A method for load optimization control of a cable branch box

Abstract

The application belongs to the technical field of power equipment, and discloses a kind of cable branch box load optimization control method;Including the following steps:S1: the health parameter of cable branch box is collected, and the health parameter is normalized processing.The application realizes the optimal configuration of distribution network resource by load transfer cost coefficient prediction and regional collaborative scheduling technology, greatly improves the utilization rate of distribution network equipment, reduces the operation cost of power grid and user power failure loss, and the traditional load transfer mode only considers the number of switch actions, which leads to unreasonable transfer scheme, while the application constructs a load transfer cost coefficient prediction model, outputs the predicted load transfer cost coefficient of transferable branch box, generates the optimal transfer scheme with the minimum total cost as the target, and defines the collaborative scheduling range of 3km power supply radius, fully excavates the idle capacity of adjacent branch box, improves the average load rate of distribution network equipment, and delays the investment demand of new distribution facilities.

Description

Technical Field

[0001] This invention relates to the field of power equipment technology, and more specifically, to a method for optimizing load control of cable branch boxes. Background Technology

[0002] Cable branch boxes, as core equipment for branching, switching and tapping cable lines in power distribution networks, are widely used in urban residential areas, industrial parks and commercial complexes. They are key nodes to ensure reliable power supply in power distribution networks. With the acceleration of urbanization and the expansion of new energy load access, the load fluctuation of power distribution networks is becoming increasingly severe. The load control mode of traditional cable branch boxes can no longer meet the operation requirements of modern power grids.

[0003] In existing technologies, cable branch boxes generally adopt a static load control method based on factory rated values. Overload risk is judged solely based on the rated capacity marked on the equipment nameplate, resulting in some equipment being in a state of hidden overload for a long time, accelerating insulation aging, and causing joint burnout, short circuits, or even fires. The existing control mode is mainly passive overload protection, which only triggers the circuit breaker to trip when the load exceeds the rated value. This can easily cause sudden large-scale power outages, seriously affecting the user's power experience. Most existing branch boxes operate in an independent mode, lacking a load coordination and scheduling mechanism between adjacent branch boxes. This often results in severe overload in local areas and a large amount of idle capacity in surrounding branch boxes, which is a waste of resources. In terms of load transfer decision-making, existing technologies mostly adopt the principle of shortest path or minimum number of switching operations, resulting in poor economic efficiency of transfer schemes and even accelerating equipment damage. Summary of the Invention

[0004] To address the problems in the background art, this invention proposes a method for optimizing the load control of cable branch boxes.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing load control of a cable branch box, comprising the following steps: S1: Collect the health parameters of the cable branch box and normalize the health parameters; S2: Generate the equipment health score of the branch box based on the processed health parameters. Obtain the dynamic load bearing coefficient based on the equipment health score and combined with the ambient temperature, humidity and running time. Then calculate the dynamic rated load and load risk score, and classify the branch box according to the load risk score. S3: Construct a load curve prediction model for the target branch box, output the predicted load curve for the target branch box for the next 24 hours through the model, simultaneously predict the dynamic rated load time series curve, calculate the hourly remaining capacity, and identify the overload time window and the maximum load to be transferred. S4: Taking the branch box that needs support as the center, define an electrical connection range with a power supply radius of 3km, calculate the minimum remaining capacity of all branch boxes within the range for the next 24 hours, and screen out the transferable branch boxes with a remaining capacity greater than 0. S5: Construct a load transfer cost coefficient prediction model for transferable branch boxes. Output the predicted load transfer cost coefficient for each transferable branch box through the model, sort the coefficients from smallest to largest, and combine the coefficient threshold with the total load to be transferred to determine the final transfer branch box and the load it will receive.

[0006] Furthermore, cable branch boxes are distributed throughout the city for cable line branching, transfer, and connection. Health parameters of each branch box are collected by data acquisition equipment. These health parameters include cable joint temperature, box internal temperature, inlet and outlet insulation resistance, grounding current, and vibration amplitude. The health parameters are then normalized.

[0007] Furthermore, the process of generating a device health score for the branch box based on the processed health parameters includes: Equipment health score H:

[0008] In the formula, , , , and These are weighting coefficients, obtained through training based on historical data.

