Machine learning-based multi-dimensional auxiliary decision analysis method and system for distribution network power supply

Abstract

The application provides a power supply multi-dimensional auxiliary decision analysis method and system based on machine learning, which comprises the following steps: S1, acquiring power distribution network historical data and historical power load data; S2, obtaining a power supply demand prediction model through a long short-term memory network and the power distribution network historical data; S3, obtaining a load change model through a time series prediction algorithm and the historical power load data; S4, identifying real-time operation data of the power distribution network based on the outputs of steps S2 and S3 to obtain a power supply capacity change trend and a load margin; S5, establishing a comprehensive objective function based on the output of step S4, optimizing and solving the objective function through a genetic algorithm, and outputting an optimal power supply decision scheme; and S6, executing the output of step S5, real-time monitoring of power distribution network power supply operation parameters, and dynamically adjusting the optimal power supply decision scheme based on the power distribution network power supply operation parameters. The application improves the reliability of power distribution network power supply and realizes dynamic adjustment and optimization of the power distribution network power supply decision scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of distribution network power supply decision analysis, and particularly relates to a machine learning-based multi-dimensional auxiliary decision analysis method and system for distribution network power supply. BACKGROUND

[0002] With the rapid development of power systems and the continuous growth of electricity demand, the distribution network power supply guarantee is facing increasingly severe challenges. The distribution network power supply guarantee mainly relies on conventional measurement equipment, including collecting basic operating parameters such as voltage and current, and analyzing power supply demand based on historical statistical data and empirical formulas. Load management adopts a fixed scheme to balance distribution by adjusting transformers and switches, while power supply decisions are often based on a single indicator, lacking comprehensive optimization capabilities. Operation monitoring relies on regular inspections and simple devices, and abnormal processing relies on manual intervention, lacking flexibility and dynamics. In response to complex and variable power supply demands, it is difficult to achieve accurate prediction and rapid response.

[0003] The existing distribution network power supply decision method mainly relies on simple statistical models, which have obvious shortcomings in power supply demand prediction accuracy, load change analysis, and decision scheme flexibility. The prediction model used cannot effectively capture the complex characteristics of power supply demand, load analysis lacks multi-dimensional consideration, decision schemes are difficult to dynamically adjust, and optimization objectives are too single, which seriously restricts the reliability of distribution network power supply. SUMMARY

[0004] In view of this, the present application aims to provide a machine learning-based multi-dimensional auxiliary decision analysis method and system for distribution network power supply, which solves the problem of lack of multi-dimensional consideration in load analysis, difficulty in dynamically adjusting decision schemes, and too single optimization objectives in the prior art.

[0005] To achieve the above-mentioned application purposes, the present application provides a machine learning-based multi-dimensional auxiliary decision analysis method for distribution network power supply in a first aspect, which comprises:

[0006] S1, obtaining distribution network historical data and historical electricity load data;

[0007] S2, constructing an initial power supply demand prediction model through a long short-term memory network, training and verifying the initial power supply demand prediction model based on the distribution network historical data, and obtaining a power supply demand prediction model;

[0008] S3, constructing an initial load change model through a time series prediction algorithm, training and verifying the initial load change model based on the historical electricity load data, and obtaining a load change model;

[0009] S4, identifying real-time operation data of the distribution network based on the power supply demand prediction model and the load change model, and obtaining power supply capacity change trend and load margin.

[0010] S5, constructing a comprehensive objective function through the power supply capacity change trend and the load margin, and optimizing and solving the objective function through a genetic algorithm to output an optimal power supply decision scheme;

[0011] S6, executing the optimal power supply decision scheme, monitoring power supply operation parameters of the distribution network in real time, and dynamically adjusting the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

[0012] Further, in step S1, the distribution network historical data and the historical power consumption load data are acquired, specifically including:

[0013] S11, collecting initial distribution network historical data and initial historical power consumption load data respectively;

[0014] S12, performing time series processing and feature extraction on the initial distribution network historical data to obtain the distribution network historical data;

[0015] S13, performing periodic analysis and outlier processing on the initial historical power consumption load data to obtain the historical power consumption load data.

[0016] Further, in step S2, a power supply demand prediction model is constructed, specifically including:

[0017] An initial power supply demand prediction model is constructed through a long short-term memory network, the distribution network historical data is input into the initial power supply demand prediction model, and a power supply demand prediction model is obtained, and a calculation formula of the power supply demand prediction model is:

[0018]

[0019] wherein, is the predicted power supply demand at time t, is a model center weight, is a hidden state at time t, is an adjustment coefficient, is the data processed by a data source i at time t, and n is the number of data sources, is a weight coefficient of the data source i, and are respectively a first error correction coefficient and a second error correction coefficient, is a trend adjustment coefficient, is a trend function;

[0020] Further, step S3 includes:

[0021] S31, using a long short-term memory network to establish a deep learning model and extracting long-term change features of the historical power consumption load data;

[0022] S32, analyze the periodic fluctuation characteristics and short-term fluctuation characteristics of the historical electricity load data by using an autoregressive integrated moving average model;

[0023] S33, combine the prediction results of the deep learning model and the autoregressive integrated moving average model by using an adaptive weight fusion algorithm to obtain a combined prediction result;

