Power distribution network line loss prediction method and device based on meteorological factors

Abstract

The application is suitable for the technical field of line loss prediction, and provides a power distribution network line loss prediction method and device based on meteorological factors, which comprises the following steps: obtaining historical line loss data and a meteorological factor data set of a target area power distribution network, and preprocessing all meteorological factor data in the historical line loss data and the meteorological factor data set; decomposing the preprocessed historical line loss data based on a preset algorithm to obtain historical line loss components; inputting the historical line loss components and all preprocessed meteorological factor data into a feature extraction model of a line loss prediction network constructed in advance to obtain a correlation coefficient of each meteorological factor and line loss data; inputting the correlation coefficient and the historical line loss components into a prediction model of the line loss prediction network to obtain a line loss prediction result of the target area power distribution network. The application selects meteorological factors having a positive impact on prediction accuracy in an iterative manner, thereby improving the prediction accuracy of line loss prediction.

Description

Method and Device for Predicting Line Losses in Distribution Networks Based on Meteorological Factors Technical Field

[0001] This invention belongs to the field of line loss prediction technology, and particularly relates to a method and device for predicting line losses in distribution networks based on meteorological factors. Background Technology

[0002] With rapid economic development, my country's demand for electricity is increasing daily. The rapid growth of new energy power equipment is constantly impacting the stability of the power system, which has a crucial influence on national economic development. As the proportion of new energy in modern power systems continues to rise, the volatility and nonlinearity of line loss sequences are far greater than in traditional power systems, further increasing the difficulty of line loss prediction. Furthermore, the increasing prevalence of smart meters provides more possibilities and challenges for line loss prediction. Traditional prediction methods, including time series analysis, fuzzy regression, and support vector regression (SVR), suffer from inherent limitations, resulting in low prediction accuracy and failing to meet the line loss prediction requirements of the ever-evolving smart grid.

[0003] In recent years, Long Short-Term Memory (LSTM) networks have been widely used in power line loss prediction due to their excellent predictive performance. For example, experiments comparing LSTM networks with feedforward neural networks highlight the high predictive power and practicality of the LSTM model for short-term power line loss prediction. Optimizing the LSTM model using the Gray Wolf Optimizer (GWO) yielded optimal hyperparameters, effectively improving the accuracy of predicting nonlinear short-term line loss data. Furthermore, the BILSTM neural network prediction model was optimized using a combination of attention mechanisms and the Whale Optimization Algorithm (WOA), and compared with the commonly used Autoregressive Integrated Moving Average (ARIMA) model. Comparison of prediction results and metrics showed that the WOA-BILSTM model yielded better results for short-term line loss prediction. However, similar to other neural network models, the LSTM model suffers from limited generalization performance and over-reliance on empirical values ​​due to factors such as the number of neurons in the hidden layers and the learning rate. The WOA-BILSTM model yields better results for short-term line loss prediction. Although optimizations have been made to the prediction model, it does not effectively handle irregular and highly volatile data. It is evident that traditional single line loss prediction models do not deeply consider the coupling characteristics between factors influencing line loss, resulting in limited prediction results and excessively long prediction times. Therefore, how to effectively select features to improve prediction accuracy has become a current research hotspot.

[0004] Some studies have analyzed the characteristics of load forecasting by comprehensively considering meteorological factors, constructing optimal input features and optimizing forecast accuracy. Other studies have proposed that changes in meteorological factors and load may limit the accuracy of learning models, combining SVR models for load forecasting to improve model reliability and predictive performance. However, the above studies still use the traditional and inefficient correlation coefficient method for selecting meteorological factors, failing to effectively calculate feature weights, thus limiting the accuracy of the forecasting models. Therefore, a more accurate method for distribution network line loss forecasting that considers key meteorological factors is urgently needed. Summary of the Invention

[0005] This invention provides a method and apparatus for predicting line losses in distribution networks based on meteorological factors, in order to solve the problem of low accuracy in prediction models caused by the inability to obtain an accurate correlation between meteorological factors and line losses.

[0006] This invention is achieved through the following technical solution:

[0007] In a first aspect, embodiments of the present invention provide a method for predicting line losses in a distribution network based on meteorological factors, including:

[0008] Historical line loss data and meteorological factor datasets of the target area distribution network are acquired, and all meteorological factor data in the historical line loss data and the meteorological factor dataset are preprocessed; wherein, the meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor;

[0009] Based on a preset algorithm, the preprocessed historical line loss data is decomposed to obtain historical line loss components;

[0010] The historical line loss components and all preprocessed meteorological factor data are input into the feature extraction model of the pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data.

[0011] The correlation coefficient and the historical line loss components are input into the prediction model of the line loss prediction network to obtain the line loss prediction results of the target area distribution network.

