Intelligent cable branch box adaptive load control method and system

Abstract

The present application belongs to the technical field of power system distribution automation, and particularly relates to a self-adaptive load control method and system for an intelligent cable branch box. The method comprises real-time acquisition and preprocessing of current data of a branch circuit and temperature data at a cable joint, obtaining of a data confidence factor, and extraction of a current change rate and a temperature rise rate. A load trend prediction model is used to obtain a load prediction value in a future set time period. A residual thermal tolerance time to reach a limit tolerance temperature is calculated. Based on a self-adaptive load control decision model and a particle swarm optimization algorithm, a load control adjustment coefficient is calculated. According to the load control adjustment coefficient, a current state interval is determined, and in combination with the residual thermal tolerance time, a hierarchical response strategy of maintaining through-flow, short-time overload monitoring and early warning, and branch current limiting is executed by an execution end. The present application combines load prediction with a transient thermal circuit model to calculate the residual thermal tolerance time, and uses a particle swarm optimized radial basis function neural network decision to realize hierarchical response of load control.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of power distribution automation, and particularly relates to a self-adaptive load control method and system for an intelligent cable branch box. BACKGROUND

[0002] The cable branch box is an intermediate connection device in a power distribution network system, and is mainly used for switching, branching and power distribution of cable lines. In the process of urban power distribution network construction, the cable branch box solves the problem of multi-branch connection in long-distance cable transmission. The intelligent cable branch box is based on a traditional metal box structure, and integrates electronic sensing technology, microcomputer processing technology and network communication technology. The intelligent cable branch box is internally configured with a high-voltage load switch, a current-voltage transformer and an intelligent monitoring terminal. The intelligent cable branch box establishes data connection with a power distribution automation master station through an optical fiber network and a wireless communication module, realizes remote monitoring of the operation state of a power distribution line, fault area positioning and load data uploading, improves the observability of the power distribution network, and is a basic support unit for building an intelligent power distribution network and realizing power distribution automation.

[0003] In order to solve the problems of insufficient load monitoring accuracy and limited power supply reliability of the power distribution line, the prior art usually adopts a method of setting a fixed current protection threshold for load control. However, zero drift generated by the sensor in a harsh environment will cause distortion of the collected data. The fixed threshold value logic cannot verify the authenticity of the data, thereby causing misoperation of the circuit breaker. In addition, the fixed threshold value ignores the influence of environmental temperature difference on cable heat dissipation and the thermal inertia physical characteristics of the cable body. In the working condition that the load is short-time over-limit but the temperature of the cable insulation layer does not reach the limit value, the power supply circuit is forcibly cut off, thereby causing power supply interruption of the cable line and waste of cable dynamic transportation potential. SUMMARY

[0004] The application aims to provide a self-adaptive load control method and system for an intelligent cable branch box. Distorted data is removed through electro-thermal coupling verification. The residual heat tolerance time is calculated by combining load prediction and transient thermal circuit model. The radial basis function neural network is optimized by particle swarm optimization to realize hierarchical response of load control, and the continuity of power supply and the safety of equipment are considered.

[0005] To solve the above technical problems, the technical solution adopted by the application is as follows: a self-adaptive load control method for an intelligent cable branch box, comprising the following steps:

[0006] Real-time collection and preprocessing of current data of a branch circuit and temperature data at a cable joint, electro-thermal coupling reliability verification of the preprocessed current data and temperature data, acquisition of a data confidence factor, and extraction of a current change rate and a temperature rise rate;

[0007] input the pretreated current data into the load trend prediction model to obtain a load prediction value in a future set time period;

[0008] In combination with the load prediction value, the residual thermal tolerance time to reach the limit tolerance temperature is calculated by using a transient thermal circuit equivalent model, according to the thermal characteristic parameters and limit tolerance temperature of the cable insulation material, and by solving through a finite difference iteration algorithm.

[0009] The data confidence factor, the current change rate, the temperature rise rate, the load prediction value and the residual thermal tolerance time are input into an adaptive load control decision model adopting a radial basis function neural network architecture, and a particle swarm optimization algorithm is used for training and optimization to calculate and output a load control adjustment coefficient.

[0010] According to the load control adjustment coefficient, a current state interval is determined, and in combination with the residual thermal tolerance time, a hierarchical response strategy of maintaining through-flow, short-time overload monitoring and early warning and disconnection current limiting is controlled to be executed by an execution end.