[0009] Furthermore, the process of calculating the dynamic rated load and load risk score by obtaining the dynamic load bearing factor based on the equipment health score and adjusting for ambient temperature, humidity, and operating time includes: By combining equipment health scores and environmental data, the factory rated load of the branch box is corrected in real time to obtain the dynamic load bearing coefficient K:

[0010] In the formula, For ambient temperature, Where z represents the ambient relative humidity, and z represents the cumulative operating time of the equipment; Branch box dynamic rated load ,in, This is the original factory rated load; Calculate the current load factor R:

[0011] In the formula, This represents the current total active power. By combining the 15-minute load change rate dP / dt, the load risk score S is obtained:

[0012] In the formula, 0.7 and 0.3 are weighting coefficients. The load change rate is 15 minutes.

[0013] Furthermore, the process of classifying branch boxes based on load risk scores includes: The branch boxes are classified into three levels based on the load risk score: Low-risk branch box: S<0.6, operating normally, monitored once per hour; Medium-risk branch box: 0.6≤S<0.8, closely monitor, once every 10 minutes; High-risk branch box: S≥0.8, immediately initiate load optimization scheduling process; Mark the medium-risk branch boxes and high-risk branch boxes as target branch boxes.

[0014] Furthermore, the process of constructing a load curve prediction model for the target branch box includes: The load curve refers to the hourly predicted load curve of the target branch box for the next 24 hours; Factors affecting the load curve include: user electricity consumption characteristics, weather forecasts, holiday information, historical load for the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use pricing information. Set a monitoring period and obtain historical load curve data of a single target branch box in different monitoring periods. The historical load curve data includes the user electricity consumption characteristics, weather forecast, holiday information, historical load at the same time, upstream transformer load, line impedance, construction activity information, emergency information, time-of-use electricity price information, and the historical load curve of the single target branch box in different monitoring periods. Based on the user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency information, time-of-use pricing information, and corresponding historical load curves of the target branch boxes in different historical load curve data, a load curve prediction set is generated and divided into the first training set and the first test set. A first convolutional neural network is constructed, and user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency event information, and time-of-use electricity price information from different historical load curve data in the first training set are used as the input data of the first convolutional neural network, and the corresponding historical load curves in the first training set are used as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the load curve prediction model. At the end of each monitoring cycle, the user electricity consumption characteristics, weather forecast, holiday information, historical load during the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use electricity price information of each target branch box within the monitoring cycle are input into the load curve prediction model to obtain the predicted load curve for each target branch box.

[0015] Furthermore, based on the dynamic load carrying capacity factor, combined with the predicted ambient temperature and cumulative equipment operating time for the next 24 hours, the dynamic rated load for each hour is predicted, and a dynamic rated load time series curve is generated. The predicted load curve and the dynamic rated load curve are plotted on the same coordinate system, and the remaining capacity at each moment is calculated.

[0016] In the formula, Let be the remaining capacity at time t. The dynamic rated load at time t, Let be the predicted load at time t; Overload judgment and load transfer calculation process: Iterate through the remaining capacity at all times in the next 24 hours. If any time exists... If the value is less than 0, mark the branch box as a branch box that needs support and record the overload start time. Overload end time and maximum overload The maximum overload is the total load that needs to be transferred. If all moments If the value is ≥0, then the branch box is marked as a transferable branch box, and its remaining capacity can be used to take over the transferred load of other branch boxes.

[0017] Furthermore, taking the branch box requiring support as the geographical center and electrical core, a power supply radius of 3km is defined as the load coordination dispatch range. Within this range, all cable branch boxes that have direct or indirect electrical connections with the branch box requiring support are selected as the initial branch boxes. For all initially selected branch boxes, based on their dynamic rated load time series curve and predicted load time series curve, the remaining capacity is calculated hourly and the minimum remaining capacity value for the next 24 hours is extracted. Finally, branch boxes with a minimum remaining capacity value greater than 0 are selected and marked as transferable branch boxes as candidates to take over the load transferred from the branch boxes that need support.

[0018] Furthermore, the process of constructing a load transfer cost coefficient prediction model for transferable branch boxes includes: The load transfer cost factor refers to the cost of a transferable branch box accommodating transferred loads; Factors affecting the load transfer cost coefficient include: the current load risk score of the transferable branch box, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score. Obtain historical load transfer cost coefficient data for a single transferable branch box. The historical load transfer cost coefficient data includes the current load risk score, electrical distance, remaining capacity margin, number of switch actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score, and historical load transfer cost coefficient of the single transferable branch box within the monitoring period. Based on the current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score and corresponding historical load transfer cost coefficient of the corresponding transferable branch box in different historical load transfer cost coefficient data, a load transfer cost coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation and current equipment health score in the different historical load transfer cost coefficient data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical load transfer cost coefficient in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the load transfer cost coefficient prediction model. The current load risk score, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score of each transferable branch box are input into the load transfer cost coefficient prediction model to obtain the predicted load transfer cost coefficient for each transferable branch box.

