Distribution network line peak load prediction method and system based on combined model

Abstract

The invention discloses a distribution network line peak load prediction method and system based on a combined model, and relates to the technical field of power grids, and the method comprises the steps: collecting distribution network line data, extracting high-correlation input features, and carrying out the overall prediction of a load rate through a support vector machine; the peak load rate and the peak load rate occurrence time form a data set, independent prediction is carried out through a combined prediction model, and a prediction result of the support vector machine is corrected; and judging an early warning level according to the distribution network line load rate, and sending distribution network line weight and overload early warning information. According to the distribution network line peak load prediction method based on the combined model, the correlation analysis method is adopted to extract high-correlation input features, the model input quality and prediction precision are improved, the distribution network line weight and overload conditions are monitored and early warned in real time, the safety and reliability of power grid operation are improved, and the method is suitable for large-scale popularization and application. The method has better effects in the aspects of accuracy, safety and reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of power grids, and in particular to a distribution network line peak load prediction method based on a combined model. Background Art

[0002] With the continuous expansion of the global power grid and the increasing demand for electricity from users, the management and optimization of distribution lines has become an important topic in the research of power systems. In recent years, traditional load rate prediction methods have mostly adopted technologies based on statistical models and machine learning, such as time series analysis, regression models, and support vector machines. These methods have achieved certain results in predicting the overall trend of load rates. However, with the advancement of smart grid technology and the popularization of big data applications, power systems have put forward higher requirements for the accuracy and real-time performance of load rate prediction, especially during peak power demand periods such as peak summer. Accurately predicting the peak load rate of distribution lines and their occurrence time has become the key to optimizing power system scheduling and preventing line overload. Existing technologies mainly focus on overall prediction errors and rarely involve accurate prediction of peak load rates, resulting in low accuracy in heavy overload prediction, which further limits the operating efficiency and safety of power systems.

[0003] The existing load rate prediction technology has significant shortcomings in some specific applications. When dealing with load rate peak prediction, the traditional methods lack accurate analysis of the peak load rate and its occurrence time, resulting in the prediction results often failing to reflect the actual load conditions, causing inaccurate heavy overload warnings and affecting the stable operation of the power grid. Secondly, most of the existing methods use a single model for prediction, failing to fully utilize the advantages of multi-source data and multi-model combinations, and it is difficult to ensure high prediction accuracy in a complex and changeable power load environment. In addition, the existing prediction models do not adequately handle abnormal data, especially in load rate peak prediction, where outliers are often ignored or handled incorrectly, resulting in reduced robustness and accuracy of the prediction model. Summary of the invention

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

[0005] Therefore, the technical problem solved by the present invention is that the existing load forecasting methods have insufficient prediction accuracy, large limitations, improper abnormal data processing, and an optimization problem of how to improve the model's ability to respond to abnormal load changes.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: a distribution network line peak load prediction method based on a combined model, comprising collecting distribution network line data, extracting highly correlated input features, and making an overall prediction of the load rate through a support vector machine; forming a data set with the peak load rate and the time when the peak load rate occurs, making independent predictions through a combined prediction model, and correcting the prediction results of the support vector machine; determining the warning level according to the load rate of the distribution network line, and sending a distribution network line heavy overload warning information.

[0007] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, wherein: the extraction of high-correlation input features includes feature correlation analysis, collecting distribution network line input feature sample values, and based on the chaos theory correlation analysis method, by calculating the correlation coefficient between features, evaluating the similarity on a specific time series, capturing the nonlinear relationship between features, and providing analysis results, which are expressed as:

[0008]

[0009]

[0010] Among them, ρ ij is the basic correlation coefficient between the i-th input feature and the j-th input feature, x ik is the kth sample value of the ith input feature, y jk is the kth sample value of the jth input feature, is the mean of the i-th input feature, is the mean of the jth input feature, N is the number of samples, ||x ik -y jk || is the Euclidean distance between feature vectors, σ is a parameter used to control the influence of distance, and the highly correlated features are adaptively weighted to generate comprehensive features. The weights are dynamically adjusted according to the relevance of the features so that the predicted contribution of each feature matches the actual relevance, which is expressed as:

[0011]

[0012]

[0013] Among them, Feature w is the weighted comprehensive feature, w i is the weight of the i-th feature, Feature i is the i-th input feature, and m is the number of features.