[0011] Preferably, the process of collecting and pre-processing the current data of the branch circuit and the temperature data at the cable joint in real time comprises:

[0012] A Hall current sensor is arranged at the conductor of the branch circuit to sense the magnetic field generated when the branch circuit conductor is energized, output a Hall voltage signal corresponding to the magnetic induction intensity, and obtain the current data at each discrete time according to the proportional relationship between the Hall voltage signal and the current;

[0013] An optical fiber Bragg grating temperature sensor is arranged on the surface of the insulation layer of the cable joint to receive a broadband light signal and reflect a narrowband light with a center wavelength, wherein the center wavelength drifts with the changes of the grating period and effective refractive index caused by the temperature change of the cable joint, the drift amount of the center wavelength is detected, and the temperature data at each discrete time is calculated according to the drift amount;

[0014] A discrete linear dynamic system model containing system state equations and observation equations is constructed, a process noise covariance matrix and a measurement noise covariance matrix are set, and a Kalman filtering algorithm is used to perform iterative calculation of time update and measurement update on the collected current data and temperature data, respectively, to output current optimal estimation values and temperature optimal estimation values filtered from noise interference, which are pretreated as current data and temperature data.

[0015] Preferably, the process of performing the electrical-thermal coupling reliability verification on the pretreated current data and temperature data to obtain the data confidence factor comprises:

[0016] The equivalent resistance parameters of the branch circuit, the joint heat dissipation coefficient and the environmental temperature parameters are obtained, an electrical-thermal coupling physical benchmark model is constructed based on Joule's law, the current optimal estimation value is input into the electrical-thermal coupling physical benchmark model, and the theoretical temperature rise value at the current time is calculated.

[0017] Calculate the absolute error between the temperature optimal estimation value and the theoretical temperature rise value, and take the absolute error as a residual index;

[0018] Based on the residual index, calculate the data confidence factor using the Gaussian membership function.

[0019] Preferably, the process of extracting the current rate of change and the temperature rise rate includes:

[0020] Obtain the current sampling time optimal current estimation value and temperature optimal estimation value, and the last sampling time optimal current estimation value and temperature optimal estimation value;

[0021] Calculate the current difference value of the current optimal estimation value between the current sampling time and the last sampling time, and divide the current difference value by the sampling time interval to obtain the current rate of change;

[0022] Calculate the temperature difference value of the temperature optimal estimation value between the current sampling time and the last sampling time, and divide the temperature difference value by the sampling time interval to obtain the temperature rise rate.

[0023] Preferably, the process of inputting the preprocessed current data into the load trend prediction model to obtain the load prediction value in the future set time period includes:

[0024] Set the length of the historical time window, adopt the sliding window mechanism, extract the historical current optimal estimation value, and construct the input time sequence vector;

[0025] Construct a load trend prediction model including an input layer, a long short-term memory network layer and an output layer, input the input time sequence vector into the long short-term memory network layer, and the long short-term memory network layer includes a preset number of long short-term memory network units, each long short-term memory network unit includes a forget gate, an input gate, an output gate and a cell state;

[0026] Receive the input data at the current time and the hidden layer state at the last time through the forget gate, calculate the forget gate coefficient through the activation function, and determine the retention degree of the cell state at the last time;

[0027] Receive the input data at the current time and the hidden layer state at the last time through the input gate, calculate the input gate coefficient and the candidate cell state, and determine the update degree of the new information at the current time;

[0028] Calculate the cell state at the current time by multiplying the cell state at the last time and the forget gate coefficient, and adding the product of the candidate cell state and the input gate coefficient;

[0029] The output gate receives the input data at the current time and the hidden layer state at the previous time, calculates the output gating coefficient, and multiplies the current time cell state after the activation function processing and the output gating coefficient to obtain the hidden layer state at the current time.

[0030] The hidden layer state at the current time is input into the output layer, linearly transformed by the weight matrix and bias term of the full connection layer, and the output current prediction sequence in the future set time period is calculated as the load prediction value.