[0024] S34, input the combined prediction result into an initial load change model for training and verification to obtain a load change model, and a calculation formula of the load change model is:

[0025]

[0026]

[0027] wherein, is the predicted electricity load at time t, is the actual electricity load at time t, is the actual electricity load at time t, is the actual electricity load at time t, is the jth feature affecting the electricity load at time t, and m is the number of features affecting the electricity load, is the feature weight of is the external interference factor at time t, is the adjustable parameter of the actual electricity load, is the adjustable parameter of the electricity load, is the adjustable parameter of the external interference factor, is the loss function value of the initial load change model, is the mean square error, is the mean absolute error, is the root mean square percentage error, is the mean square error, is the mean absolute error, is the weight coefficient of the root mean square percentage error, is the accuracy, is the mean of the predicted electricity load, is the mean of the actual electricity load. Further, step S4 comprises: S41, real-time acquisition of power distribution network real-time operation data in power distribution network operation by using sensors and monitoring devices;

[0028]

[0029] S41, real-time acquisition of power distribution network real-time operation data in power distribution network operation by using sensors and monitoring devices;

[0030] ​​​​S42, identifying the real-time operation data of the distribution network based on the power supply demand prediction model and the load change model to obtain real-time predicted power supply demand and real-time predicted power consumption load;

[0031] S43, classifying and regression analyzing the real-time predicted power supply demand and the real-time predicted power consumption load by a support vector machine algorithm to obtain a power supply capacity change trend in a future period, and a calculation formula of the power supply capacity change trend is:

[0032]

[0033] is the power supply capacity change trend at time , is the voltage at time , is the current at time , is the frequency at time , is the weight coefficient of the voltage, is the weight coefficient of the current, is the weight coefficient of the frequency;

[0034] S44, calculating the real-time predicted power consumption load based on a fuzzy logic algorithm to obtain a load margin, and a calculation formula of the load margin is:

[0035]

[0036] is the load margin at time , is the real-time predicted power consumption load at time , is the real-time actual power consumption load at time

[0037] Further, step S5 includes:

[0038] S51, determining power supply capacity indicators and load margin indicators based on the power supply capacity change trend and the load margin;

[0039] S52, assigning different importance weights to the power supply capacity indicators and the load margin indicators by a weight distribution method to construct a comprehensive objective function, and a calculation formula of the comprehensive objective function is:

[0040]

[0041] wherein F is a value of the comprehensive objective function, is an actual value of the kth power supply capacity indicator, ​​​​​​​​is a real value of the first load margin index, is a maximum value normalized by the power supply capacity index, is a maximum value normalized by the load margin index, is a weight coefficient of the power supply capacity index, is a weight coefficient of the load margin index, N is the number of the power supply capacity indexes, and M is the number of the load margin indexes;

[0042] S53, the comprehensive objective function is optimized and solved by a genetic algorithm, an initial population is initialized, and an initial population is generated;

[0043] S54, the fitness of each individual in the initial population is evaluated based on the comprehensive objective function, and the individual with the highest fitness is selected based on an elite reservation strategy;

[0044] S55, new individuals of a new generation are generated through partial random crossover and mutation, new individuals are repeatedly iterated, when the iteration reaches a preset iteration number or the fitness change meets the requirements, the iteration is stopped, and an optimal power supply decision scheme is output.

[0045] Further, step S6 comprises:

[0046] S61, according to the optimal power supply decision scheme, the operation parameters of the equipment in the distribution network are adjusted, and the load demand of the equipment is redistributed;

[0047] S62, the adjusted operation parameters and the redistributed load demand are issued to the corresponding equipment in the form of control commands through the distribution network control center;

[0048] S63, an Internet of Things sensor is deployed at a key equipment of the distribution network, and power supply operation parameters are collected in real time based on the Internet of Things sensor;

[0049] S64, the power supply operation parameters are cleaned, filtered and normalized to obtain distribution network power supply operation data, and the distribution network power supply operation data is transmitted to a centralized monitoring platform for real-time storage and backup;

[0050] S65, based on the distribution network power supply operation data, the current power supply state is analyzed, potential abnormalities and operation bottlenecks are identified, and a state evaluation result is obtained;

[0051] S66, according to the state evaluation result, combined with historical data, a corresponding adjustment strategy is formulated and executed, and the optimal power supply decision scheme is dynamically adjusted.

[0052] The second aspect of the present application provides a distribution network power supply multi-dimensional auxiliary decision analysis system based on machine learning, which comprises:

[0053] A data acquisition module is configured to acquire historical data of the distribution network and historical power consumption load data.

[0054] a power supply demand module, configured to construct an initial power supply demand prediction model through a long short-term memory network, train and verify the initial power supply demand prediction model based on the historical data of the distribution network, and obtain a power supply demand prediction model;

[0055] a load change module, configured to construct an initial load change model through a time series prediction algorithm, train and verify the initial load change model based on the historical power consumption load data, and obtain a load change model;

[0056] a trend margin module, configured to identify real-time operation data of the distribution network based on the power supply demand prediction model and the load change model, and obtain a power supply capacity change trend and a load margin;

[0057] a target optimization module, configured to construct a comprehensive target function through the power supply capacity change trend and the load margin, optimize and solve the comprehensive target function through a genetic algorithm, and output an optimal power supply decision scheme;

[0058] a scheme adjustment module, configured to execute the optimal power supply decision scheme, monitor real-time power supply operation parameters of the distribution network, and dynamically adjust the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

[0059] The third aspect of the present application provides an electronic device, comprising: at least one processor, at least one memory, a communication interface and a bus;

[0060] The processor, the memory and the communication interface can communicate with each other through the bus, the memory stores program instructions executable by the processor, and the processor invokes the program instructions to implement the method of any one of claims 1-7.