[0019] Furthermore, based on the load transfer cost coefficient prediction model, the predicted load transfer cost coefficients of all candidate transferable branch boxes are obtained. All candidate transferable branch boxes are sorted in ascending order of predicted load transfer cost coefficients. A coefficient threshold is set, and only transferable branch boxes with coefficients less than the coefficient threshold are selected to participate in load transfer. Starting with the first transferable branch box in the sorting list, the transfer load is allocated sequentially, with each allocated load not exceeding the minimum remaining capacity of that transferable branch box, until the cumulative allocated load reaches the total load to be transferred. Finally, the list of transferable branch boxes participating in the transfer and the load capacity of each transferable branch box are determined.

[0020] The technical effects and advantages of the cable branch box load optimization control method of the present invention are as follows: (1) By using load transfer cost coefficient prediction and regional collaborative scheduling technology, the optimal allocation of distribution network resources has been achieved, the utilization rate of distribution network equipment has been greatly improved, the power grid operation cost and user power outage losses have been reduced. Traditional load transfer methods only consider the number of switching actions, resulting in unreasonable transfer schemes. However, this invention constructs a load transfer cost coefficient prediction model, outputs the predicted load transfer cost coefficient of transferable branch boxes, and generates the optimal transfer scheme with the goal of minimizing the total cost. By defining a collaborative scheduling range of 3km power supply radius, the idle capacity of adjacent branch boxes is fully explored, the average load rate of distribution network equipment is increased, the investment demand for new power distribution facilities is delayed, and non-critical user loads are transferred first, avoiding power outages for primary and secondary critical users.

[0021] (2) Through the dynamic load carrying capacity assessment driven by equipment health and the overload prediction technology of dual time-series curves, a fundamental transformation from passive trip protection to active and proactive prevention and control has been achieved, which significantly improves the reliability of power supply and the service life of equipment. The traditional static rated value assessment method cannot reflect the actual operating status of the equipment, resulting in insufficient accuracy of overload warning. However, this invention combines ambient temperature, humidity and running time to correct the dynamic rated load of the equipment in real time. By cross-comparing the load curve prediction model with the dynamic rated load time-series curve, the overload time window and the maximum load to be transferred can be accurately identified 24 hours in advance, which fundamentally avoids power outage accidents caused by overload tripping, improves the reliability of power supply, and avoids long-term overload operation of equipment by reasonably controlling the equipment load rate, which extends the average service life of cable branch boxes and reduces the cost of equipment replacement and maintenance. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0024] Reference Figure 1 A method for optimizing load control of cable branch boxes includes the following steps: S1: Data acquisition module: collects the health parameters of the cable branch box and normalizes the health parameters; S2: Data Calculation Module: Generates equipment health score for branch boxes based on processed health parameters, obtains dynamic load bearing coefficient based on equipment health score and combined with ambient temperature, humidity and running time, then calculates dynamic rated load and load risk score, and classifies branch boxes according to load risk score; S3: Load curve prediction module: Constructs a load curve prediction model for the target branch box, outputs the predicted load curve of the target branch box for the next 24 hours through the model, synchronously predicts the dynamic rated load time series curve, calculates the hourly remaining capacity, and identifies the overload time window and the maximum load to be transferred. S4: Portable Branch Box Screening Module: Taking the branch box that needs to be supported as the center, define an electrical connection range with a power supply radius of 3km, calculate the minimum remaining capacity of all branch boxes within the range for the next 24 hours, and screen out portable branch boxes with a remaining capacity greater than 0. S5: Load Transfer Cost Coefficient Prediction Module: Constructs a load transfer cost coefficient prediction model for transferable branch boxes, outputs the predicted load transfer cost coefficient for each transferable branch box through the model, sorts the coefficients from smallest to largest, and combines the coefficient threshold with the total load to be transferred to determine the final transfer branch box and the load it will receive.