[0031] Preferably, the process of calculating the residual thermal tolerance time to reach the limit tolerance temperature is solved by a finite difference iterative algorithm, which includes:

[0032] The thermal characteristic parameters include the equivalent thermal resistance parameter and the equivalent thermal capacity parameter, the equivalent thermal resistance parameter is calculated by using the thermal conductivity of the cable insulation material combined with the geometric size of the cable insulation layer, the equivalent thermal capacity parameter is calculated by using the specific heat capacity parameter of the cable insulation material combined with the density and volume of the cable insulation layer, and the transient thermal circuit equivalent model containing the equivalent thermal resistance parameter and the equivalent thermal capacity parameter is constructed;

[0033] The optimal estimation value of the temperature at the current time is taken as the initial temperature state for iterative calculation;

[0034] Based on the current prediction sequence in the future set time period and the transient thermal circuit equivalent model, the finite difference iterative algorithm is adopted to deduce the cable temperature value at each discrete time in the future step by step;

[0035] The calculated cable temperature value at each discrete time in the future is compared with the limit tolerance temperature of the cable insulation material one by one, when it is determined that the cable temperature value at a discrete time first exceeds the limit tolerance temperature, the time difference between the discrete time and the current time is calculated, and the time difference is taken as the residual thermal tolerance time.

[0036] Preferably, the particle swarm optimization algorithm is used for training and optimization, and the process of calculating the output load control adjustment coefficient includes:

[0037] The data confidence factor, the current change rate, the temperature rise rate, the load prediction value and the residual thermal tolerance time are normalized and combined to construct an input feature vector;

[0038] An adaptive load control decision model is constructed, which includes a decision input layer, a decision hidden layer and a decision output layer, wherein the decision hidden layer adopts a Gaussian radial basis function as an activation function, a historical operation data set containing input feature vectors at historical times and corresponding ideal load control adjustment coefficients is obtained as a model training sample;

[0039] The particle swarm optimization algorithm is used to optimize network parameters of the adaptive load control decision model, and an optimized adaptive load control decision model is obtained, wherein the network parameters include a basis function center vector of a decision hidden layer, a basis function width vector of the decision hidden layer, and output weights from the decision hidden layer to a decision output layer.

[0040] The particle swarm is initialized, the particle population size, the learning factor, the inertia weight, and the maximum iteration number are set, and the network parameters are encoded into the position vector of the particle.

[0041] The particle fitness evaluation is performed, the historical operation data set is input into the adaptive load control decision model, the network parameters corresponding to the current particle position vector are used to calculate the model output value, the mean square error between the model output value and the ideal load control adjustment coefficient is calculated, and the reciprocal of the mean square error is taken as the fitness function value.

[0042] The optimization extreme value is updated, the fitness function value of the current particle is compared with the individual historical optimal fitness function value of the current particle and the global historical optimal fitness function value of the population, respectively, if the fitness function value of the current particle is greater, the position of the current particle is used to replace the corresponding individual historical optimal position and global historical optimal position, respectively.

[0043] The iterative optimization is performed, the flight speed and spatial position of the particle are adjusted according to the individual historical optimal position and the global historical optimal position, the particle fitness evaluation and the optimization extreme value updating process are cyclically executed, until one of the maximum iteration number and the fitness function value meeting the convergence condition is reached, and the final global historical optimal position is decoded into the optimized basis function center vector, the basis function width vector and the output weight.

[0044] The input feature vector of the current time is input into the optimized adaptive load control decision model, the output of the Gaussian radial basis function is calculated through the decision hidden layer, and the linear weighted sum is performed through the decision output layer, so as to calculate the load control adjustment coefficient.

[0045] Preferably, the process of controlling the execution end to execute the hierarchical response strategy of maintaining through flow, short-time overload monitoring and early warning, and breaking limit current includes:

[0046] Based on the calculated load control adjustment coefficient and the remaining heat tolerance time, the preset yellow warning threshold and the preset red action threshold are set, and the load control adjustment coefficient is compared with the yellow warning threshold and the red action threshold, respectively.

[0047] When the load control adjustment coefficient is less than or equal to the yellow warning threshold, it is determined that the current state is in the normal interval, and the control execution end maintains the through flow state at the current time.