[0012] Secondly, embodiments of the present invention provide a distribution network line loss prediction device based on meteorological factors, comprising:

[0013] The processing module is used to acquire historical line loss data and meteorological factor dataset of the distribution network in the target area, and to preprocess all meteorological factor data in the historical line loss data and the meteorological factor dataset; wherein, the meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor;

[0014] The decomposition module is used to decompose the preprocessed historical line loss data based on a preset algorithm to obtain historical line loss components.

[0015] The extraction module is used to input the historical line loss components and preprocessed meteorological factor data into the feature extraction model of the pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data.

[0016] The prediction module is used to input the correlation coefficient and the historical line loss component into the prediction model of the line loss prediction network to obtain the line loss prediction result of the target area distribution network.

[0017] This invention provides a method and apparatus for predicting line losses in a distribution network based on meteorological factors. The method involves acquiring historical line loss and meteorological factor datasets of the distribution network in a target area, preprocessing them, optimizing a variational mode decomposition algorithm, and then decomposing the preprocessed historical line losses using the optimized algorithm to obtain historical line loss components. A line loss prediction network, including a feature extraction model and a prediction model, is constructed. The processed historical line loss and meteorological datasets are input into the feature extraction network to obtain the correlation coefficient between each meteorological factor and the line loss data. These correlation coefficients and the historical line loss components are then input into the line loss prediction network to obtain the line loss prediction results for the target area. Optimizing the preprocessed data effectively improves data quality. The correlation coefficients between climate factors and line loss data are obtained through the feature extraction network, and line loss prediction is performed through the line loss prediction network, effectively improving the prediction accuracy of line loss prediction. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 is a flowchart illustrating a distribution network line loss prediction method based on meteorological factors according to an embodiment of the present invention.

[0020] Figure 2 is a flowchart of the ZOA algorithm for a distribution network line loss prediction method based on meteorological factors provided in an embodiment of the present invention;

[0021] Figure 3 is a flowchart of the implementation of a distribution network line loss prediction method based on meteorological factors provided in an embodiment of the present invention;

[0022] Figure 4 is a schematic diagram of the structure of the distribution network line loss prediction device based on meteorological factors provided in an embodiment of the present invention. Detailed Implementation

[0023] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0024] Figure 1 is a flowchart illustrating a distribution network line loss prediction method based on meteorological factors according to an embodiment of the present invention. Referring to Figure 1, the detailed description of the distribution network line loss prediction method based on meteorological factors is as follows:

[0025] S110: Obtain historical line loss data and meteorological factor dataset of the target area distribution network, and preprocess all meteorological factor data in the historical line loss data and meteorological factor dataset; wherein, the meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor.

[0026] In an optional embodiment, the historical line loss data and meteorological factor dataset are preprocessed, including:

[0027] The historical line loss data and meteorological factor datasets are standardized to obtain standard historical line loss data and standard meteorological factor datasets.

[0028] The standard historical line loss data and the standard meteorological factor dataset are divided into training set and test set according to the preset partitioning rules;

[0029] The standard historical line loss data and standard meteorological factor dataset in the training set are used as the preprocessed historical line loss data and meteorological factor dataset.

[0030] Among them, the standard historical line loss data is obtained by standardizing all historical line loss data, and the standard meteorological factor dataset is obtained by standardizing the data of each meteorological factor in the meteorological factor dataset.

[0031] Optionally, meteorological factors may include: temperature, humidity, azimuth, cloud opacity, dew point temperature, direct horizontal irradiance, direct normal irradiance, global horizontal irradiance, global tilted irradiance, precipitable water, relative humidity, surface pressure, wind direction, wind speed, and zenith angle. Other meteorological factors may also be included, and this application does not impose specific restrictions on them.

[0032] Preprocessing refers to standardization. Labeling all historical line loss data and meteorological factor data can effectively eliminate the influence of dimensions, enabling the method of this application to adapt to various types of datasets.

[0033] The standardized formula is:

[0034]

[0035] in, This is historical line loss data or meteorological factor data. For standardized data, This is the average of historical line loss data or meteorological factor data. The standard deviation is the historical line loss data or meteorological factor data.

[0036] S120: Based on a preset algorithm, the preprocessed historical line loss data is decomposed to obtain historical line loss components.

[0037] In an optional embodiment, the preprocessed historical line loss data is decomposed based on a preset algorithm to obtain historical line loss components, including:

[0038] The parameters of the variational mode decomposition algorithm are optimized to obtain the preset algorithm;

[0039] The preprocessed historical line loss data is decomposed based on a preset algorithm to obtain a preset number of historical line loss components.

[0040] Optionally, the default algorithm is ZOA-VMD (Zebra Optimization Algorithm-Variational Modal Decomposition) algorithm, which is the VMD algorithm optimized by ZOA.

[0041] As shown in Figure 2, the optimal parameters of the VMD algorithm can be obtained by optimizing the VMD algorithm using the ZOA algorithm. The ZOA algorithm is an optimization algorithm inspired by the natural behavior of zebras in the wild; it simulates the foraging behavior of zebras and their defensive behaviors against predators.