[0025] It should be further explained that, in the specific implementation process, cable branch boxes are distributed in the city and are used for cable line branching, transfer and connection. The health parameters of each branch box are collected by the data acquisition equipment. The health parameters include cable joint temperature, box internal temperature, insulation resistance of incoming and outgoing lines, grounding current and vibration amplitude. The health parameters are then normalized. PT100 platinum resistance temperature sensors were installed at all incoming and outgoing cable joints in the cable branch box to measure the cable joint temperature, and the cable joint temperature was normalized.

[0026] In the formula, To normalize the cable joint temperature, This is the measured temperature of the cable joint. and These are the lower and upper limits for the cable joint temperature, specifically 20℃ and 120℃ respectively; A digital temperature and humidity sensor is installed at the top center of the branch box to measure the internal temperature, which is then normalized.

[0027] In the formula, To normalize the internal temperature of the chamber, This is the measured internal temperature of the chamber. and These are the lower and upper limits of the internal temperature of the chamber, specifically 20℃ and 80℃ respectively; The insulation resistance to ground of the incoming and outgoing cables is automatically measured weekly using an online insulation monitoring instrument and the DC superposition method. The insulation resistance of the incoming and outgoing cables is then normalized.

[0028] In the formula, To normalize the insulation resistance of incoming and outgoing lines, The measured insulation resistance of the incoming and outgoing lines. and These are the lower and upper limits of the insulation resistance of the incoming and outgoing lines, specifically 1MΩ and 200MΩ, respectively. A zero-sequence current transformer is installed on the grounding busbar of the branch box to monitor the current in the grounding loop in real time and normalize the grounding current.

[0029] In the formula, For normalized grounding current, This is the measured grounding current. and These are the lower and upper limits of the grounding current, specifically 0A and 1A respectively; Piezoelectric vibration sensors are installed at the circuit breaker contacts and busbar connections to collect mechanical vibration signals during equipment operation. The sampling frequency is 1 time / 10 minutes. The vibration amplitude is extracted by FFT analysis and then normalized.

[0030] In the formula, To normalize the vibration amplitude, The measured vibration amplitude is... and These represent the lower and upper limits of the vibration amplitude, specifically 0g and 1g, respectively.

[0031] It should be further explained that, in the specific implementation process, the process of generating the branch box's equipment health score based on the processed health parameters includes: Equipment health score H:

[0032] In the formula, , , , and These are weighting coefficients, obtained from training on historical data, and set to 0.4, 0.2, 0.2, 0.15, and 0.05 respectively. If the health parameter of a branch box is It is 70℃. It is 40℃. It is 80MΩ. It is 0.2A. If the concentration is 0.15g, then the equipment health parameter H of this branch box is 0.575.

[0033] It should be further explained that, in the specific implementation process, the dynamic load bearing coefficient is obtained based on the equipment health score and adjusted according to ambient temperature, humidity, and operating time. The process for calculating the dynamic rated load and load risk score includes: By combining equipment health scores and environmental data, the factory rated load of the branch box is corrected in real time to obtain the dynamic load bearing coefficient K:

[0034] In the formula, For ambient temperature, Where z represents the ambient relative humidity, and z represents the cumulative operating time of the equipment; A coefficient of 0.006 indicates that for every 1°C increase in ambient temperature, the equipment's load-bearing capacity decreases by 0.6%. A coefficient of 0.003 indicates that for every 1% increase in relative humidity, the load-bearing capacity decreases by 0.3%. A coefficient of 0.0002 indicates that the load-bearing capacity decreases by 0.02% for every 1000 hours of operation. Branch box dynamic rated load ,in, This is the original factory rated load; Calculate the current load factor R:

[0035] In the formula, The current total active power is measured in real time by a multi-function energy meter installed at the inlet of the branch box; By combining the 15-minute load change rate dP / dt, the load risk score S is obtained:

[0036] In the formula, 0.7 and 0.3 are weighting coefficients, determined through analysis of historical overload accident data. The 15-minute load change rate; The scaling factor is used because, under normal circumstances, the load change within 15 minutes usually does not exceed 0.1 times the rated load. Multiplying it by 10 can map the range of dP / dt to the interval [0, 1], making it consistent with the dimensions of the load factor R.

[0037] It should be further explained that, in the specific implementation process, the process of classifying branch boxes based on load risk scores includes: The branch boxes are classified into three levels based on the load risk score: Low-risk branch box: S<0.6, operating normally, monitored once per hour; Medium-risk branch box: 0.6≤S<0.8, closely monitor, once every 10 minutes; High-risk branch box: S≥0.8, immediately initiate load optimization scheduling process; Mark the medium-risk branch boxes and high-risk branch boxes as target branch boxes.