[0061] The fourth aspect of the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions enable a computer to implement the method of any one of claims 1-7.

[0062] Compared with the prior art, the present application has the following advantages:

[0063] 1. By acquiring power distribution history data and power load data, and using long short-term memory network to build power demand prediction model, using time series prediction algorithm to establish load change model, combining multi-dimensional data to analyze load change characteristics, based on power demand and load change model, identifying real-time operation data of distribution network, obtaining power supply capacity change trend and load margin, according to genetic algorithm, constructing and solving comprehensive objective function, outputting optimal power supply decision scheme, realizing dynamic adjustment and optimization of decision scheme, through real-time monitoring and feedback mechanism, continuously optimizing power supply decision, improving the reliability and economy of distribution network power supply, overcoming the limitations of traditional methods, realizing the efficiency of power supply guarantee.

[0064] (2) By long short-term memory network to build power demand prediction model, combining its deep learning ability to effectively capture the long-term change characteristics of power demand, through the recursive neural network structure to model the time series, and using adaptive learning rate optimization algorithm to dynamically adjust the network parameters, improving the accuracy of power demand prediction.

[0065] (3) By combining long short-term memory network and autoregressive integrated moving average model, the long-term change characteristics and periodic fluctuation characteristics of historical power load data are comprehensively analyzed, and through the adaptive weight fusion algorithm, the prediction results of deep learning model and traditional statistical model are combined to build an accurate load change model, which improves the accuracy of load change prediction and enhances the adaptability to complex load change. BRIEF DESCRIPTION OF DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only embodiments of the present application, and those skilled in the art can obtain other drawings according to the provided drawings without creating any creative labor.

[0067] Figure 1 The flow chart of the multi-dimensional auxiliary decision analysis method for distribution network power supply based on machine learning provided by the embodiment of the present application;

[0068] Figure 2 The structural diagram of the multi-dimensional auxiliary decision analysis system for distribution network power supply based on machine learning provided by the embodiment of the present application. DETAILED DESCRIPTION

[0069] The present application will be further described in detail below in combination with the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application. In addition, it should be noted that, in order to facilitate the description, only the parts related to the present application are shown in the drawings, not all the structures.

[0070] Referring Figure 1 The first aspect of the embodiment provides a multi-dimensional auxiliary decision analysis method for power supply of distribution network based on machine learning. Specifically, the embodiment obtains distribution network historical data and power load data, constructs a power demand prediction model using a long short-term memory network, establishes a load change model using a time series prediction algorithm, analyzes load change characteristics based on multi-dimensional data, identifies real-time operation data of the distribution network based on the power demand and the load change model, obtains power supply capacity change trend and load margin, constructs a comprehensive objective function, solves the comprehensive objective function through a genetic algorithm, outputs an optimal power supply decision scheme, realizes dynamic adjustment and optimization of the decision scheme, continuously optimizes power supply decisions through real-time monitoring and feedback mechanisms, improves the reliability and economy of power supply of the distribution network, overcomes the limitations of traditional methods, and realizes efficient power supply guarantee.

[0071] The method comprises:

[0072] S1, obtaining distribution network historical data and historical power load data.

[0073] In step S1, the distribution network historical data and the historical power load data are obtained, specifically including:

[0074] S11, collecting initial distribution network historical data and initial historical power load data respectively. The original distribution network historical data includes distribution network equipment operation state data, distribution network fault record data and distribution network maintenance record data.

[0075] S12, performing time series processing and feature extraction on the initial distribution network historical data to obtain the distribution network historical data.

[0076] S13, performing periodic analysis and outlier processing on the initial historical power load data to obtain the historical power load data. The original historical power load data includes user power consumption data, load power data and load change rate data.

[0077] Specifically, the embodiment collects multi-dimensional data such as distribution network equipment operation state, fault record and maintenance record, as well as load data such as user power consumption, load power and change rate, to ensure the completeness of the basic data. The distribution network historical data is processed by time series and feature extraction, so that the data has time sequence. The historical power load data is periodically analyzed and outlier processed to effectively remove data noise and improve data reliability.

[0078] The distribution network historical data includes distribution network equipment operation state data, distribution network fault record data, and distribution network maintenance record data. The distribution network equipment operation state data is used to reflect the working state of the equipment; the distribution network fault record data is used to record the fault conditions in the system operation; and the distribution network maintenance record data is used to record the maintenance information of the equipment. After the above raw data is subjected to time sequence processing and feature extraction, the distribution network historical data that can be used for model training is formed.

[0079] The historical power consumption load data includes user power consumption data, load power data, and load change rate data; the user power consumption data is used to record the actual power consumption of the user; the load power data is used to reflect the load level of the system; and the load change rate data is used to represent the dynamic characteristics of the load change. After the above raw data is subjected to periodic analysis and outlier processing, the historical power consumption load data that can be used for model training is formed.