[0014] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, the overall prediction of the load rate by the support vector machine includes a support vector machine algorithm, which maps the input data to a high-dimensional space through a kernel function, so that the linearly inseparable data becomes linearly separable in the high-dimensional space, which is expressed as:

[0015]

[0016]

[0017] Among them, f(x) is the prediction function, α i is the Lagrange multiplier in the support vector machine, x i is the i-th support vector, x is the input feature vector, b is the bias term in the support vector machine, K(x i ,x) is the input x and support vector x i The kernel function value between the two is used to introduce an error correction mechanism based on the support vector machine. The prediction results are dynamically adjusted through the Kalman filter. The prediction results are continuously updated and corrected according to the real-time observation data to minimize the prediction error, which is expressed as:

[0018]

[0019] in, is the corrected predicted value at time t+1, f(x t ) is the support vector machine prediction value at time t, y t is the observed value at time t, K t is the Kalman gain at the tth moment.

[0020] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, wherein: the peak load rate and the peak load rate occurrence time are combined into a data set, including generating a box plot data set, calculating the interquartile range of the data set, measuring the degree of variation of the data, identifying outliers in the data set, extracting outliers above the upper quartile, constructing a peak data set, performing multi-scale analysis on the peak data, capturing peak features on different time scales through wavelet transform, and identifying changes and emergencies in the signal, which is expressed as:

[0021]

[0022] Outliers = {x is |x is <Q1-λ×IQR W x i >Q3+λ×IQR W}

[0023] Among them, IQRW is the interquartile range calculated by wavelet transform, Q3 s and Q1 s are the third quartile at the sth scale and the first quartile at the sth scale, S is the number of scales of the wavelet transform, Outliers is the set of outliers, x is is the i-th data point, Q3 and Q1 are the third quartile and the first quartile respectively.

[0024] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, the independent prediction through the combined prediction model includes a combined prediction based on the attention mechanism, the recurrent neural network and the improved recurrent unit neural network, the weight of each feature is calculated through the attention weight matrix, and the weight is applied to the feature vector to generate a weighted feature vector, which is expressed as:

[0025] Att=softmax(W a ·h t )

[0026]

[0027] Among them, Att is the attention weight matrix, W a is the weight matrix of the attention mechanism, h t is the hidden state vector at time t, is the weighted hidden state vector. The recurrent neural network captures the temporal dependency in the time series data. The improved recurrent unit neural network introduces a gating mechanism and combines it with an adaptive filter based on the Monte Carlo method to dynamically correct the hidden state. The distribution of the hidden state is estimated through multiple sampling and weight adjustment, which is expressed as:

[0028]

[0029] Z t =σ(W z x t +U z h t-1 +b z )

[0030]

[0031] Among them, z t is the update gate, x t is the input vector at time t, U z is the weight matrix of the update gate, W z and W h is the weight matrix of the input vector, b z is the bias term of the update gate, is the candidate hidden state, U h is the weight matrix of the reset gate, b h is the bias term of the candidate hidden state, r t is the reset gate, P is the number of particles, is the candidate hidden state of the p-th particle, is the weight of the pth particle.

[0032] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, wherein: the warning level is determined according to the load rate of the distribution network line, including determining the load condition of the distribution network line, calculating the dynamic load rate, reflecting the load condition of the distribution network line within a preset time period, the dynamic load rate is the ratio of the peak power to the rated power, reflecting the load condition of the line, after obtaining the dynamic load rate, based on the preset judgment standard, the load condition is layered, and the load rate is divided into normal, heavy load and overload. When the load rate is between 0 and 0.8, it is normal, when the load rate is between 0.8 and 1.0, it is heavy load, and when the load rate exceeds 1.0, it is overload, expressed as:

[0033]

[0034]

[0035] Among them, L r (t) is the load rate at the tth moment, P p (t) is the peak power at the tth moment, P r () is the rated power at the tth moment, and WL(y) ​​is the warning level at the tth moment.

