, and assign it to the cluster where the nearest cluster center is located;
[0068] Determining the number of clusters to be divided involves calculating the total variance of the K-means clustering, or the within-cluster sum of squares (WCSS), for different K values (usually starting from 1 and gradually increasing). Plot a graph of the relationship between K value and WCSS. As the K value increases, the WCSS gradually decreases because each cluster contains fewer and fewer data points. The ideal K value is the point where the rate of decrease of WCSS suddenly slows down, which is like the "elbow" of the graph, indicating that adding more clusters will not significantly improve the clustering results.
[0069] For each data point, calculate how tightly it fits within its own cluster (the a value) and how well it is separated from other clusters (the b value). Calculate the silhouette coefficient for each data point, which is the result of (ba) / max(a,b). The silhouette coefficient ranges from -1 to 1, with values closer to 1 indicating that the data point fits its cluster well and is well separated from other clusters. For each value of K, calculate the average silhouette coefficient for all data points. Select the K value with the highest average silhouette coefficient as the optimal number of clusters.
[0070] S202: Calculate the average value of the power output of each cluster to update the position of the cluster center;
[0071] S203: Repeat the steps of allocating data points and updating cluster centers until the position of the cluster center no longer changes significantly or a predetermined number of iterations is reached.
[0072] In an exemplary embodiment of the present invention, clustering power stations in a target area using a clustering algorithm includes:
[0073] Calculate the distance between each power station and construct a distance matrix;
[0074] Create the initial cluster: each power station acts as a separate cluster;
[0075] Merge the most similar clusters: According to the distance matrix, merge the two clusters with the closest distance to form a new cluster;
[0076] Update the distance matrix: calculate the distance between the new cluster and other clusters and update the distance matrix;
[0077] The steps of merging the most similar clusters and updating the distance matrix are repeated until all power stations are merged into one large cluster or a predetermined number of clusters is reached.
[0078] In an exemplary embodiment of the present invention, the step of acquiring data of a reference power station to establish a training sample includes:
[0079] S301: Acquire data of a reference power station, and obtain relationship coefficients of a plurality of data through the data of the reference power station and the output power of the power station;
[0080] Collect data on factors related to power station power output, such as historical power output data, weather data, equipment status data, etc., calculate the relationship coefficient between these data and power station power output, and obtain the correlation results; use the obtained correlation results as the analysis results of factors affecting the power of the group power stations.
[0081] In this embodiment, the relationship coefficient can be understood as the rate of change of the power station output power with the change of data. The larger the relationship coefficient is, the greater the impact of the data on the power station output power is.
[0082] S302: setting a threshold of the relationship coefficient, and retaining information features of data corresponding to a value greater than the relationship coefficient;
[0083] S303: Selecting information features of cluster power plant power at the predicted moment as a reference sequence, and selecting information feature vectors of cluster power plant power at the corresponding moment in the historical samples of the previous n days as a comparison sequence;
[0084] S304: Select samples with high sample data integrity and accuracy, calculate the correlation between the comparison sequence and the reference sequence, sort the correlations from large to small, and select samples with a large correlation with the prediction moment as the final training samples.
[0085] In an exemplary embodiment of the present invention, building a prediction model includes:
[0086] After the training samples are standardized or normalized, a data set is constructed and divided into a training set, a validation set, and a test set;
[0087] Input the training set into CNN for training to obtain the initial model;
[0088] The performance of the initial model is evaluated using the validation set, and the final prediction model is obtained after adjusting the parameters based on the evaluation results.
[0089] In this embodiment, the prediction model includes a cluster power prediction model and an electric energy data model, wherein the cluster power prediction model may be:
[0090]
[0091] Where P z is the output of the cluster power prediction model, ω ij is the output from the i-th input layer to the j-th hidden layer, λ a is the regularization coefficient, β j is the input of the j-th output layer, μ is the calculation weight, σ is the activation function, is the learning rate, η is the bias of the jth hidden layer, and N is the number of hidden layers.
[0092] The electric energy data model may be:
[0093] E z =A e ξ
[0094] In the formula, E z is the output of the electric energy data model, A e is the initial electric energy value, ξ is the rate of change.
[0095] In a second aspect, the present invention further provides a new energy power prediction system based on cluster division, comprising:
[0096] The data processing module is configured to obtain historical data of the power station and preprocess the historical data; obtain the preprocessed historical data and perform feature selection;
[0097] A partitioning module is configured to use a cluster partitioning algorithm to perform cluster partitioning on the power stations in the target area and determine a benchmark power station;
[0098] The prediction module is configured to establish training samples by acquiring data from a benchmark power station, construct a prediction model, train the prediction model through the training samples, and output an optimal prediction model, wherein the prediction model includes a cluster power prediction model and an electric energy data model, and outputs a power prediction curve based on the cluster power prediction model and the electric energy data model.
[0099] A main control device is connected to the data processing module, the partitioning module and the prediction module, and the main control device is used to execute the above-mentioned new energy power prediction method based on cluster partitioning.
[0100] For ease of understanding, the implementation process of the method provided by the present invention is briefly described below using a specific usage scenario as an example:
[0101] The power characteristics and influencing factors of cluster power plants are analyzed, and the concepts of cluster effective influencing factors and cluster power plant power are proposed. Through correlation analysis, factors that have a greater impact on cluster power plant power are found as important influencing factors of cluster power plant power at the prediction moment.