[0080] S2, an initial power supply demand prediction model is constructed through a long short-term memory network, the initial power supply demand prediction model is trained and verified based on the distribution network historical data, and a power supply demand prediction model is obtained.

[0081] In step S2, the power supply demand prediction model is constructed, specifically including:

[0082] The initial power supply demand prediction model is constructed through the long short-term memory network, the distribution network historical data is input into the initial power supply demand prediction model, and the power supply demand prediction model is obtained. The calculation formula of the power supply demand prediction model is:

[0083]

[0084] wherein, is the predicted power supply demand at time t, is the model center weight, is the hidden state at time t, is the adjustment coefficient, is the data processed by the data source i at time t, and n is the number of data sources, is the weight coefficient of the data source i, and are respectively the first error correction coefficient and the second error correction coefficient, is the trend adjustment coefficient, is the trend function;

[0085] Specifically, the long short-term memory network is used to train and verify the power supply demand prediction model, the recurrent neural network structure is used for time series modeling, and the adaptive learning rate optimization algorithm is used to dynamically adjust the network parameters. The specific steps of constructing the power supply demand prediction model through the long short-term memory network are as follows:

[0086] 1) Data preprocessing:

[0087] Collect and clean the distribution network historical data, including:

[0088] Device operating status data (voltage, current, frequency, etc.), fault record data (fault type, occurrence time, duration), maintenance record data (maintenance type, maintenance period, maintenance effect), ensure the integrity and consistency of the data, and standardize the data, such as: time series processing of data, using 15 minutes as the basic sampling interval, through standardized processing to map the data to the [-1,1] interval, to adapt to the input requirements of the long short-term memory network model.

[0089] 2) Model construction:

[0090] Select the long short-term memory network structure, determine the number of layers and the number of neurons in each layer, define the structure of the input layer, hidden layer and output layer, and ensure that the model can process time series data.

[0091] Build a 3-layer LSTM network structure: the input layer includes 128 neurons; the hidden layer includes 256 neurons; the output layer includes 1 neuron; the time step is set to 96 (corresponding to 24 hours); the batch size is set to 64.

[0092] 3) Model initialization:

[0093] Initialize the various parameters of the model, including center weight, data source weight coefficient, adjustment coefficient, error correction coefficient and trend adjustment coefficient, determine the form of the trend function, so as to capture the trend of power supply demand in the model.

[0094] The model center weight is randomly initialized in the [-0.1, 0.1] interval; the initial value of the data source weight coefficient is set to 1 / n; the initial value of the adjustment coefficient is set to 0.8; the error correction coefficients and are initialized to 0.3 and 0.7, respectively; the initial value of the trend adjustment coefficient is set to 0.5.

[0095] 4) Model training:

[0096] Input the standardized distribution network historical data into the long short-term memory network model, use the recurrent neural network structure for time series modeling, use the adaptive learning rate optimization algorithm (such as Adam or RMSprop) to dynamically adjust the network parameters, improve the training efficiency and prediction accuracy of the model. Specifically: use Adam optimizer, set the initial learning rate to 0.001, use the recurrent neural network structure, the time window size is 24, the training round (epoch) is set to 100, and the learning rate is updated every 50 batches.

[0097] 5) Model validation:

[0098] The model is evaluated using the validation dataset to ensure its generalization ability, and the model parameters are adjusted according to the validation results to optimize the model performance.

[0099] 20% of the data is used as the validation set, and the evaluation indicators include: mean square error (MSE) target value <0.05, mean absolute error (MAE) target value <0.08, and accuracy target value >90%.

[0100] 6) Model output:

[0101] The trained long short-term memory network model outputs the predicted power supply demand at time t, and the final prediction result is calculated using model center weight, hidden state, adjustment coefficient, etc. The predicted power supply demand at time t includes: voltage prediction value (accurate to 0.1V), current prediction value (accurate to 0.01A), and frequency prediction value (accurate to 0.01Hz).

[0102] This embodiment introduces multiple adjustment coefficients and error correction coefficients in the calculation formula to realize fine adjustment of the prediction result, considering the weight influence of multiple data sources , improving the comprehensiveness of the prediction, using a recurrent neural network structure for time series modeling, which can effectively capture the time sequence characteristics of power supply demand, dynamically adjusting network parameters through an adaptive learning rate optimization algorithm, so that the model can adaptively adapt to data changes, and by introducing a trend function modeling the trend of power supply demand changes, preserving and transmitting historical information through hidden state improving the long-term memory ability of the model and improving the accuracy of power supply demand prediction.

[0103] S3, an initial load change model is constructed by a time series prediction algorithm, and the initial load change model is trained and verified based on the historical electricity load data to obtain a load change model.

[0104] Step S3 includes:

[0105] S31, a long short-term memory network is used to establish a deep learning model to extract long-term change characteristics of historical electricity load data.

[0106] S32, an autoregressive integrated moving average model is used to analyze the periodic fluctuation characteristics and short-term fluctuation characteristics of historical electricity load data.

[0107] S33, the prediction results of the deep learning model and the autoregressive integrated moving average model are combined by an adaptive weight fusion algorithm to obtain a combined prediction result.