[0042] First, initialize the optimization object.

[0043] Second, the zebra population is initialized. From a mathematical perspective, each zebra is a candidate solution to the problem, and the plain where the zebras are located represents the search space. The position of each zebra in the search space determines the value of the decision variable. Therefore, each zebra participating in the ZOA algorithm can be modeled using a vector, where the elements of the vector represent the values ​​of the problem variables. The number of zebras can be mathematically modeled using a matrix, and the initial positions of the zebras in the search space are randomly assigned. The ZOA population matrix is ​​as follows:

[0044]

[0045] in, For zebra population, For the first zebra, For the first Only zebras, For the first Only zebras, The first question posed to the first zebra The first zebra proposed One question, The first zebra proposed One question, For the first The first question raised by the zebra For the first The first zebra proposed the first One question, For the first The first zebra proposed the first One question, For the first The first question raised by the zebra For the first The first zebra proposed the first One question, For the first The first zebra proposed the first One question, For the number of population members, This represents the number of decision variables.

[0046] Third, an evaluation function is established, and the optimal zebra is updated based on this function. In each iteration, the zebra's position and the objective function value are updated. The members participating in the ZOA algorithm are updated using two natural behaviors of zebras in the wild: foraging and predator defense. In the first stage, the population members are updated based on simulations of the zebras' foraging behavior. Among the members participating in the ZOA algorithm, the best member of the population is considered the pioneer zebra, leading other population members to its search space position within the population. Therefore, updating the zebra's position during the foraging phase can be mathematically modeled using the following formula.

[0047]

[0048]

[0049] in, For the first The new state of zebras based on the first stage, Let l be the value of the k-th generation zebra in the l-th dimension. Indicates the first stage. For the first Based on the first stage, the zebra Values ​​of each dimension The value of the objective function. As the pioneer zebra, A random number in the interval [0,1]. ,in, Let [the number be] another random number in the interval [0,1]. The closer the parameter is to 2, the more drastic the change.

[0050] Fourth, in the second phase, the positions of ZOA population members in the search space are updated by simulating the defensive behavior of zebras against predator attacks. Assume that one of the following two scenarios occurs with equal probability:

[0051] Scenario (1): The lion attacks the zebra, and the zebra chooses to run away;

[0052] Situation (2): Other predators attack the zebra, and the zebra chooses to attack.

[0053] In scenario (1), when a zebra is attacked by a lion, the zebra will seek shelter in its surroundings to avoid the lion's attack. The position state update can be modeled using the following formula:

[0054]

[0055] in, For the first Zebra's second phase Values ​​of each dimension This is the second phase. Let l be the value of the k-th generation zebra in the l-th dimension. The maximum number of iterations, This represents the total number of iterations in the zebra population. A constant of 0.01 A random number in the interval [0,1]. It is a random number within the interval [0,1].

[0056] In scenario (2), when other predators attack one of the zebras in the herd, the other zebras in the herd will move toward the attacked zebra and attempt to establish a defensive mechanism to attack the predator. In this case, the position state update can be modeled using the following formula:

[0057]

[0058] in, For the first Zebra's second phase Values ​​of each dimension This is the second phase. Let l be the value of the k-th generation zebra in the l-th dimension. The state of the zebra being attacked. A random number in the interval [0,1]. It is a random number within the interval [0,1].

[0059] Fifth, update the optimal position based on the calculation results and determine whether the updated optimal position satisfies the iteration conditions. The iteration conditions can be reaching the maximum number of iterations, the optimal position meeting the accuracy requirements, the fitness no longer changing, or other iteration conditions; there are no specific limitations. If the updated optimal position satisfies the iteration conditions, the optimal solution is saved. If the updated optimal position does not satisfy the iteration conditions, the process returns to the first stage to iteratively update the optimal position.

[0060] Optionally, historical line losses can be decomposed and defined as follows:

[0061]

[0062] in, For the first One center frequency; For the first One line loss modal component; This is the original load or line loss sequence; For partial derivative operators; It is a unit impulse function.

[0063] Based on the process and principles of VMD, it can be seen that the number of decompositions and the bandwidth parameter have a significant impact on the results. In this application, the minimum envelope entropy is selected as the fitness function to optimize the number of decompositions and the bandwidth parameter. The envelope entropy formula is:

[0064]

[0065]

[0066] in, For the minimum envelope entropy, The modal component represents the proportion or weight of the entire signal. For the sample size, This is the Hilt-modulated envelope sequence of the VMD modal components.

[0067] When using ZOA with minimum envelope entropy as the fitness function to optimize the number of decompositions and bandwidth parameters, if the ZOA parameters are set as shown in Table 1:

[0068] Table 1 ZOA Parameter Settings

[0069]

[0070] After the optimization process of ZOA-VMD, it can be concluded that the system tends to stabilize when the number of iterations is 12. At this time, the optimal value of the number of decompositions is 6, the optimal value of the bandwidth parameter is 1250, and the minimum envelope entropy is 0.31.