[0038] It should be further explained that, in the specific implementation process, the process of constructing the load curve prediction model for the target branch box includes: The load curve refers to the hourly predicted load curve of the target branch box for the next 24 hours; Factors affecting the load curve include: user electricity consumption characteristics, weather forecasts, holiday information, historical load for the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use pricing information. User electricity consumption characteristics are obtained from the electricity consumption information collection system, which collects user types (residential, commercial, industrial) and historical electricity consumption curves for each outgoing circuit. Weather forecasts obtain hourly temperature, humidity, and rainfall data for the next 24 hours from meteorological APIs. Holiday information is obtained from a calendar database, including information on holidays, weekends, and weekdays. Historical load for the same period refers to the actual load data for the same period over the past 7 days; The load of the upstream transformer refers to the real-time load and the predicted load of the upstream transformer. Line impedance parameters are obtained from the distribution network GIS system. Construction activity information is obtained from municipal departments regarding construction plans within the target area; Emergency information refers to early warnings for large-scale events and disasters; Time-of-use pricing information refers to how electricity prices at different times affect users' electricity consumption behavior; Set a monitoring period and obtain historical load curve data of a single target branch box in different monitoring periods. The historical load curve data includes the user electricity consumption characteristics, weather forecast, holiday information, historical load at the same time, upstream transformer load, line impedance, construction activity information, emergency information, time-of-use electricity price information, and the historical load curve of the single target branch box in different monitoring periods. Based on the user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency information, time-of-use pricing information, and corresponding historical load curves of the target branch boxes in different historical load curve data, a load curve prediction set is generated and divided into the first training set and the first test set. A first convolutional neural network is constructed, and user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency event information, and time-of-use electricity price information from different historical load curve data in the first training set are used as the input data of the first convolutional neural network, and the corresponding historical load curves in the first training set are used as the output data of the first convolutional neural network. The first convolutional neural network adopts a structure of 3 convolutional layers + 2 fully connected layers, as follows: Input layer: The input dimension is 9×24, corresponding to 24-hour time series data of 9 input features; First convolutional layer: 32 filters, 3×1 kernel size, stride 1, padding method is same, activation function is ReLU; Second convolutional layer: 64 filters, 3×1 kernel size, stride 1, padding method is same, activation function is ReLU; Third convolutional layer: 128 filters, 3×1 kernel size, stride 1, padding method is same, activation function is ReLU; Pooling layer: Global average pooling is used to compress the feature map into a 128-dimensional vector; First fully connected layer: 64 neurons, activation function is ReLU; The second fully connected layer has 24 neurons, the activation function is a linear function, and it outputs the hourly load values ​​for the next 24 hours. Training hyperparameter settings: The optimizer used is Adam, with an initial learning rate of 0.001, which decreases by 0.1 every 10 epochs; the batch size is 32; the loss function is mean squared error (MSE); training termination conditions: training stops when the mean squared error of the test set is less than or equal to 0.005 (after normalization) or when the number of training epochs reaches 100; measures to prevent overfitting: a Dropout layer is added after each convolutional layer and fully connected layer, with a dropout rate of 0.2; L2 regularization is added with a regularization coefficient of 0.0001; The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the load curve prediction model. At the end of each monitoring cycle, the user electricity consumption characteristics, weather forecast, holiday information, historical load during the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use electricity price information of each target branch box within the monitoring cycle are input into the load curve prediction model to obtain the predicted load curve for each target branch box.

[0039] It should be further explained that, in the specific implementation process, based on the dynamic load bearing coefficient, combined with the predicted ambient temperature value and the cumulative running time of the equipment in the next 24 hours, the dynamic rated load of each hour is predicted, and a dynamic rated load time series curve is generated. The predicted load curve and the dynamic rated load curve are plotted on the same coordinate system, and the remaining capacity at each moment is calculated.

[0040] In the formula, Let be the remaining capacity at time t. The dynamic rated load at time t, Let be the predicted load at time t; Overload judgment and load transfer calculation process: Iterate through the remaining capacity at all times in the next 24 hours. If any time exists... If the value is less than 0, mark the branch box as a branch box that needs support and record the overload start time. Overload end time and maximum overload The maximum overload is the total load that needs to be transferred. If all moments If the value is ≥0, then the branch box is marked as a transferable branch box, and its remaining capacity can be used to take over the transferred load of other branch boxes.