[0108] S34. Input the combined prediction results into the initial load change model for training and validation to obtain the load change model. The calculation formula of the load change model is:

[0109]

[0110]

[0111] in, For time Predicted electricity load at that time For time The actual power load at that time For time The actual power load at that time For time The j-th characteristic affecting the electricity load at any given time, where m is the number of characteristics affecting the electricity load. for Feature weights, For time External interference factors at the time Adjustable parameters based on actual power load. To adjust parameters that affect electricity load, Adjustable parameters to account for external interference factors. The loss function value for the initial load change model. Mean square error, The mean absolute error, The root mean square percentage error. Mean square error, Mean absolute error The weighting coefficient for the root mean square percentage error. For accuracy, To predict the average electricity load, This represents the average of the actual electricity load.

[0112] Specifically, historical electricity load data is divided into training and testing sets. The initial load change model is trained and validated using these sets. Cross-validation is used to evaluate the model, and mean squared error (MSE) and mean absolute error (MAE) are used as performance evaluation metrics. Model parameters are adjusted based on the validation results. This embodiment extracts long-term variation characteristics of historical electricity load data using a long short-term memory (LSTM) network, analyzes periodic and short-term fluctuations using an autoregressive integral moving average (ARM) model, and optimizes the combination of prediction results from the two models using an adaptive weight fusion algorithm. The MSE is then introduced into the model. Mean absolute error and root mean square percentage error Three evaluation indexes, and weight coefficients 、 、 The accuracy index is used for the weighted combination of different evaluation indexes The comprehensive evaluation is carried out, the reliability of model verification is improved, the historical power load data is divided into training set and test set for model training and verification, the cross-validation method is used to evaluate the model performance, the model parameters are dynamically adjusted according to the verification result, and the precision of the model is improved.

[0113] S4, based on the power supply demand prediction model and the load change model, identifying the real-time operation data of the distribution network to obtain the power supply capacity change trend and the load margin.

[0114] Step S4 includes:

[0115] S41, real-time acquisition of real-time operation data of the distribution network in the operation of the distribution network through sensors and monitoring equipment.

[0116] S42, based on the power supply demand prediction model and the load change model, identifying the real-time operation data of the distribution network to obtain the real-time predicted power supply demand and the real-time predicted power load. The real-time predicted power supply demand includes voltage, current and frequency.

[0117] S43, classifying and regression analyzing the real-time predicted power supply demand and the real-time predicted power load by support vector machine algorithm to obtain the power supply capacity change trend of the future period, and the calculation formula of the power supply capacity change trend is:

[0118]

[0119] Among them, is the power supply capacity change trend at time t, is the voltage at time t, is the current at time t, is the frequency at time t, is the weight coefficient of voltage, is the weight coefficient of current, is the weight coefficient of frequency. S44, calculating the real-time predicted power load based on fuzzy logic algorithm to obtain the load margin, and the calculation formula of the load margin is:

[0120]

[0121]

[0122] is the time​​​​ the load margin at the time, the real-time predicted power load at the time the real-time predicted power load at the time the real-time predicted power load at the time the real-time actual power load at the time.

[0123] Specifically, the embodiment realizes real-time collection of distribution network operation data through sensors and monitoring devices, real-time acquisition of key power supply parameters such as voltage, current and frequency, ensures the timeliness and accuracy of data, identifies real-time data based on a power supply demand prediction model and a load change model, obtains real-time predicted power supply demand and real-time predicted power load, and realizes real-time evaluation of power supply status. By using a support vector machine algorithm for classification and regression analysis, the change trend of power supply capacity is accurately predicted, the load margin is calculated using a fuzzy logic algorithm, and the rationality of the calculation result is improved.

[0124] S5, constructing a comprehensive objective function through the power supply capacity change trend and the load margin, optimizing and solving the objective function through a genetic algorithm, and outputting an optimal power supply decision scheme.

[0125] Step S5 includes:

[0126] S51, determining power supply capacity indicators and load margin indicators based on the power supply capacity change trend and the load margin.

[0127] S52, assigning different importance weights to the power supply capacity indicators and the load margin indicators using a weight distribution method, constructing a comprehensive objective function, and the calculation formula of the comprehensive objective function is:

[0128]

[0129] wherein F is the value of the comprehensive objective function, is the actual value of the kth power supply capacity indicator, is the actual value of the 1st load margin indicator, is the maximum value standardization of the power supply capacity indicator, is the maximum value standardization of the load margin indicator, is the weight coefficient of the power supply capacity indicator, is the weight coefficient of the load margin indicator, N is the number of power supply capacity indicators, and M is the number of load margin indicators.

[0130] S53, optimizing and solving the comprehensive objective function through a genetic algorithm, initializing a population, and generating an initial population.

[0131] S54, evaluating the fitness of each individual in the initial population based on the comprehensive objective function, and selecting the individual with the highest fitness based on an elite retention strategy.

[0132] S55, generating new individuals by partial random crossover and mutation, repeatedly iterating to generate new individuals, stopping iteration when the iteration reaches a preset iteration number or the fitness change meets the requirement, and outputting the optimal power supply decision scheme.