[0071] As shown in Figure 2, the historical line loss data after ZOA-VMD decomposition exhibits a significant periodicity, with a marked reduction in data volatility and randomness, thus improving prediction accuracy. The accuracy of signal decomposition can be verified through reconstruction error, calculated using the following formula:

[0072]

[0073] in, For reconstruction error, The original signal, For the reconstructed signal, For signal length, For time scale.

[0074] In traditional line loss prediction, Virtual Mode Decomposition (VMD) is typically used to decompose historical line loss data. However, the VMD parameters are often set manually, which fails to fully extract the periodicity and trend information from the historical line loss data. This application uses the ZOA algorithm with envelope entropy as the objective function to optimize the VMD parameters. This decomposes a single set of historical line loss data into multiple modal components, enabling the learning of sufficient effective information from historical line loss data even with a small dataset. This significantly improves the robustness of the prediction framework and lays a solid foundation for subsequent data prediction.

[0075] S130: Input the historical line loss components and all preprocessed meteorological factor data into the feature extraction model of the pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data.

[0076] In an optional embodiment, the pre-built line loss prediction network includes a feature extraction model and a prediction model.

[0077] Optionally, the feature extraction model is CNN-LSTM-Attention (Convolutional Neural Network-Long Short-Term Memory-Attention, combining a convolutional-long short-term memory neural network with an attention mechanism). The LSTM network forms the basis of the feature extraction model, specifically designed for time-series data. Through memory and forgetting mechanisms, it can better capture long-term dependencies in sequence data. The input gate updates unit information, the forget gate determines whether to retain or reset information, and the output gate controls the output. The computational flow of the LSTM information flow is as follows:

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] in, The output of the input gate determines how much of the input information is written into the cell state at the current moment; This represents the input at the current moment; This is the hidden state from the previous moment; The output of the forget gate determines how much information in the cell from the previous time step should be forgotten; This is the output at the current moment; Candidate cell state; This represents the current cell state. The hidden state for output; , , , This is the weight matrix of the model. , , , This is the offset term of the model; It is the sigmoid activation function. It is the hyperbolic tangent activation function.

[0085] In the feature extraction model, the CNN layer can perform feature extraction and recognition to learn local trend features of time series. This application mainly uses convolutional layers to effectively extract nonlinear local features from meteorological data and pooling layers to reduce the dimensionality of the data for sampling, in order to generate key feature information. After fusing the CNN network, it is used for feature extraction to capture the correlation in the input data. This upper-layer network that combines CNN and LSTM simplifies the feature engineering part in line loss prediction, while achieving high parallelism in the feature extraction process, thus improving the efficiency of training and feature selection. The one-dimensional convolution formula is as follows:

[0086]

[0087] in, To output data, For data length, For convolution kernel, For input data, It is biased.

[0088] In the feature extraction process of time series data, due to the long-term dependence and seasonal periodicity of line loss time series, incorporating a self-attention mechanism during feature extraction allows for the automatic learning and allocation of different weights based on the importance of different parts of the input sequence. This more effectively captures key information from the sequence data. The calculation formula for the attention mechanism is as follows:

[0089]

[0090] Where k is the length of the data sequence, For the final result obtained from the calculation, Calculate weights for the attention mechanism. This is the existing data.

[0091] A combined CNN-LSTM-Attention neural network is used as the feature extraction model, which exhibits better robustness and adaptability compared to traditional feature extraction methods. This method can accurately extract features from various types of historical line loss data, increasing the adaptability of the entire prediction framework and reducing its complexity.

[0092] In an optional embodiment, historical line loss components and preprocessed meteorological factor datasets are input into a pre-built feature extraction model of a line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data in the meteorological factor dataset, including:

[0093] The historical line loss components and all preprocessed meteorological factor data are transformed into time series; each time point of the time series includes the value of each historical line loss component and the value of each meteorological factor data.

[0094] The control feature extraction model learns according to the time series to obtain the correlation coefficient between each meteorological factor and each historical line loss component.

[0095] Optionally, the feature extraction model reads the historical line loss component values ​​and the data of each meteorological factor at each time point in a time series, learns and captures the long-term dependence between each meteorological factor and the historical line loss data, and outputs the correlation coefficient between the meteorological factor and the line loss data.

[0096] S140: Input the correlation coefficient and historical line loss components into the prediction model of the line loss prediction network to obtain the line loss prediction results of the target area distribution network.

[0097] In an optional embodiment, the correlation coefficient and historical line loss components are input into the prediction model of the line loss prediction network to obtain the line loss prediction results for the target area distribution network, including:

[0098] The correlation coefficient is converted into a weight parameter using a transformation formula.

[0099] The weight parameters and historical line loss components are input into the prediction model of the line loss prediction network to obtain multiple predicted values ​​for line loss components.