[0041] It should be further explained that, in the specific implementation process, the area with a power supply radius of 3km is defined as the load coordination and dispatch range, with the branch box that needs to be supported as the geographical center and electrical core. Within this range, all cable branch boxes that have direct or indirect electrical connections with the branch box that needs to be supported are selected as the initial branch boxes. For all initially selected branch boxes, based on their dynamic rated load time series curve and predicted load time series curve, the remaining capacity is calculated hourly and the minimum remaining capacity value for the next 24 hours is extracted. Finally, branch boxes with a minimum remaining capacity value greater than 0 are selected and marked as transferable branch boxes as candidates to take over the load transferred from the branch boxes that need support.

[0042] It should be further explained that, in the specific implementation process, the process of constructing the load transfer cost coefficient prediction model for the transferable branch box includes: The load transfer cost factor refers to the cost of a transferable branch box accommodating transferred loads; Factors affecting the load transfer cost coefficient include: the current load risk score of the transferable branch box, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score. Electrical distance refers to the electrical distance between the branch box that needs support and the relocatable branch box; Remaining capacity margin refers to the ratio of the minimum remaining capacity of a transferable branch box to its own dynamic rated load; The number of switch operations refers to the number of times the outgoing switch of the transferable branch box has been operated in the past 30 days. User importance level refers to the user importance level (Level 1, Level 2, Level 3) of the outgoing circuit of the transferable branch box. Historical transfer success rate refers to the proportion of successful load transfers in the past for this transferable branch box out of the total number of transfers; Line load rate refers to the current load rate of the line connecting the branch box that needs support and the transferable branch box; Voltage deviation refers to the current voltage deviation value of the transferable branch box; Obtain historical load transfer cost coefficient data for a single transferable branch box. The historical load transfer cost coefficient data includes the current load risk score, electrical distance, remaining capacity margin, number of switch actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score, and historical load transfer cost coefficient of the single transferable branch box within the monitoring period. Based on the current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score and corresponding historical load transfer cost coefficient of the corresponding transferable branch box in different historical load transfer cost coefficient data, a load transfer cost coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation and current equipment health score in the different historical load transfer cost coefficient data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical load transfer cost coefficient in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network adopts a structure of 2 convolutional layers + 2 fully connected layers, as follows: Input layer: The input dimension is 9×1, corresponding to one-dimensional data of 9 input features; First convolutional layer: 16 filters, 3×1 kernel size, stride 1, padding method is same, activation function is ReLU; Second convolutional layer: 32 filters, 3×1 kernel size, stride 1, padding method is same, activation function is ReLU; Pooling layer: Global average pooling is used to compress the feature map into a 32-dimensional vector; First fully connected layer: 16 neurons, ReLU activation function; The second fully connected layer has 1 neuron, uses the sigmoid function as the activation function, and outputs a load transfer cost coefficient between 0 and 1. Training hyperparameter settings: The optimizer used is Adam, with an initial learning rate of 0.0005, which decays by 0.1 every 15 epochs; the batch size is 16; the loss function is binary cross-entropy; training termination conditions: training stops when the test set accuracy is greater than or equal to 95% or the number of training epochs reaches 80; measures to prevent overfitting: a Dropout layer is added after each fully connected layer with a dropout rate of 0.3; L2 regularization is added with a regularization coefficient of 0.0005. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the load transfer cost coefficient prediction model. The current load risk score, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score of each transferable branch box are input into the load transfer cost coefficient prediction model to obtain the predicted load transfer cost coefficient for each transferable branch box.