[0036] As a preferred solution of the distribution network line peak load prediction method based on the combined model described in the present invention, wherein: the sending of distribution network line heavy overload warning information includes sending warning information according to different warning levels, generating alarm information based on the load rate discrimination result, and adopting different sending strategies according to different warning levels. When the warning level is overload, a high priority sending strategy is adopted. When the warning level is heavy load, a normal priority sending strategy is adopted. When the load rate is within the preset normal range, no action is taken to avoid interference, which is expressed as:

[0037] Alert(t)=GA(WL(t))

[0038]

[0039] Among them, Alert(t) is the warning information at the tth moment, GA(x) is the function of generating warning information, SA(t) is the warning information sending action at the tth moment, Send(x) is the function of sending warning information, H_P is the high priority sending strategy, N_P is the normal priority sending strategy, and N_A means taking no action.

[0040] Another object of the present invention is to provide a distribution network line peak load prediction system based on a combined model, which can form a data set of peak load rate and peak load rate occurrence time through a combined prediction module, perform independent prediction through a combined prediction model, and correct the prediction results of a support vector machine, thereby solving the current problem of insufficient peak load rate prediction accuracy.

[0041] As a preferred solution of the distribution network line peak load prediction system based on the combined model described in the present invention, it includes: an extraction prediction module, a combined prediction module, and a discrimination and warning module; the extraction prediction module is used to collect distribution network line data, extract high-correlation input features, and perform an overall prediction of the load rate through a support vector machine; the combined prediction module is used to form a data set of peak load rate and peak load rate occurrence time, perform independent prediction through a combined prediction model, and correct the prediction result of the support vector machine; the discrimination and warning module is used to discriminate the warning level according to the distribution network line load rate, and send distribution network line heavy overload warning information.

[0042] A computer device includes a memory and a processor, wherein the memory stores a computer program, and is characterized in that the processor executes the computer program to implement a step of a distribution network line peak load prediction method based on a combined model.

[0043] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a distribution network line peak load prediction method based on a combined model are implemented.

[0044] Beneficial effects of the present invention: The distribution network line peak load prediction method based on the combined model provided by the present invention improves the quality of model input and prediction accuracy by extracting high-correlation input features by adopting the correlation analysis method, and improves the accuracy of the overall load rate prediction. By effectively extracting and analyzing the load rate peak, the accuracy of the peak load rate prediction is improved. By real-time monitoring and early warning of heavy overload conditions of the distribution network lines, the safety and reliability of the power grid operation are improved. The present invention achieves better results in terms of accuracy, safety and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0046] Figure 1 An overall flow chart of a distribution network line peak load prediction method based on a combined model provided for the first embodiment of the present invention.

[0047] Figure 2 A distribution network line load rate box plot of a distribution network line peak load prediction method based on a combined model provided in a second embodiment of the present invention.

[0048] Figure 3 A distribution network line load rate prediction effect diagram of a distribution network line peak load prediction method based on a combined model provided in the second embodiment of the present invention.

[0049] Figure 4 A distribution network line load rate prediction error diagram of a distribution network line peak load prediction method based on a combined model provided in a second embodiment of the present invention.

[0050] Figure 5 An overall module diagram of a distribution network line peak load prediction system based on a combined model provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0052] Example 1

[0053] Reference Figure 1 , as an embodiment of the present invention, provides a distribution network line peak load prediction method based on a combined model, comprising:

[0054] S1: Collect distribution network line data, extract highly relevant input features, and use support vector machine to make an overall prediction of the load rate.

[0055] Furthermore, extracting high-relevance input features includes feature correlation analysis.