[0100] The predicted values ​​of multiple line loss components are superimposed to obtain the line loss prediction results of the distribution network in the target area.

[0101] In an optional embodiment, the conversion formula is:

[0102]

[0103] in, For the first The weighting parameters of each meteorological factor For the first The correlation coefficient of the meteorological factors This represents the total number of meteorological factors.

[0104] Optionally, the prediction model is GA-BIGRU-LSTM (Genetic Algorithm-Bidirectional Gated Recurrent Unit-Long Short-Term Memory). First, the genetic algorithm generates an initial population consisting of multiple individuals, each representing a potential solution for the prediction model parameters. The fitness of each individual, i.e., the prediction evaluation function of the model, is evaluated to determine which individuals will be selected to generate the next generation. Second, a crossover process is introduced, generating offspring by exchanging chromosome segments from parent individuals. Random mutations are then introduced into the population through mutation operations to increase population diversity and prevent getting trapped in local optima. Finally, after the crossover and mutation operations, the fitness of the generated offspring is evaluated, and they are used to replace a portion of the current population, ensuring the population evolves towards the optimal solution. Through repeated selection, crossover, mutation, and replacement until the maximum number of iterations is met, the genetic algorithm continuously optimizes the population and gradually calculates the optimal solution for the prediction model parameters.

[0105] The BIGRU layer contains a two-way recurrent structure: one processes the sequence from front to back, and the other from back to front, simultaneously updating and outputting the hidden states in both directions, and concatenating them as the final feature representation. An LSTM layer, as the subsequent processing layer, further extracts complex models and dependencies in the sequence, and controls the flow of information through three gating units (input gate, forget gate, and output gate), effectively handling the vanishing and exploding gradient problems, thus improving the prediction accuracy of the second-layer prediction model for long-length sequences with linear loss. After obtaining the prediction results of each component, they are linearly superimposed in a fully connected layer to obtain the final linear loss prediction result. The linear loss prediction result is used to calculate the evaluation index and backpropagated to the first-layer feature extraction network to update the weight coefficients.

[0106] GRU combines the forget gate and output gate into an update gate, which can effectively save time when training large amounts of data. The calculation formulas for the update gate, reset gate, and hidden state in the GRU network are as follows:

[0107]

[0108]

[0109]

[0110]

[0111] in, The output of the update gate in the model determines how much information from the previous hidden state needs to be retained, and combines it with the information from the current time step to generate a new hidden state; This is the input for the current moment; To reset the door's output, the amount of information from the previous hidden state that would be forgotten was controlled. The candidate hidden state is calculated by combining the current input and the hidden state at the previous moment before resetting the control of the gate; The final hidden state is controlled by the update gate, which performs a weighted average between retaining the hidden state information from the previous time step and the candidate hidden state information. , , This is the weight matrix of the model. , , This is the bias term of the model; It is the sigmoid activation function. It is the hyperbolic tangent activation function.

[0112] In BIGRU-LSTM, the selection of various parameters is crucial to prediction accuracy; optimal network hyperparameters cannot be chosen based on experience alone. Compared to other optimization algorithms, the GA algorithm possesses unique advantages and characteristics in hyperparameter selection, enabling it to explore a broad search space, exhibiting strong global search capabilities, and performing well in complex, multi-peak optimization problems. Algorithms that fuse GA with prediction networks can quickly and accurately find the optimal network parameters.

[0113] In an optional embodiment, the method further includes:

[0114] The error value is determined based on the line loss prediction results;

[0115] If the error value does not meet the preset conditions, then according to the preset rules, the historical line loss components and the meteorological factors corresponding to the target correlation coefficient are input into the line loss prediction network to re-predict the line loss, obtain the line loss prediction result, and return to the step of determining the error value based on the line loss prediction result.

[0116] If the error value meets the preset conditions, the previous line loss prediction result will be used as the final line loss prediction result.

[0117] The formula for calculating the error value is as follows:

[0118]

[0119] Where MAE represents the error value. The total number of data points. This represents the actual line loss value. This is the predicted value for line loss.

[0120] Optionally, the preset condition is: calculate the difference between the MAE obtained this time and the MAE calculated last time, and the MAE obtained this time is greater than the MAE calculated last time. That is, when the value of MAE shows an upward trend, the line loss prediction result is used as the final line loss prediction result.

[0121] In an optional embodiment, according to preset rules, the meteorological factor data corresponding to the historical line loss components and the target correlation coefficient are input into the line loss prediction network to re-predict line losses, including:

[0122] The correlation coefficients are arranged in ascending order, and the number of each meteorological factor is determined based on the ranking results;

[0123] Obtain the number of times line loss prediction is performed, and determine the target correlation coefficient based on the number of times;

[0124] The historical line loss components and the meteorological factor data corresponding to the target correlation coefficient are input into the line loss prediction network to re-predict line losses.