[0043] It should be further explained that, in the specific implementation process, the predicted load transfer cost coefficients of all candidate transferable branch boxes are obtained based on the load transfer cost coefficient prediction model. All candidate transferable branch boxes are sorted in ascending order of predicted load transfer cost coefficients. A coefficient threshold is set, and only transferable branch boxes with coefficients less than the coefficient threshold are selected to participate in load transfer. Starting with the first transferable branch box in the sorting list, the transfer load is allocated sequentially, with each allocated load not exceeding the minimum remaining capacity of that transferable branch box, until the cumulative allocated load reaches the total load to be transferred. Finally, the list of transferable branch boxes participating in the transfer and the load capacity of each transferable branch box are determined.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0045] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method of load optimization control of a cable branch box, characterized by, Includes the following steps: S1: Collect the health parameters of the cable branch box and normalize the health parameters; S2: Generate the equipment health score of the branch box based on the processed health parameters. Obtain the dynamic load bearing coefficient based on the equipment health score and combined with the ambient temperature, humidity and running time. Then calculate the dynamic rated load and load risk score, and classify the branch box according to the load risk score. S3: Construct a load curve prediction model for the target branch box, output the predicted load curve for the target branch box for the next 24 hours through the model, simultaneously predict the dynamic rated load time series curve, calculate the hourly remaining capacity, and identify the overload time window and the maximum load to be transferred. S4: Taking the branch box that needs support as the center, define an electrical connection range with a power supply radius of 3km, calculate the minimum remaining capacity of all branch boxes within the range for the next 24 hours, and screen out the transferable branch boxes with a remaining capacity greater than 0. S5: Construct a load transfer cost coefficient prediction model for transferable branch boxes. Output the predicted load transfer cost coefficient for each transferable branch box through the model, sort the coefficients from smallest to largest, and combine the coefficient threshold with the total load to be transferred to determine the final transfer branch box and the load it will receive.

2. The cable branch box load optimization control method of claim 1, wherein, Cable branch boxes are distributed throughout the city and are used for branching, switching and connecting cable lines. Health parameters of each branch box are collected by data acquisition equipment. These health parameters include cable joint temperature, box internal temperature, insulation resistance of incoming and outgoing lines, grounding current and vibration amplitude. The health parameters are then normalized.

3. The cable branch box load optimization control method according to claim 2, characterized in that, The process of generating a branch box equipment health score based on the processed health parameters includes: Equipment health score H: In the formula, , , , and These are weighting coefficients, obtained through training based on historical data.

4. The cable branch box load optimization control method according to claim 3, characterized in that, The process of calculating the dynamic load capacity factor based on the equipment health score and adjustments made for ambient temperature, humidity, and operating time includes: By combining equipment health scores and environmental data, the factory rated load of the branch box is corrected in real time to obtain the dynamic load bearing coefficient K: In the formula, For ambient temperature, Where z represents the ambient relative humidity, and z represents the cumulative operating time of the equipment; Branch box dynamic rated load ,in, This is the original factory rated load; Calculate the current load factor R: In the formula, This represents the current total active power. By combining the 15-minute load change rate dP / dt, the load risk score S is obtained: In the formula, 0.7 and 0.3 are weighting coefficients. The load change rate is 15 minutes.

5. The cable branch box load optimization control method according to claim 4, characterized in that, The process of classifying branch boxes based on load risk scores includes: The branch boxes are classified into three levels based on the load risk score: Low-risk branch box: S<0.6, operating normally, monitored once per hour; Medium-risk branch box: 0.6≤S<0.8, closely monitor, once every 10 minutes; High-risk branch box: S≥0.8, immediately initiate load optimization scheduling process; Mark the medium-risk branch boxes and high-risk branch boxes as target branch boxes.

6. The cable branch box load optimization control method according to claim 5, characterized in that, The process of constructing a load curve prediction model for the target branch box includes: The load curve refers to the hourly predicted load curve of the target branch box for the next 24 hours; Factors affecting the load curve include: user electricity consumption characteristics, weather forecasts, holiday information, historical load for the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use pricing information. Set a monitoring period and obtain historical load curve data of a single target branch box in different monitoring periods. The historical load curve data includes the user electricity consumption characteristics, weather forecast, holiday information, historical load at the same time, upstream transformer load, line impedance, construction activity information, emergency information, time-of-use electricity price information, and the historical load curve of the single target branch box in different monitoring periods. Based on the user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency information, time-of-use pricing information, and corresponding historical load curves of the target branch boxes in different historical load curve data, a load curve prediction set is generated and divided into the first training set and the first test set. A first convolutional neural network is constructed, and user electricity consumption characteristics, weather forecasts, holiday information, historical loads during the same period, upstream transformer loads, line impedance, construction activity information, emergency event information, and time-of-use electricity price information from different historical load curve data in the first training set are used as the input data of the first convolutional neural network, and the corresponding historical load curves in the first training set are used as the output data of the first convolutional neural network. The first convolutional neural network is trained to obtain the first initial convolutional neural network. The first initial convolutional neural network is validated using the first test set. The first initial convolutional neural network that outputs a value less than or equal to the preset first test error threshold is used as the load curve prediction model. At the end of each monitoring cycle, the user electricity consumption characteristics, weather forecast, holiday information, historical load during the same period, upstream transformer load, line impedance, construction activity information, emergency event information, and time-of-use electricity price information of each target branch box within the monitoring cycle are input into the load curve prediction model to obtain the predicted load curve for each target branch box.