[0133] Specifically, the embodiment simultaneously considers the power supply capacity index and the load margin index, assigns reasonable importance weights to different indexes through a weight allocation method, realizes unified optimization of multiple objectives through construction of a comprehensive objective function. The embodiment uses a genetic algorithm to optimize and solve the comprehensive objective function, retains high-quality individuals with high fitness through an elite retention strategy, and uses a partial random crossover and mutation mechanism to avoid local optimum. Based on iterative optimization to continuously improve the quality of the solution, a reasonable iteration termination condition (such as a preset iteration number or a fitness change requirement) is set, so as to output the optimal power supply decision scheme that has been fully optimized.

[0134] S6, executing the optimal power supply decision scheme, monitoring the power supply operation parameters of the distribution network in real time, and dynamically adjusting the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

[0135] Step S6 includes:

[0136] S61, adjusting the operation parameters of the equipment in the distribution network according to the optimal power supply decision scheme, and reallocating the load demand of the equipment. The operation parameters include switch state, transformer tap position, and distributed energy output power.

[0137] S62, through the distribution network control center, the adjusted operation parameters and the reallocated load demand are issued to the corresponding equipment in the form of control commands.

[0138] S63, deploying Internet of Things sensors on key equipment in the distribution network, and collecting power supply operation parameters in real time based on the Internet of Things sensors. The power supply operation parameters include voltage, current, frequency, and power factor.

[0139] S64, cleaning, filtering and normalizing the power supply operation parameters to obtain distribution network power supply operation data, and transmitting the distribution network power supply operation data to the centralized monitoring platform for real-time storage and backup.

[0140] S65, based on the distribution network power supply operation data, analyzing the current power supply state, identifying potential abnormalities and operation bottlenecks, and obtaining a state evaluation result.

[0141] S66, according to the state evaluation result, combining historical data, formulating and executing corresponding adjustment strategies, and dynamically adjusting the optimal power supply decision scheme. Continuously monitor the adjustment effect, and verify the performance improvement of the adjusted optimal power supply decision scheme through a feedback mechanism.

[0142] In a specific embodiment, according to the optimal power supply decision scheme, the operating parameters of each device in the distribution network are adjusted, including the switch state, the transformer tap position and the distributed energy output power. The load demand of each node is redistributed, and the adjusted operating parameters and load distribution scheme are issued to each related device in the form of control commands through the distribution network control center. Internet of Things sensors are deployed at key devices in the distribution network to collect real-time power supply operating parameters such as voltage, current, frequency and power factor. The collected power supply operating parameters are cleaned, filtered and normalized to obtain distribution network power supply operating data, which is transmitted to the centralized monitoring platform for real-time storage and backup. Based on the distribution network power supply operating data, the current power supply state is analyzed to identify potential abnormalities and operating bottlenecks, and the state evaluation result is obtained. According to the state evaluation result, combined with historical data, the corresponding adjustment strategy is formulated, the adjustment strategy is executed, and the optimal power supply decision scheme is adjusted. The adjustment effect is continuously monitored, and the performance improvement of the adjusted optimal power supply decision scheme is verified through the feedback mechanism.

[0143] Specifically, the embodiment ensures the stability of the power supply system by precisely adjusting the operating parameters of the distribution network devices and the load distribution, and ensures the timeliness of the power supply operating data through the deployment of Internet of Things sensors and real-time data collection. The embodiment also dynamically adjusts by analyzing the power supply state in real time, identifying abnormalities and bottlenecks, and combining historical data to ensure continuous optimization of the power supply decision scheme.

[0144] As another alternative embodiment, with reference to Figure 2 The application also provides a machine learning-based multi-dimensional auxiliary decision analysis system for power supply of a distribution network, which comprises:

[0145] A data acquisition module is configured to acquire historical data of the distribution network and historical electricity load data.

[0146] A power supply demand module is configured to construct an initial power supply demand prediction model through a long short-term memory network, train and verify the initial power supply demand prediction model based on the historical data of the distribution network, and obtain a power supply demand prediction model.

[0147] A load change module is configured to construct an initial load change model through a time series prediction algorithm, train and verify the initial load change model based on the historical electricity load data, and obtain a load change model.

[0148] A trend margin module is configured to identify real-time operating data of the distribution network based on the power supply demand prediction model and the load change model, and obtain a power supply capacity change trend and a load margin.

[0149] A target optimization module is configured to construct a comprehensive target function through the power supply capacity change trend and the load margin, optimize and solve the comprehensive target function through a genetic algorithm, and output an optimal power supply decision scheme.

[0150] A scheme adjustment module is configured to execute the optimal power supply decision scheme, monitor power supply operation parameters of the distribution network in real time, and dynamically adjust the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

[0151] Specifically, the data collection module can obtain comprehensive historical data of the distribution network and historical power consumption load data, and ensure the accuracy of the data through time series processing and feature extraction of the data. The power supply demand module and the load change module respectively utilize a long short-term memory network and a time series prediction algorithm to construct and train a power supply demand prediction model and a load change model, which can identify and predict the power supply demand and load change trend. The trend margin module identifies the operation data of the distribution network in real time, and provides power supply capacity change trend and load margin information. The target optimization module optimizes and solves a comprehensive target function constructed based on the power supply capacity and the load margin through a genetic algorithm, and outputs an optimal power supply decision scheme. The scheme adjustment module can dynamically adjust the optimal power supply decision scheme according to real-time monitoring data, and ensure the stability of the power supply system.