[0125] In this embodiment, the relationship between the number of line loss predictions and the target correlation coefficient is as follows: the target correlation coefficient for the nth line loss prediction is one or more correlation coefficients ranked in the top n-1.

[0126] After obtaining the line loss prediction results, the following formulas can be used to evaluate the prediction results:

[0127]

[0128]

[0129] Where MAPE is the mean absolute percentage error and RMSE is the root mean square error. To measure the number of times, This represents the actual line loss value. This is the predicted value for line loss.

[0130] The weighted correlation coefficients output by the feature extraction model, combined with the historical line loss components after ZOA-VMD decomposition, are used as input to the prediction model. The BIGRU layer contains a two-way recurrent structure: one processes the sequence from front to back, and the other from back to front, simultaneously updating and outputting the hidden states in both directions, and concatenating them as the final feature representation. LSTM, as the subsequent processing layer, further extracts complex models and dependencies in the sequence, and controls the flow of information through three gating units (input gate, forget gate, and output gate), effectively handling the vanishing and exploding gradient problems, improving the prediction accuracy of the prediction module for long line loss sequences. After obtaining the prediction results for each component, the prediction results of each component are linearly superimposed to obtain the final line loss prediction result. Finally, the line loss prediction result is used to calculate the evaluation index and backpropagated to the feature extraction model to update the correlation coefficients and weight coefficients. This process can exclude weakly correlated meteorological factors and meteorological factors that have an adverse impact on the prediction results, making the line loss prediction results more accurate.

[0131] Traditional prediction networks, lacking feature engineering, cannot automatically select and optimize input features, leading to significant prediction accuracy deviations and poor adaptability to new datasets. In contrast, the meteorological factor-based distribution network line loss prediction method proposed in this application adaptively selects features and feeds back the correlation between meteorological factors and line loss data to the prediction model for re-prediction. Even with changes in the dataset, the line loss prediction method provided in this application can better perform line loss prediction and achieve the desired prediction results. Because the original data is preprocessed and the line loss prediction results can be validated and updated, this application can adapt to various datasets.

[0132] As shown in Figure 3, the implementation process of this application is as follows:

[0133] First, obtain historical line loss data and meteorological factor datasets for the target area's power distribution network.

[0134] Second, all historical line loss data and all data in the meteorological factor dataset are preprocessed to obtain standard historical line loss data and standard meteorological factor dataset.

[0135] Third, the standard historical line loss data and standard meteorological factor data are divided into training and testing sets. The standard historical line loss data in the testing set is used as the preprocessed historical line loss data, and the standard meteorological factor dataset in the testing set is used as the preprocessed meteorological factor dataset.

[0136] Fourth, the VMD algorithm is optimized using the ZOA algorithm, and the ZOA-VMD algorithm is used to decompose the preprocessed historical line loss data to obtain multiple historical line loss components.

[0137] Fifth, the historical line loss components and all meteorological factor data from the preprocessed meteorological factor dataset are input into the line loss prediction network. The correlation coefficient between each meteorological factor and the line loss data is obtained through the feature extraction model of the line loss prediction network (i.e., CNN-LSTM-Attention). The correlation is converted into weight parameters, and the line loss prediction result of the target area distribution network is obtained through the prediction model of the line loss prediction network (i.e., GA-BIGRU-LSTM), and the error value is calculated.

[0138] Sixth, sort the correlation coefficients between meteorological factor data and line loss data in descending order.

[0139] Seventh, the preprocessed meteorological factor data corresponding to the correlation coefficient ranked first and the preprocessed historical line loss data are re-input into the line loss prediction network to obtain new line loss prediction results and calculate new error values.

[0140] Eighth, compare the new error value with the error value of the line loss prediction result obtained in the previous line loss prediction.

[0141] If the new error value is smaller than the previous error value, the number of calculations n is obtained, and the data corresponding to the top n-1 meteorological factors in terms of correlation are input into the line loss prediction network to re-predict the line loss, so as to obtain the final line loss prediction result. The final line loss prediction result is then used as the line loss prediction result for the target area distribution network.

[0142] If the new error value is larger than the previous error value, the weights are re-determined based on the correlation coefficient of the previous error prediction, and the weights and historical line loss components are input into the prediction model for prediction.

[0143] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0144] Corresponding to the meteorological factor-based distribution network line loss prediction method described in the above embodiments, Figure 4 shows a schematic diagram of a meteorological factor-based distribution network line loss prediction device provided by an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown.

[0145] Referring to Figure 4, a distribution network line loss prediction device 4 based on meteorological factors in an embodiment of the present invention may include:

[0146] Processing module 41 is used to acquire historical line loss data and meteorological factor dataset of the target area distribution network, and to preprocess all meteorological factor data in the historical line loss data and meteorological factor dataset; wherein, the meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor;

[0147] The decomposition module 42 is used to decompose the preprocessed historical line loss data based on a preset algorithm to obtain historical line loss components.