[0056] It should be noted that the characteristic sample values ​​of the distribution network line input are collected, and based on the chaos theory correlation analysis method, the correlation coefficient between the features is calculated to evaluate the similarity in a specific time series, capture the nonlinear relationship between the features, and provide analysis results, which are expressed as:

[0057]

[0058]

[0059] Among them, ρ ij is the basic correlation coefficient between the i-th input feature and the j-th input feature, x ik is the kth sample value of the ith input feature, y jk is the kth sample value of the jth input feature, is the mean of the i-th input feature, is the mean of the jth input feature, N is the number of samples, ||x ik -y jk || is the Euclidean distance between feature vectors, σ is a parameter used to control the influence of distance, and the highly correlated features are adaptively weighted to generate comprehensive features. The weights are dynamically adjusted according to the relevance of the features so that the predicted contribution of each feature matches the actual relevance, which is expressed as:

[0060]

[0061]

[0062] Among them, Feature w is the weighted comprehensive feature, w i is the weight of the i-th feature, Feature i is the i-th input feature, and m is the number of features.

[0063] Furthermore, overall prediction of the load factor by support vector machine includes support vector machine algorithm.

[0064] It should be noted that by mapping the input data to a high-dimensional space through a kernel function, linearly inseparable data becomes linearly separable in the high-dimensional space, which can be expressed as:

[0065]

[0066]

[0067] Among them, f(x) is the prediction function, α i is the Lagrange multiplier in the support vector machine, x i is the i-th support vector, x is the input feature vector, b is the bias term in the support vector machine, K(xi ,x) is the input x and support vector x i The kernel function value between the two is used to introduce an error correction mechanism based on the support vector machine. The prediction results are dynamically adjusted through the Kalman filter. The prediction results are continuously updated and corrected according to the real-time observation data to minimize the prediction error, which is expressed as:

[0068]

[0069] in, is the corrected predicted value at time t+1, f(x t ) is the support vector machine prediction value at time t, y t is the observed value at time t, K t is the Kalman gain at the tth moment.

[0070] S2: The peak load rate and the time when the peak load rate occurs are combined into a data set, and independent predictions are made through the combined prediction model, and the prediction results of the support vector machine are corrected.

[0071] Furthermore, forming a data set of the peak load rate and the time when the peak load rate occurs includes generating a box plot data set.

[0072] It should be noted that the interquartile range of the data set is calculated to measure the degree of variation of the data, outliers in the data set are identified, outliers above the upper quartile are extracted, peak data sets are constructed, multi-scale analysis is performed on the peak data, peak features at different time scales are captured through wavelet transform, and changes and emergencies in the signal are identified, which can be expressed as:

[0073]

[0074] Outliers = {x is |x is <Q1-λ×IQR W x i >Q3+λ×IQR W}

[0075] Among them, IQR W is the interquartile range calculated by wavelet transform, Q3 s and Q1 s are the third quartile at the sth scale and the first quartile at the sth scale, S is the number of scales of the wavelet transform, Outliers is the set of outliers, x is is the i-th data point, Q3 and Q1 are the third quartile and the first quartile respectively.

[0076] Furthermore, independent predictions are made through combined prediction models including combined predictions based on attention mechanism, recurrent neural network and improved recurrent unit neural network.

[0077] It should be noted that the weight of each feature is calculated through the attention weight matrix, and the weight is applied to the feature vector to generate a weighted feature vector, which is expressed as:

[0078] Att=softmax(W a ·h t )

[0079]

[0080] Among them, Att is the attention weight matrix, W a is the weight matrix of the attention mechanism, h t is the hidden state vector at time t, is the weighted hidden state vector. The recurrent neural network captures the temporal dependency in the time series data. The improved recurrent unit neural network introduces a gating mechanism and combines it with an adaptive filter based on the Monte Carlo method to dynamically correct the hidden state. The distribution of the hidden state is estimated through multiple sampling and weight adjustment, which is expressed as:

[0081]

[0082] z t =σ(W z x t +U z h t-1 +b z )

[0083]

[0084] Among them, z t is the update gate, x t is the input vector at time t, U z is the weight matrix of the update gate, W z and W h is the weight matrix of the input vector, b z is the bias term of the update gate, is the candidate hidden state, Y h is the weight matrix of the reset gate, b h is the bias term of the candidate hidden state, r t is the reset gate, P is the number of particles, is the candidate hidden state of the p-th particle, is the weight of the pth particle.