[0152] As another optional embodiment, the present application further provides an electronic device, comprising at least one processor, at least one memory, a communication interface and a bus,

[0153] Wherein, the processor, memory, communication interface complete mutual communication through the bus, the memory stores program instructions that can be executed by the processor, the processor calls the program instructions, to realize a kind of based on machine learning's distribution network power supply multidimensional auxiliary decision analysis method.

[0154] As another optional embodiment, the present application further provides a computer readable storage medium, the computer readable storage medium stores computer instructions, the computer instructions make computer realize the all or part steps of the kind of based on machine learning's distribution network power supply multidimensional auxiliary decision analysis method described in the embodiment of the present application.The storage medium includes: U disk, mobile hard disk, read-only memory ROM, random access memory RAM, magnetic disk or optical disk and various can store program code medium.

[0155] The above only describes the preferred embodiments of the present application, and does not limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A multi-dimensional auxiliary decision analysis method for power distribution network based on machine learning, characterized in that, Includes the following steps: S1. Obtain historical data of the power distribution network and historical power load data; S2. Construct an initial power demand prediction model through a long short-term memory network, and train and verify the initial power demand prediction model based on the historical data of the distribution network to obtain the power demand prediction model. In step S2, a power demand forecasting model is constructed, which specifically includes: An initial power demand forecasting model is constructed using a Long Short-Term Memory (LSTM) network. Historical distribution network data is input into the initial power demand forecasting model to obtain the power demand forecasting model. The calculation formula for the power demand forecasting model is as follows: in, For the predicted power demand at time t, For the model center weights, The hidden state at time t. To adjust the coefficient, Here, n represents the processed data from data source i at time t, and n is the number of data sources. The weight coefficients for data source i. and These are the first error correction coefficient and the second error correction coefficient, respectively. This is the trend adjustment coefficient. For trend functions; S3. Construct an initial load change model using a time-series prediction algorithm, and train and validate the initial load change model based on the historical electricity load data to obtain the load change model; S4. Based on the power demand prediction model and the load change model, identify the real-time operation data of the distribution network to obtain the power supply capacity change trend and load margin; S5. Construct a comprehensive objective function based on the trend of power supply capacity change and load margin, optimize the objective function using a genetic algorithm, and output the optimal power supply decision scheme; Step S5 includes: S51. Determine the power supply capacity index and load margin index based on the power supply capacity change trend and load margin. S52. A weighted allocation method is used to assign different importance weights to the power supply capacity index and the load margin index, constructing a comprehensive objective function. The formula for the comprehensive objective function is: Where F is the value of the comprehensive objective function, This represents the actual value of the k-th power supply capacity indicator. This is the actual value of the first load margin index. To standardize the maximum value of the power supply capacity index, To standardize the maximum value of the load margin index, The weighting coefficients for the power supply capacity index. , where N is the weighting coefficient of the load margin index, N is the number of power supply capacity indexes, and M is the number of load margin indexes. S53. Optimize and solve the comprehensive objective function using a genetic algorithm, initialize the population, and generate an initial population; S54. Evaluate the fitness of each individual in the initial population based on the comprehensive objective function, and select the individual with the highest fitness based on the elite preservation strategy; S55. Generate a new generation of individuals through partial random crossover and mutation, and repeatedly iterate to generate new individuals. When the iteration reaches the preset number of iterations or the fitness change meets the requirements, stop the iteration and output the optimal power supply decision scheme. S6. Execute the optimal power supply decision scheme, monitor the power supply operation parameters of the distribution network in real time, and dynamically adjust the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

2. The multi-dimensional auxiliary decision analysis method for power distribution network based on machine learning as described in claim 1, characterized in that, In step S1, historical distribution network data and historical electricity load data are obtained, specifically including: S11. Collect initial historical data of the distribution network and initial historical power load data respectively; S12. Perform time-series processing and feature extraction on the initial distribution network historical data to obtain distribution network historical data; S13. Perform periodic analysis and outlier processing on the initial historical electricity load data to obtain historical electricity load data.

3. The multi-dimensional auxiliary decision analysis method for power distribution network based on machine learning as described in claim 1, characterized in that, Step S3 includes: S31. Use a long short-term memory network to build a deep learning model and extract the long-term variation characteristics of historical electricity load data; S32. The periodic and short-term fluctuation characteristics of historical electricity load data are analyzed using an autoregressive integral moving average model. S33. The prediction results of the deep learning model and the autoregressive integral moving average model are combined using an adaptive weight fusion algorithm to obtain the combined prediction result. S34. Input the combined prediction results into the initial load change model for training and validation to obtain the load change model. The calculation formula of the load change model is: in, For time Predicted electricity load at that time For time The actual power load at that time For time The actual power load at that time For time The j-th characteristic affecting the electricity load at any given time, where m is the number of characteristics affecting the electricity load. for Feature weights, For time External interference factors at the time Adjustable parameters based on actual power load. To adjust parameters that affect electricity load, Adjustable parameters to account for external interference factors. The loss function value for the initial load change model. Mean square error, The mean absolute error, The root mean square percentage error. Mean square error, Mean absolute error The weighting coefficient for the root mean square percentage error. For accuracy, To predict the average electricity load, This represents the average of the actual electricity load.