[0148] Extraction module 43 is used to input historical line loss components and preprocessed multiple meteorological factor data into the feature extraction model of the pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data;

[0149] The prediction module 44 is used to input the correlation coefficient and historical line loss components into the prediction model of the line loss prediction network to obtain the line loss prediction results of the target area distribution network.

[0150] In one possible implementation, processing module 41 is specifically used for:

[0151] The historical line loss data and meteorological factor datasets are standardized to obtain standard historical line loss data and standard meteorological factor datasets.

[0152] The standard historical line loss data and the standard meteorological factor dataset are divided into training set and test set according to the preset partitioning rules;

[0153] The standard historical line loss data and standard meteorological factor dataset in the training set are used as the preprocessed historical line loss data and meteorological factor dataset.

[0154] In one possible implementation, the pre-built line loss prediction network includes a feature extraction model and a prediction model.

[0155] In one possible implementation, decomposition module 42 is specifically used for:

[0156] The parameters of the variational mode decomposition algorithm are optimized to obtain the preset algorithm;

[0157] The preprocessed historical line loss data is decomposed based on a preset algorithm to obtain a preset number of historical line loss components.

[0158] In one possible implementation, extraction module 43 is specifically used for:

[0159] The historical line loss components and all preprocessed meteorological factor data are transformed into time series; each time point of the time series includes the value of each historical line loss component and the value of each meteorological factor data.

[0160] The control feature extraction model learns according to the time series to obtain the correlation coefficient between each meteorological factor and each historical line loss component.

[0161] In one possible implementation, the prediction module 44 is specifically used for:

[0162] The correlation coefficient is converted into a weight parameter using a transformation formula.

[0163] The weight parameters and historical line loss components are input into the prediction model of the line loss prediction network to obtain multiple predicted values ​​for line loss components.

[0164] The predicted values ​​of multiple line loss components are superimposed to obtain the line loss prediction results of the distribution network in the target area.

[0165] In one possible implementation, the transformation formula is:

[0166]

[0167] in, For the first The weighting parameters of each meteorological factor For the first The correlation coefficient of the meteorological factors This represents the total number of meteorological factors.

[0168] In one possible implementation, the prediction module 44 is also used for:

[0169] The error value is determined based on the line loss prediction results;

[0170] If the error value does not meet the preset conditions, then according to the preset rules, the historical line loss components and the meteorological factors corresponding to the target correlation coefficient are input into the line loss prediction network to re-predict the line loss, obtain the line loss prediction result, and return to the step of determining the error value based on the line loss prediction result.

[0171] If the error value meets the preset conditions, the previous line loss prediction result will be used as the final line loss prediction result.

[0172] In one possible implementation, the prediction module 44 is also used for:

[0173] The correlation coefficients are arranged in ascending order, and the number of each meteorological factor is determined based on the ranking results;

[0174] Obtain the number of times line loss prediction is performed, and determine the target correlation coefficient based on the number of times;

[0175] The historical line loss components and the meteorological factor data corresponding to the target correlation coefficient are input into the line loss prediction network to re-predict line losses.

[0176] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0177] Those skilled in the art will recognize that the templates, units, and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0178] If the module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above embodiments of the distribution network line loss prediction method based on meteorological factors. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0179] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting line losses in a distribution network based on meteorological factors, characterized in that, include: Historical line loss data and meteorological factor datasets of the target area distribution network are acquired, and all meteorological factor data in the historical line loss data and meteorological factor datasets are preprocessed. The meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor. The preprocessed historical line loss data is decomposed based on a preset algorithm to obtain historical line loss components. The historical line loss components and all preprocessed meteorological factor data are input into a pre-constructed feature extraction model of a line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data. The correlation coefficients and the historical line loss components are input into the prediction model of the line loss prediction network to obtain the line loss prediction result of the target area distribution network. The prediction model is GA-BIGRU-L. STM; The step of decomposing the preprocessed historical line loss data based on a preset algorithm to obtain historical line loss components includes: optimizing the parameters of the variational mode decomposition algorithm to obtain a preset algorithm; decomposing the preprocessed historical line loss data based on the preset algorithm to obtain a preset number of historical line loss components; wherein, the preset algorithm is the ZOA-VMD algorithm; The step of decomposing the preprocessed historical line loss data based on the preset algorithm to obtain a preset number of historical line loss components includes: optimizing the parameters of the VMD algorithm based on the ZOA algorithm using a preset optimization formula and a preset number of iterations to obtain a parameter-optimized VMD algorithm; inputting the preprocessed historical line loss data into the parameter-optimized VMD algorithm to obtain a preset number of historical line loss components; the preset optimization formula is: in, For the first Based on the second stage, the zebra Values ​​of each dimension This is the second phase. Let l be the value of the k-th generation zebra in the l-th dimension. The state of the zebra being attacked. The maximum number of iterations, This represents the total number of iterations in the zebra population. A constant of 0.01 A random number in the interval [0,1]. It is a random number within the interval [0,1].