[0085] S3: Determine the warning level according to the load rate of the distribution network line and send the distribution network line heavy overload warning information.

[0086] Furthermore, determining the warning level according to the load rate of the distribution network line includes determining the load condition of the distribution network line.

[0087] It should be noted that the calculation of the dynamic load rate reflects the load condition of the distribution network line within a preset time period. The dynamic load rate is the ratio of peak power to rated power, reflecting the load condition of the line. After obtaining the dynamic load rate, the load condition is stratified based on the preset judgment criteria, and the load rate is divided into normal, heavy load and overload. When the load rate is between 0 and 0.8, it is normal; when the load rate is between 0.8 and 1.0, it is heavy load; when the load rate exceeds 1.0, it is overload, expressed as:

[0088]

[0089]

[0090] Among them, L r (t) is the load rate at the tth moment, P p (t) is the peak power at the tth moment, P r (t) is the rated power at the tth moment, and WL(t) is the warning level at the tth moment.

[0091] Furthermore, sending the distribution network line heavy overload warning information includes sending the warning information according to different warning levels.

[0092] It should be noted that based on the load rate discrimination result, alarm information is generated. Different sending strategies are adopted according to different warning levels. When the warning level is overload, a high priority sending strategy is adopted. When the warning level is heavy load, a normal priority sending strategy is adopted. When the load rate is within the preset normal range, no action is taken to avoid interference, which is expressed as:

[0093] Alert(t)=GA(WL(t))

[0094]

[0095] Among them, Alert(t) is the warning information at the tth moment, GA(x) is the function of generating warning information, SA(t) is the warning information sending action at the tth moment, Send(x) is the function of sending warning information, H_P is the high priority sending strategy, N_P is the normal priority sending strategy, and N_A means taking no action.

[0096] Example 2

[0097] Reference Figure 2-Figure 4, which is an embodiment of the present invention, provides a distribution network line peak load prediction method based on a combined model. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0098] from Figure 2 It can be seen that the load rate fluctuates greatly over time. By extracting the outliers in the box plot to form a data set, a total of 144 points were extracted.

[0099] from Figure 3-Figure 4 It can be seen that the corrected traditional prediction model constructed by the present invention has the strongest ability to approximate the actual value and the best prediction effect.

[0100] As shown in Table 1, the prediction accuracy of the traditional support vector machine model is 90.79%, while the prediction accuracy of the corrected traditional combined prediction model constructed by the present invention is as high as 96.98%, and the accuracy is improved by 6.19%.

[0101] As shown in Tables 2 and 3, the distribution network lines in the test data set had a total of 69 heavy overloads, including 68 heavy loads and 1 overload. The traditional support vector machine model only predicted heavy loads 59 times, with a correct warning rate of 85.50%. The corrected traditional combined prediction model constructed by the present invention predicted heavy loads 65 times and overloads 1 time, with a correct warning rate of up to 96.65%, which is 11.15% higher than the traditional model.

[0102] Table 1 Prediction performance indicators of each model

[0103] Model Type Root mean square error Average Percent Error Traditional forecasting models 0.09 9.21％ Correcting traditional forecasting models 0.03 3.02％

[0104] Table 2 Distribution transformer overload judgment standard

[0105] Line load rate range Load level Line Status [0,0.8] 0 normal (0.8,1] 1 Overload (1,+∞] 2 Overload

[0106] Table 3. Results of heavy overload warning for each model

[0107] Model Type Actual overload Predictive reload Actual overload Predicted overload Traditional Model 68 times 59 times 1 time 0 times Correction model 68 times 65 times 1 time 1 time

[0108] Example 3

[0109] Reference Figure 5 , which is an embodiment of the present invention, provides a distribution network line peak load prediction system based on a combined model, including: an extraction prediction module, a combined prediction module, and a discrimination and warning module.