7. The cable branch box load optimization control method according to claim 6, characterized in that, Based on the dynamic load carrying capacity factor, combined with the predicted ambient temperature and cumulative equipment operating time for the next 24 hours, the dynamic rated load for each hour is predicted, and a dynamic rated load time series curve is generated. The predicted load curve and the dynamic rated load curve are plotted on the same coordinate system, and the remaining capacity at each moment is calculated. In the formula, Let be the remaining capacity at time t. The dynamic rated load at time t, Let be the predicted load at time t; Overload judgment and load transfer calculation process: Iterate through the remaining capacity at all times in the next 24 hours. If any time exists... If the value is less than 0, mark the branch box as a branch box that needs support and record the overload start time. Overload end time and maximum overload The maximum overload is the total load that needs to be transferred. If all moments If the value is ≥0, then the branch box is marked as a transferable branch box, and its remaining capacity can be used to take over the transferred load of other branch boxes.

8. The cable branch box load optimization control method according to claim 7, characterized in that, Taking the branch box requiring support as the geographical center and electrical core, a power supply radius of 3km is defined as the load coordination dispatch range. Within this range, all cable branch boxes that have direct or indirect electrical connections with the branch box requiring support are selected as the initial branch boxes. For all initially selected branch boxes, based on their dynamic rated load time series curve and predicted load time series curve, the remaining capacity is calculated hourly and the minimum remaining capacity value for the next 24 hours is extracted. Finally, branch boxes with a minimum remaining capacity value greater than 0 are selected and marked as transferable branch boxes as candidates to take over the load transferred from the branch boxes that need support.

9. The cable branch box load optimization control method according to claim 8, characterized in that, The process of constructing a load transfer cost coefficient prediction model for a transferable branch box includes: The load transfer cost factor refers to the cost of a transferable branch box accommodating transferred loads; Factors affecting the load transfer cost coefficient include: the current load risk score of the transferable branch box, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score. Obtain historical load transfer cost coefficient data for a single transferable branch box. The historical load transfer cost coefficient data includes the current load risk score, electrical distance, remaining capacity margin, number of switch actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score, and historical load transfer cost coefficient of the single transferable branch box within the monitoring period. Based on the current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation, current equipment health score and corresponding historical load transfer cost coefficient of the corresponding transferable branch box in different historical load transfer cost coefficient data, a load transfer cost coefficient prediction set is generated and divided into a second training set and a second test set. A second convolutional neural network is constructed. The current load risk score, electrical distance, remaining capacity margin, number of switching actions, user importance level, historical transfer success rate, line load rate, voltage deviation and current equipment health score in the different historical load transfer cost coefficient data in the second training set are used as the input data of the second convolutional neural network, and the corresponding historical load transfer cost coefficient in the second training set is used as the output data of the second convolutional neural network. The second convolutional neural network is trained to obtain the second initial convolutional neural network. The second initial convolutional neural network is validated using the second test set. The second initial convolutional neural network that outputs a second test error threshold less than or equal to the preset second test error threshold is used as the load transfer cost coefficient prediction model. The current load risk score, electrical distance, remaining capacity margin, number of switch operations, user importance level, historical transfer success rate, line load rate, voltage deviation, and current equipment health score of each transferable branch box are input into the load transfer cost coefficient prediction model to obtain the predicted load transfer cost coefficient for each transferable branch box.

10. The cable branch box load optimization control method according to claim 9, characterized in that, Based on the load transfer cost coefficient prediction model, the predicted load transfer cost coefficients of all candidate transferable branch boxes are obtained. All candidate transferable branch boxes are sorted in ascending order of predicted load transfer cost coefficients. A coefficient threshold is set, and only transferable branch boxes with coefficients less than the coefficient threshold are selected to participate in load transfer. Starting with the first transferable branch box in the sorting list, the transfer load is allocated sequentially, with each allocated load not exceeding the minimum remaining capacity of that transferable branch box, until the cumulative allocated load reaches the total load to be transferred. Finally, the list of transferable branch boxes participating in the transfer and the load capacity of each transferable branch box are determined.

Images