2. The distribution network line loss prediction method based on meteorological factors as described in claim 1, characterized in that, The method further includes: determining an error value based on the line loss prediction result; if the error value does not meet a preset condition, then according to a preset rule, inputting multiple meteorological factor data corresponding to the historical line loss components and the target correlation coefficient into the line loss prediction network to re-predict the line loss, obtaining the line loss prediction result, and returning to the step of determining the error value based on the line loss prediction result; if the error value meets the preset condition, then using the previous line loss prediction result as the final line loss prediction result.

3. The distribution network line loss prediction method based on meteorological factors as described in claim 1, characterized in that, The step of inputting the historical line loss components and all preprocessed meteorological factor data into the feature extraction model of a pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data includes: converting the historical line loss components and all preprocessed meteorological factor data into a time series; each time point of the time series includes the value of each historical line loss component and the value of each meteorological factor data; controlling the feature extraction model to learn according to the time series to obtain the correlation coefficient between each meteorological factor and each historical line loss component.

4. The distribution network line loss prediction method based on meteorological factors as described in claim 1, characterized in that, The step of inputting the correlation coefficient and the historical line loss components into the prediction model of the line loss prediction network to obtain the line loss prediction result of the target area distribution network includes: converting the correlation coefficient into a weight parameter using a conversion formula; inputting the weight parameter and the historical line loss components into the prediction model of the line loss prediction network to obtain multiple line loss component prediction values; and superimposing the multiple line loss component prediction values ​​to obtain the line loss prediction result of the target area distribution network.

5. The distribution network line loss prediction method based on meteorological factors as described in claim 2, characterized in that, The step of inputting the historical line loss components and the meteorological factor data corresponding to the target correlation coefficient into the line loss prediction network to re-predict line losses according to preset rules includes: arranging the correlation coefficients in ascending order and determining the number of each meteorological factor based on the sorting result; obtaining the number of times line loss prediction is performed and determining the target correlation coefficient based on the number of times; and inputting the historical line loss components and the meteorological factor data corresponding to the target correlation coefficient into the line loss prediction network to re-predict line losses.

6. The distribution network line loss prediction method based on meteorological factors as described in claim 1, characterized in that, The preprocessing of the historical line loss data and all meteorological factor data in the meteorological factor dataset includes: standardizing the historical line loss data and the meteorological factor dataset to obtain standard historical line loss data and standard meteorological factor dataset; dividing the standard historical line loss data and the standard meteorological factor dataset into a training set and a test set according to a preset partitioning rule; and using the standard historical line loss data and the standard meteorological factor dataset in the training set as the preprocessed historical line loss data and meteorological factor dataset.

7. The distribution network line loss prediction method based on meteorological factors as described in claim 1, characterized in that, The pre-built line loss prediction network includes a feature extraction model and a prediction model.

8. The distribution network line loss prediction method based on meteorological factors as described in claim 4, characterized in that, The conversion formula is: in, For the first The weighting parameters of each meteorological factor For the first The correlation coefficient of the meteorological factors This represents the total number of meteorological factors.

9. A distribution network line loss prediction device based on meteorological factors, characterized in that, include: The processing module is used to acquire historical line loss data and meteorological factor datasets of the target area distribution network, and preprocess all meteorological factor data in the historical line loss data and the meteorological factor dataset; wherein, the meteorological factor dataset includes multiple meteorological factors and multiple meteorological factor data corresponding to each meteorological factor; the decomposition module is used to decompose the preprocessed historical line loss data based on a preset algorithm to obtain historical line loss components; the extraction module is used to input the historical line loss components and the preprocessed multiple meteorological factor data into the feature extraction model of a pre-constructed line loss prediction network to obtain the correlation coefficient between each meteorological factor and the line loss data; the prediction module is used to input the correlation coefficient and the historical line loss components into the line loss prediction network. The prediction model of the loss prediction network is used to obtain the line loss prediction results of the distribution network in the target area; the prediction model is GA-BIGRU-LSTM; the decomposition module is also used to optimize the parameters of the variational mode decomposition algorithm to obtain a preset algorithm; based on the preset algorithm, the preprocessed historical line loss data is decomposed to obtain a preset number of historical line loss components; wherein, the preset algorithm is the ZOA-VMD algorithm; the decomposition module is also used to optimize the parameters of the VMD algorithm based on the ZOA algorithm with a preset optimization formula and a preset number of iterations to obtain a parameter-optimized VMD algorithm; the preprocessed historical line loss data is input into the parameter-optimized VMD algorithm to obtain a preset number of historical line loss components; the preset optimization formula is: in, For the first Based on the second stage, the zebra Values ​​of each dimension This is the second phase. Let l be the value of the k-th generation zebra in the l-th dimension. The state of the zebra being attacked. The maximum number of iterations, This represents the total number of iterations in the zebra population. A constant of 0.01 A random number in the interval [0,1]. It is a random number within the interval [0,1].

Images