[0110] The extraction prediction module is used to collect distribution network line data, extract highly correlated input features, and make an overall prediction of the load rate through a support vector machine; the combined prediction module is used to combine the peak load rate and the time when the peak load rate occurs into a data set, make independent predictions through a combined prediction model, and correct the prediction results of the support vector machine; the discrimination and warning module is used to discriminate the warning level according to the load rate of the distribution network line, and send heavy overload warning information of the distribution network line.

[0111] If the function 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. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment 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, etc. Various media that can store program codes.

[0112] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0113] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0114] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution network line peak load prediction method based on a combined model, characterized in that: include: Collect distribution network line data, extract highly relevant input features, and use support vector machines to make an overall prediction of the load rate; The peak load rate and the time when the peak load rate occurs are combined into a data set, and independent predictions are made through the combined prediction model, and the prediction results of the support vector machine are corrected; The warning level is determined according to the load rate of the distribution network line, and the distribution network line overload warning information is sent.

2. The distribution network line peak load prediction method based on the combined model according to claim 1, characterized in that: The extraction of high-correlation input features includes feature correlation analysis, collecting distribution network line input feature sample values, and evaluating the similarity on a specific time series by calculating the correlation coefficient between features based on the chaos theory correlation analysis method, capturing the nonlinear relationship between features, and providing analysis results, which are expressed as: Among them, ρ ij is the basic correlation coefficient between the i-th input feature and the j-th input feature, x ik is the kth sample value of the ith input feature, y jk is the kth sample value of the jth input feature, is the mean of the i-th input feature, is the mean of the jth input feature, N is the number of samples, ||x ik -y jk || is the Euclidean distance between feature vectors, σ is a parameter used to control the influence of distance, and the highly correlated features are adaptively weighted to generate comprehensive features. The weights are dynamically adjusted according to the relevance of the features so that the predicted contribution of each feature matches the actual relevance, which is expressed as: Among them, Feature w is the weighted comprehensive feature, w i is the weight of the i-th feature, Feature i is the i-th input feature, and m is the number of features.

3. The distribution network line peak load prediction method based on the combined model as claimed in claim 2, characterized in that: The overall prediction of the load rate by the support vector machine includes a support vector machine algorithm, which maps the input data to a high-dimensional space through a kernel function, so that the linearly inseparable data becomes linearly separable in the high-dimensional space, which is expressed as: Among them, f(x) is the prediction function, α i is the Lagrange multiplier in the support vector machine, x i is the i-th support vector, x is the input feature vector, b is the bias term in the support vector machine, K(x i ,x) is the input x and support vector x i The kernel function value between the two is used to introduce an error correction mechanism based on the support vector machine. The prediction results are dynamically adjusted through the Kalman filter. The prediction results are continuously updated and corrected according to the real-time observation data to minimize the prediction error, which is expressed as: in, is the corrected predicted value at time t+1, f(x t ) is the support vector machine prediction value at time t, y t is the observed value at time t, K t is the Kalman gain at the tth moment.

4. The distribution network line peak load prediction method based on the combined model as claimed in claim 3, characterized in that: The method of forming a data set of peak load rate and peak load rate occurrence time includes generating a box plot data set, calculating the interquartile range of the data set, measuring the degree of data variation, identifying outliers in the data set, extracting outliers above the upper quartile, constructing a peak data set, performing multi-scale analysis on the peak data, capturing peak features on different time scales through wavelet transform, and identifying changes and emergencies in the signal, which is expressed as: Outliers={x is |x is <Q1-λ×IQR W or x i >Q3+λ×IQR W } Among them, IQR W is the interquartile range calculated by wavelet transform, Q3 s and Q1 s are the third quartile at the sth scale and the first quartile at the sth scale, S is the number of scales of the wavelet transform, Outliers is the set of outliers, and x is is the i-th data point, Q3 and Q1 are the third quartile and the first quartile respectively.