4. The multi-dimensional auxiliary decision analysis method for power distribution network based on machine learning as described in claim 3, characterized in that, Step S4 includes: S41. Real-time operation data of the distribution network during operation is collected through sensors and monitoring equipment; S42. Based on the power supply demand prediction model and the load change model, identify the real-time operation data of the distribution network to obtain the real-time predicted power supply demand and the real-time predicted power load. S43. The real-time predicted power demand and the real-time predicted power load are classified and regressed using the support vector machine algorithm to obtain the future power supply capacity change trend. The formula for calculating the power supply capacity change trend is: in, For time The trend of power supply capacity changes over time. For time Voltage at that time For time Current at time For time Frequency of time For voltage weighting coefficients, For the current, the weighting factor is... The weighting coefficients for frequency; S44. The real-time predicted electricity load is calculated based on the fuzzy logic algorithm to obtain the load margin. The formula for calculating the load margin is: For time Load margin at that time For time Real-time forecast of electricity load For time The real-time actual power load at that time.

5. The multi-dimensional auxiliary decision analysis method for power distribution network based on machine learning as described in claim 1, characterized in that, Step S6 includes: S61. Based on the optimal power supply decision scheme, adjust the operating parameters of the equipment in the distribution network and redistribute the load demand of the equipment; S62. Through the distribution network control center, the adjusted operating parameters and redistributed load requirements are issued to the corresponding equipment in the form of control commands; S63. Deploy IoT sensors on key equipment in the power distribution network, and collect power supply operation parameters in real time based on the IoT sensors; S64. The power supply operation parameters are cleaned, filtered and normalized to obtain the power distribution network operation data, and the power distribution network operation data is transmitted to the centralized monitoring platform for real-time storage and backup. S65. Based on the power distribution network operation data, analyze the current power supply status, identify potential anomalies and operational bottlenecks, and obtain status assessment results; S66. Based on the status assessment results and combined with historical data, formulate and implement corresponding adjustment strategies to dynamically adjust the optimal power supply decision scheme.

6. A multi-dimensional auxiliary decision analysis system for power distribution network based on machine learning, characterized in that, The system includes: The data acquisition module is used to acquire historical data of the distribution network and historical electricity load data; The power supply demand module is used to construct an initial power supply demand prediction model through a long short-term memory network, and to train and verify the initial power supply demand prediction model based on the historical data of the distribution network to obtain the power supply demand prediction model. Constructing a power demand forecasting model specifically includes: An initial power demand forecasting model is constructed using a Long Short-Term Memory (LSTM) network. Historical distribution network data is input into the initial power demand forecasting model to obtain the power demand forecasting model. The calculation formula for the power demand forecasting model is as follows: in, For the predicted power demand at time t, For the model center weights, The hidden state at time t. To adjust the coefficient, Here, n represents the processed data from data source i at time t, and n is the number of data sources. The weight coefficients for data source i. and These are the first error correction coefficient and the second error correction coefficient, respectively. This is the trend adjustment coefficient. For trend functions; The load change module is used to construct an initial load change model through a time-series prediction algorithm, and to train and verify the initial load change model based on the historical electricity load data to obtain the load change model. The trend margin module, based on the power demand prediction model and the load change model, identifies the real-time operation data of the distribution network to obtain the power supply capacity change trend and load margin. The objective optimization module is used to construct a comprehensive objective function based on the trend of power supply capacity changes and load margin, and to optimize and solve the comprehensive objective function through a genetic algorithm to output the optimal power supply decision scheme. The target optimization module is specifically used for execution: S51. Determine the power supply capacity index and load margin index based on the power supply capacity change trend and load margin. S52. A weighted allocation method is used to assign different importance weights to the power supply capacity index and the load margin index, constructing a comprehensive objective function. The formula for the comprehensive objective function is: Where F is the value of the comprehensive objective function, This represents the actual value of the k-th power supply capacity indicator. This is the actual value of the first load margin index. To standardize the maximum value of the power supply capacity index, To standardize the maximum value of the load margin index, The weighting coefficients for the power supply capacity index. , where N is the weighting coefficient of the load margin index, N is the number of power supply capacity indexes, and M is the number of load margin indexes. S53. Optimize and solve the comprehensive objective function using a genetic algorithm, initialize the population, and generate an initial population; S54. Evaluate the fitness of each individual in the initial population based on the comprehensive objective function, and select the individual with the highest fitness based on the elite preservation strategy; S55. Generate a new generation of individuals through partial random crossover and mutation, and repeatedly iterate to generate new individuals. When the iteration reaches the preset number of iterations or the fitness change meets the requirements, stop the iteration and output the optimal power supply decision scheme. The scheme adjustment module is used to execute the optimal power supply decision scheme, monitor the power supply operation parameters of the distribution network in real time, and dynamically adjust the optimal power supply decision scheme based on the power supply operation parameters of the distribution network.

7. An electronic device, characterized in that, include: At least one processor, at least one memory, a communication interface, and a bus; The processor, memory, and communication interface communicate with each other through the bus. The memory stores program instructions that can be executed by the processor. The processor calls the program instructions to implement the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause the computer to perform the method as described in any one of claims 1 to 5.

Images