[0048] When the load control adjustment coefficient is greater than the yellow warning threshold and less than the red action threshold, it is determined that the current state is located in the short-time overload interval, a warning signal is generated, the control execution end maintains the current time flow state and enters the short-time overload monitoring and warning state, and the remaining heat tolerance time is used as an initial time length to start countdown monitoring;

[0049] When the load control adjustment coefficient is greater than or equal to the red action threshold, and the countdown monitoring is zero in the short-time overload monitoring and warning state, it is determined that the current state is located in the danger interval, and the control execution end executes a breaking and current limiting instruction to cut off and limit the current of the branch circuit;

[0050] Real-time acquisition of control response feedback data of the execution end, packaging and sending the control response feedback data, the current optimal estimation value, the temperature optimal estimation value, the load prediction value, the remaining heat tolerance time and the load control adjustment coefficient to the monitoring terminal for storage and recording.

[0051] Preferably, a visual load state atlas is further generated according to the load control adjustment coefficient:

[0052] A load state monitoring coordinate system of the visual load state atlas is constructed on a display interface of the monitoring terminal, and the horizontal axis is set as the time dimension and the vertical axis is set as the load control adjustment coefficient dimension;

[0053] A first boundary line corresponding to the yellow warning threshold and a second boundary line corresponding to the red action threshold are drawn in the load state monitoring coordinate system, and the load state monitoring coordinate system is divided into a normal operation region, a short-time overload warning region and a dangerous action region by the first boundary line and the second boundary line;

[0054] The historical sequence and the current value of the load control adjustment coefficient are mapped into the load state monitoring coordinate system, and a continuous load state change trajectory curve is generated by spline interpolation;

[0055] Each data point on the load state change trajectory curve is determined to be located in a region, and the load state change trajectory curve is configured with a corresponding color display attribute according to the region;

[0056] When the load control adjustment coefficient is in the short-time overload warning region, a countdown dynamic icon is generated in the load state monitoring coordinate system, and a real-time value of the countdown monitoring is mapped to the countdown dynamic icon for superimposed display.

[0057] The intelligent cable branch box adaptive load control system is used to implement the above method, comprising:

[0058] A perception verification module collects and pre-processes current data of the branch circuit and temperature data at the cable joint in real time, performs electro-thermal coupling credibility verification on the pre-processed current data and temperature data, obtains a data confidence factor, and extracts a current rate of change and a temperature rise rate;

[0059] A trend prediction module inputs the pre-processed current data into a load trend prediction model to obtain a load prediction value in a future set time period;

[0060] A thermal state deduction module combines the load prediction value, uses a transient thermal circuit equivalent model, and solves by a finite difference iteration algorithm according to thermal characteristic parameters and a limit tolerance temperature of the cable insulation material to calculate a residual thermal tolerance time to the limit tolerance temperature;

[0061] An intelligent control decision module inputs the data confidence factor, the current rate of change, the temperature rise rate, the load prediction value, and the residual thermal tolerance time into an adaptive load control decision model using a radial basis function neural network architecture, and trains and optimizes by a particle swarm optimization algorithm to calculate and output a load control adjustment coefficient;

[0062] A hierarchical response module determines a current state interval according to the load control adjustment coefficient, and controls an execution end to execute a hierarchical response strategy of maintaining through-flow, short-time overload monitoring and early warning, and disconnection current limiting in combination with the residual thermal tolerance time.

[0063] The present application has the following advantages:

[0064] The present application introduces an electro-thermal coupling credibility verification mechanism in the data collection stage, verifies the logical consistency of the current data and the temperature data by using a Joule's law physical model, calculates a data confidence factor as a weight parameter for subsequent control decisions, effectively identifies and eliminates data distortion caused by sensor zero drift and damage, avoids false tripping actions of circuit breakers caused by single sensor data anomalies, and significantly improves the monitoring data accuracy and operation reliability of the intelligent cable branch box in harsh operating environments.

[0065] The present application uses a transient thermal circuit equivalent model in combination with a load prediction value and thermal characteristic parameters of the cable insulation material to deduce the temperature rise trajectory of the cable insulation layer and calculate the residual thermal tolerance time, breaks through the limitation of the traditional fixed current threshold protection mode on the short-time overload capacity of the cable, allows the branch line to operate in short-time overload under the premise of ensuring that the temperature of the cable insulation layer does not exceed the limit, fully taps the dynamic transmission potential of the cable line, and maximizes the continuity of power supply while ensuring the safety of equipment.