[0102] Extract information reflecting the power characteristics of cluster power plants. According to the analysis results of factors affecting the power of cluster power plants, select factors that have a greater impact on the power of cluster power plants as representative information features.
[0103] Determine the reference sequence and the comparison sequence. Select the information feature of the cluster power plant power at the predicted moment as the reference sequence, and the information feature vector of the cluster power plant power at the corresponding moment in the historical samples of the previous n days as the comparison sequence.
[0104] Select samples with high data integrity and accuracy, and use the correlation and association calculation formula to calculate the gray association between the comparison sequence and the reference sequence. Sort the associations from large to small, and select the samples with the largest association with the predicted moment as the final training samples.
[0105] Establish a prediction model. Select m samples with a large correlation with the prediction time as the final training samples. The input variable is the power information feature of the power plant at the prediction time, and the output variable is the predicted cluster power plant power.
[0106] By combining physical and statistical methods, an electric energy prediction model is established to output maximum and minimum output curves.
[0107] The feasibility of the model is evaluated by case analysis. The normalized mean absolute error (NMAEE) and the normalized root mean square error (NRMSEE) are selected as the basis for evaluating the prediction results.
[0108] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0109] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0110] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
Claims
1. A new energy power prediction method based on cluster division, characterized in that: include: Obtain historical data of the power station and pre-process the historical data; Obtain preprocessed historical data and perform feature selection; Use a clustering algorithm to cluster the power stations in the target area and determine the benchmark power stations; By acquiring the data of the benchmark power station to establish training samples, build a prediction model, train the prediction model through the training samples, and output the optimal prediction model; The prediction model includes a cluster power prediction model and an electric energy data model, and outputs a power prediction curve based on the cluster power prediction model and the electric energy data model.
2. The new energy power prediction method based on cluster division according to claim 1 is characterized in that: The pre-processing comprises: Identify outliers in the data and remove them; Get the average of the two data points adjacent to the outlier on the left and right, and insert the average value into the empty space of the removed outlier.
3. The new energy power prediction method based on cluster division according to claim 1 is characterized in that: The clustering algorithm is used to cluster the power stations in the target area, including: Determine the number of clusters to be divided, randomly select K data points as the initial cluster centers, and for each power station, calculate its distance to each cluster center and assign it to the cluster with the nearest cluster center; Calculate the average power output of each cluster's plants to update the location of the cluster center; The steps of assigning data points and updating cluster centers are repeated until the positions of cluster centers no longer change significantly, or a predetermined number of iterations is reached.
4. The new energy power prediction method based on cluster division according to claim 1 is characterized in that: The clustering algorithm is used to cluster the power stations in the target area, including: Calculate the distance between each power station and construct a distance matrix; Create the initial cluster: each power station acts as a separate cluster; Merge the most similar clusters: According to the distance matrix, merge the two clusters with the closest distance to form a new cluster; Update the distance matrix: calculate the distance between the new cluster and other clusters and update the distance matrix; The steps of merging the most similar clusters and updating the distance matrix are repeated until all power stations are merged into one large cluster or a predetermined number of clusters is reached.
5. The new energy power prediction method based on cluster division according to claim 1 is characterized in that: The step of acquiring the data of the reference power station to establish the training samples includes: Obtain data of a reference power station, and obtain relationship coefficients of a number of data through the data of the reference power station and the output power of the power station; And set the threshold of the relationship coefficient to retain the information characteristics of the data corresponding to the relationship coefficient; Select the information features of the cluster power plant power at the predicted moment as the reference sequence, and the information feature vectors of the cluster power plant power at the corresponding moment in the historical samples of the previous n days as the comparison sequence; Select samples with high sample data integrity and accuracy, calculate the correlation between the comparison sequence and the reference sequence, sort the correlations from large to small, and select samples with a large correlation with the predicted moment as the final training samples.
6. The new energy power prediction method based on cluster division according to claim 1 is characterized in that: The prediction models constructed include: After the training samples are standardized or normalized, a data set is constructed and divided into a training set, a validation set, and a test set; Input the training set into CNN for training to obtain the initial model; The performance of the initial model is evaluated using the validation set, and the final prediction model is obtained after adjusting the parameters based on the evaluation results.
7. A new energy power prediction system based on cluster division, characterized in that: include: A data processing module is configured to obtain historical data of the power station and pre-process the historical data; Obtain preprocessed historical data and perform feature selection; A partitioning module is configured to use a cluster partitioning algorithm to perform cluster partitioning on the power stations in the target area and determine a benchmark power station; A prediction module is configured to establish a training sample by acquiring data of a reference power station, construct a prediction model, train the prediction model by the training sample, and output an optimal prediction model, wherein the prediction model includes a cluster power prediction model and an electric energy data model, and output a power prediction curve based on the cluster power prediction model and the electric energy data model; A main control device is connected to the data processing module, the partitioning module and the prediction module, and the main control device is used to execute the new energy power prediction method based on cluster partitioning as described in any one of claims 1-6.
8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the new energy power prediction method based on cluster division according to any one of claims 1 to 6 is implemented.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the new energy power prediction method based on cluster division as described in any one of claims 1 to 6 is implemented.