5. The distribution network line peak load prediction method based on the combined model as claimed in claim 4, characterized in that: The independent prediction by the combined prediction model includes combined prediction based on the attention mechanism, the recurrent neural network and the improved recurrent unit neural network, the weight of each feature is calculated by the attention weight matrix, and the weight is applied to the feature vector to generate a weighted feature vector, which is expressed as: That=softmax(W a ·h t ) Among them, Att is the attention weight matrix, W a is the weight matrix of the attention mechanism, h t is the hidden state vector at time t, is the weighted hidden state vector. The recurrent neural network captures the temporal dependency in the time series data. The improved recurrent unit neural network introduces a gating mechanism and combines it with an adaptive filter based on the Monte Carlo method to dynamically correct the hidden state. The distribution of the hidden state is estimated through multiple sampling and weight adjustment, which is expressed as: z t =σ(W z x t +U z h t-1 +b z ) Among them, z t is the update gate, x t is the input vector at time t, U z is the weight matrix of the update gate, W z and W h is the weight matrix of the input vector, b z is the bias term of the update gate, is the candidate hidden state, U h is the weight matrix of the reset gate, b h is the bias term of the candidate hidden state, r t is the reset gate, P is the number of particles, is the candidate hidden state of the p-th particle, is the weight of the pth particle.

6. The distribution network line peak load prediction method based on the combined model as claimed in claim 5, characterized in that: The method of judging the warning level according to the load rate of the distribution network line includes judging the load condition of the distribution network line, calculating the dynamic load rate, reflecting the load condition of the distribution network line within a preset time period, and the dynamic load rate is the ratio of the peak power to the rated power, reflecting the load condition of the line. After obtaining the dynamic load rate, the load condition is layered based on the preset judgment standard, and the load rate is divided into normal, heavy load and overload. When the load rate is between 0 and 0.8, it is normal; when the load rate is between 0.8 and 1.0, it is heavy load; when the load rate exceeds 1.0, it is overload, which is expressed as: Among them, L r (t) is the load rate at the tth moment, P p (t) is the peak power at the tth moment, P r (t) is the rated power at the tth moment, and WL(t) is the warning level at the tth moment.

7. The distribution network line peak load prediction method based on the combined model according to claim 6, characterized in that: The sending of distribution network line heavy overload warning information includes sending warning information according to different warning levels, generating alarm information based on the load rate discrimination result, and adopting different sending strategies according to different warning levels. When the warning level is overload, a high priority sending strategy is adopted. When the warning level is heavy load, a normal priority sending strategy is adopted. When the load rate is within a preset normal range, no action is taken to avoid interference, which is expressed as: Alert(t)=GA(WL(t)) Among them, Alert(t) is the warning information at the tth moment, GA(x) is the function of generating warning information, SA(t) is the warning information sending action at the tth moment, Send(x) is the function of sending warning information, H_P is the high priority sending strategy, N_P is the normal priority sending strategy, and N_A means taking no action.

8. A system using the distribution network line peak load prediction method based on the combined model as claimed in any one of claims 1 to 7, characterized in that: It includes extraction prediction module, combination prediction module and discrimination warning module; The extraction prediction module is used to collect distribution network line data, extract highly relevant input features, and make an overall prediction of the load rate through a support vector machine; The combined prediction module is used to combine the peak load rate and the peak load rate occurrence time into a data set, perform independent prediction through the combined prediction model, and correct the prediction result of the support vector machine; The identification and warning module is used to identify the warning level according to the load rate of the distribution network line and send the distribution network line heavy overload warning information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution network line peak load prediction method based on the combined model described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distribution network line peak load prediction method based on a combined model as described in any one of claims 1 to 7 are implemented.

Images