[0066] The radial basis function neural network optimized by the particle swarm optimization algorithm is adopted to construct an adaptive load control decision model, multi-dimensional feature vectors of data confidence, load change trend and residual heat tolerance time are comprehensively processed, a nonlinear mapping relationship between input features and load control adjustment coefficients is established, a change from passive post-trip to active hierarchical response strategy is realized, current operation state can be accurately determined according to the load control adjustment coefficient and the residual heat tolerance time, and maintaining through-flow, short-time overload monitoring and early warning and disconnection current limiting operation are executed, and the fine and intelligent load management of the power distribution network is realized. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 A flowchart of the method of the present application is shown in the figure.

[0068] Figure 2 A simulation test graph in the verification process of the present application is shown in the figure.

[0069] Figure 3 A structure block diagram of the system of the present application is shown in the figure. DETAILED DESCRIPTION

[0070] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the accompanying drawings.

[0071] Embodiment 1: As shown in the figure, the adaptive load control method of the intelligent cable branch box comprises the following steps: Figure 1

[0072] S1, real-time acquisition and preprocessing of current data of the branch circuit and temperature data at the cable joint, performing electric-thermal coupling confidence verification on the preprocessed current data and temperature data, obtaining a data confidence factor, and extracting a current change rate and a temperature rise rate.

[0073] A Hall current sensor is arranged at the conductor of the branch circuit to sense the magnetic field generated when the branch circuit conductor is powered on, output a Hall voltage signal corresponding to the magnetic induction intensity, and obtain the current data at each discrete time according to the proportional relationship between the Hall voltage signal and the current.

[0074] A fiber Bragg grating temperature sensor is arranged on the surface of the cable joint insulating layer to receive a broadband light signal and reflect a narrowband light of a center wavelength, wherein the center wavelength drifts with the change of the grating period and the effective refractive index caused by the temperature change of the cable joint, the drift amount of the center wavelength is detected, and the temperature data at each discrete time is calculated according to the drift amount.

[0075] ​A discrete linear dynamic system model including a system state equation and an observation equation is constructed, a process noise covariance matrix and a measurement noise covariance matrix are set, and based on the discrete linear dynamic system model, the process noise covariance matrix and the measurement noise covariance matrix, the collected current data and temperature data are iteratively calculated by using a Kalman filtering algorithm for time updating and measurement updating, and the current optimal estimation value and the temperature optimal estimation value after filtering out noise interference are output as the pretreated current data and temperature data respectively.

[0076] The equivalent resistance parameter of the branch circuit, the joint heat dissipation coefficient and the environmental temperature parameter are acquired, a physical benchmark model of electro-thermal coupling is constructed based on Joule's law, the current optimal estimation value is input into the physical benchmark model of electro-thermal coupling, and the theoretical temperature rise value at the current time is calculated .

[0077] According to the Joule's law formula, the heat generated by the current passing through the cable is calculated. Then, combined with the thermal characteristic parameters of the cable (such as equivalent thermal resistance and equivalent thermal capacity) and the environmental temperature, a dynamic model of cable temperature change is established. By combining the heat generated by the current with the heat dissipation process of the cable (including convection and radiation heat dissipation), the differential equation of cable temperature change with time is derived. Finally, the temperature change of the cable under different current conditions is calculated using the equation, and a physical benchmark model of electro-thermal coupling is constructed for evaluating the thermal state of the cable in actual operation.

[0078] The absolute error between the temperature optimal estimation value and the theoretical temperature rise value is calculated, and the absolute error is taken as a residual index.

[0079] Based on the residual index, a data confidence factor is calculated using a Gaussian membership function , wherein, represents a preset residual standard deviation threshold, represents an exponential function with a natural constant as the base.

[0080] The current optimal estimation value and the temperature optimal estimation value at the current sampling time are acquired, as well as the current optimal estimation value and the temperature optimal estimation value at the last sampling time.

[0081] The current difference value of the current optimal estimation value between the current sampling time and the last sampling time is calculated, and the current difference value is divided by the sampling time interval to obtain the current change rate.

[0082] The temperature difference value of the temperature optimal estimation value between the current sampling time and the last sampling time is calculated, and the temperature difference value is divided by the sampling time interval to obtain the temperature rise rate.

[0083] S2, input the pretreated current data into the load trend prediction model, and acquire the load prediction value in the future set time period.

[0084] The historical time window length is set, and a sliding window mechanism is used to extract the optimal historical current estimate and construct the input time series vector.

[0085] A load trend prediction model is constructed, which includes an input layer, a long short-term memory network layer, and an output layer. The input time series vector is input into the long short-term memory network layer. The long short-term memory network layer includes a preset number of long short-term memory network units. Each long short-term memory network unit includes a forget gate, an input gate, an output gate, and a cell state.

[0086] The forget gate is used to receive the input data at the current time step and the hidden layer state at the previous time step. The forget gate coefficient (the control coefficient of the forget gate) is calculated through the activation function to determine the degree of retention of the cell state at the previous time step.

[0087] The input gate is used to receive the input data at the current time step and the hidden layer state at the previous time step. The input gating coefficient and the candidate cell state are calculated to determine the degree of update of the new information at the current time step.

[0088] The current cell state is calculated by multiplying the cell state of the previous time step by the forgetting gating coefficient, and then adding the product of the candidate cell state and the input gating coefficient.

[0089] The input data at the current time step and the hidden layer state at the previous time step are received by the output gate. The output gating coefficient is calculated, and the cell state at the current time step is processed by the activation function and multiplied by the output gating coefficient to obtain the hidden layer state at the current time step.

[0090] The current hidden layer state is input into the output layer, and a linear transformation is performed through the weight matrix and bias term of the fully connected layer to calculate and output the current prediction sequence for a future set time period, which is used as the load prediction value.

[0091] S3. Combining the load prediction value, using the transient thermal circuit equivalent model, and based on the thermal characteristic parameters and ultimate tolerance temperature of the cable insulation material, the remaining thermal tolerance time to reach the ultimate tolerance temperature is calculated using the finite difference iterative algorithm.

[0092] Thermal characteristic parameters include equivalent thermal resistance parameters. and equivalent heat capacity parameters The equivalent thermal resistance parameter is calculated by combining the thermal conductivity of the cable insulation material with the geometric dimensions of the cable insulation layer. The equivalent heat capacity parameter is calculated by combining the specific heat capacity parameter of the cable insulation material with the density and volume of the cable insulation layer. Construct a system that includes equivalent thermal resistance parameters. and equivalent heat capacity parameters The transient thermal circuit equivalent model.

[0093] The temperature optimal estimation value at the current time is taken as the initial temperature state of the iterative calculation.

[0094] Based on the current prediction sequence and the transient thermal circuit equivalent model in the future setting time period, the finite difference iterative algorithm is used to deduce the cable temperature value at each discrete time in the future in time steps, wherein the calculation formula of the cable temperature value at the k+1 time step is as follows:

[0095]

[0096] In the formula, T(k+1) represents the cable temperature value at the k+1 time step, represents the cable temperature value at the k time step, represents the time step, represents the current prediction value corresponding to the k time step in the current prediction sequence, represents the conductor resistance of the branch circuit, represents the ambient temperature parameter.

[0097] The calculated cable temperature value at each discrete time in the future is compared with the limit temperature of the cable insulation material one by one, and when it is determined that the cable temperature value at a discrete time first exceeds the limit temperature, the time difference between the discrete time and the current time is calculated, and the time difference is taken as the residual thermal tolerance time.

[0098] S4, input the data confidence factor, the current change rate, the temperature rise rate, the load prediction value and the residual thermal tolerance time into the adaptive load control decision model adopting the radial basis function neural network architecture, and train and optimize it using the particle swarm optimization algorithm to calculate and output the load control adjustment coefficient.

[0099] The data confidence factor, the current change rate, the temperature rise rate, the load prediction value and the residual thermal tolerance time are normalized and combined to form an input feature vector X.

[0100] An adaptive load control decision model is constructed, which includes a decision input layer, a decision hidden layer and a decision output layer, wherein the decision hidden layer adopts Gaussian radial basis function as the activation function, obtains a historical running data set containing the input feature vector X of the historical time and the corresponding ideal load control adjustment coefficient as the model training sample.

[0101] The particle swarm optimization algorithm is used to optimize the network parameters of the adaptive load control decision model, and the optimized adaptive load control decision model is obtained, wherein the network parameters include the basis function center vector of the decision hidden layer, the basis function width vector of the decision hidden layer and the output weight from the decision hidden layer to the decision output layer.

[0102] ​​Initialize the particle swarm, set the particle swarm size, learning factor, inertia weight and maximum iteration number, and encode the network parameters into the position vector of the particle.

[0103] Perform particle fitness evaluation, input the historical operation data set into the adaptive load control decision model, calculate the model output value using the network parameters corresponding to the current particle position vector, calculate the mean square error between the model output value and the ideal load control adjustment coefficient, and take the reciprocal of the mean square error as the fitness function value.

[0104] Update the optimization extreme value, compare the fitness function value of the current particle with the individual historical optimal fitness function value and the global historical optimal fitness function value of the population, respectively, if the fitness function value of the current particle is larger, then use the position of the current particle to replace the corresponding individual historical optimal position and global historical optimal position.

[0105] Perform iterative optimization, adjust the flight speed and spatial position of the particle according to the individual historical optimal position and the global historical optimal position, and loop the above particle fitness evaluation and optimization extreme value updating process until one of the maximum iteration number and the fitness function value meets the convergence condition, and decode the final global historical optimal position into the optimized basis function center vector, basis function width vector and output weight.

[0106] The input feature vector of the current time is constructed Input the optimized adaptive load control decision model, calculate the output of the Gaussian radial basis function through the decision hidden layer, and perform linear weighted summation through the decision output layer to calculate the load control adjustment coefficient where m represents the number of nodes of the decision hidden layer neurons, represents the output weight of the jth decision hidden layer neuron to the decision output layer, represents the basis function center vector of the jth decision hidden layer neuron, represents the Euclidean norm, represents the basis function width vector of the jth decision hidden layer neuron.

[0107] S5, determine the current state interval according to the load control adjustment coefficient, and control the execution end to execute the hierarchical response strategy of maintaining through-flow, short-time overload monitoring and early warning, and breaking limit current.

[0108] Based on the calculated load control adjustment coefficient and the remaining heat tolerance time, set the preset yellow warning threshold and the preset red action threshold, and compare the load control adjustment coefficient with the yellow warning threshold and the red action threshold, respectively.

[0109] When the load control adjustment coefficient is less than or equal to the yellow early warning threshold, it is determined that the current state is located in the normal interval, and the control execution end maintains the current time flow state.

[0110] When the load control adjustment coefficient is greater than the yellow early warning threshold and less than the red action threshold, it is determined that the current state is located in the short-time overload interval, an early warning signal is generated, the control execution end maintains the current time flow state and enters the short-time overload monitoring and early warning state, and the remaining heat tolerance time is used as an initial time length to start countdown monitoring.

[0111] When the load control adjustment coefficient is greater than or equal to the red action threshold, and the countdown monitoring is zero in the short-time overload monitoring and early warning state, it is determined that the current state is located in the dangerous interval, and the control execution end executes a breaking and current limiting instruction to cut off and limit the current of the branch circuit.

[0112] Real-time control response feedback data of the execution end is collected, and the control response feedback data, the current optimal estimation value, the temperature optimal estimation value, the load prediction value, the remaining heat tolerance time and the load control adjustment coefficient are packaged and sent to the monitoring terminal for storage and recording.

[0113] In the embodiment, the method further includes: S6, generating a visual load state atlas according to the load control adjustment coefficient.

[0114] A load state monitoring coordinate system of the visual load state atlas is constructed on a display interface of the monitoring terminal, and the horizontal axis is set as the time dimension and the vertical axis is set as the load control adjustment coefficient dimension.

[0115] A first boundary line corresponding to the yellow early warning threshold and a second boundary line corresponding to the red action threshold are drawn in the load state monitoring coordinate system, and the load state monitoring coordinate system is divided into a normal operation region, a short-time overload warning region and a dangerous action region by the first boundary line and the second boundary line.

[0116] The historical sequence and the current value of the load control adjustment coefficient are mapped into the load state monitoring coordinate system, and a continuous load state change trajectory curve is generated by spline interpolation.

[0117] Each data point on the load state change trajectory curve is determined to be located in a region, and the load state change trajectory curve is configured with a corresponding color display attribute according to the region, wherein a curve segment located in the normal operation region is configured with a first color, a curve segment located in the short-time overload warning region is configured with a second color, and a curve segment located in the dangerous action region is configured with a third color.

[0118] When the load control adjustment coefficient is in the short-time overload warning region, a countdown dynamic icon is generated in the load state monitoring coordinate system, and a real-time value of the countdown monitoring is mapped to the countdown dynamic icon for superimposed display.

[0119] like Figure 2 As shown, through simulation experiments, the ability of the method in this embodiment to perform adaptive graded response and tap the dynamic transmission potential of the cable under dynamic time-varying load and sudden overload conditions is verified.

[0120] With the resistance of the wires in the branch circuit Equivalent thermal resistance parameters and equivalent heat capacity parameters The physical model parameters are used, with current data containing sinusoidal fluctuations and step disturbances for 600 seconds as input test excitation. The simulation test diagram is shown below. Figure 2 As shown.

[0121] Figure 2 In the diagram, the horizontal axis represents the simulation time t (in seconds), and the vertical axis represents the load control adjustment coefficient. Figure 2 The lower dashed line is the first boundary line, and the upper dashed line is the second boundary line. The first boundary line corresponds to the yellow warning threshold (0.6), and the second boundary line corresponds to the red action threshold (0.85). Using these two boundary lines, the load status monitoring coordinate system is divided into the normal operation area at the bottom, the short-term overload warning area in the middle, and the dangerous action area at the top. The area below the first boundary line is the normal operation area, the area between the first and second boundary lines is the short-term overload warning area, and the area above the second boundary line is the dangerous action area.

[0122] Figure 2 The 0s-300s interval shows the adaptive monitoring phase of the method in this embodiment under dynamic load. During this period, the input current data exhibits periodic fluctuations, and the load control adjustment coefficient adaptively switches between approximately 0.4 (normal operation range) and approximately 0.7 (short-term overload warning range). When the load control adjustment coefficient jumps to 0.7 and enters the short-term overload warning range, although it exceeds the first boundary line, the remaining thermal withstand time calculated by combining the transient thermal circuit equivalent model indicates that the cable insulation temperature has not yet reached the limit withstand temperature. Therefore, only short-term overload monitoring and warning and countdown monitoring are executed, and the disconnection and current limiting command is not executed. This process fully verifies that the method in this embodiment can utilize the thermal inertia of the cable to tap its dynamic transmission potential and avoid the frequent malfunctions of traditional fixed threshold protection during load fluctuations.

[0123] Figure 2The active defense stage of the method of the embodiment in the 300s-600s interval is shown. At 300s, a sudden strong load disturbance causes the load control adjustment coefficient to quickly rise to about 0.9, break through the second boundary line and continuously be in the dangerous action region, it is determined that the cable insulation faces the risk of irreversible thermal damage, and the breaking current limiting instruction is triggered immediately. Subsequently, at about 450s, the current of the branch circuit is forced to be limited, the load control adjustment coefficient instantaneously drops and returns to the normal operation region of about 0.4, and this process verifies that the method of the embodiment can realize accurate state identification and decisive hierarchical response when facing real overload faults, and effectively guarantees the safety of equipment.

[0124] It can be seen through the verification process that the method of the embodiment effectively solves the problems of misoperation caused by distorted sensor data and insufficient flexibility of traditional fixed threshold protection by combining the electrical-thermal coupling credibility verification and the transient thermal circuit equivalent model. The adaptive load control decision model optimized by the particle swarm optimization algorithm ensures the maximum transmission efficiency of the distribution line within the safety boundary.

[0125] In summary, the method of the embodiment exhibits accurate state identification and reliable active defense capability in a complex distribution network load fluctuation environment. At the same time, through multi-dimensional feature fusion and hierarchical response mechanism, an intelligent management and control system for the end of distribution is constructed, which takes into account power supply reliability and equipment safety, has significant technical advantages and application value.

[0126] Embodiment 2: As shown in Figure 3 The adaptive load control system of the intelligent cable branch box is used to implement the method in embodiment 1, and includes a perception verification module, a trend prediction module, a thermal state deduction module, a smart control decision module, and a hierarchical response module. The modules are connected by electrical signals.

[0127] The perception verification module collects and pre-processes the current data of the branch circuit and the temperature data at the cable joint in real time, performs electrical-thermal coupling credibility verification on the pre-processed current data and temperature data, obtains data confidence factors, and extracts the current change rate and temperature rise rate.

[0128] The trend prediction module inputs the pre-processed current data into a load trend prediction model to obtain a load prediction value in a future set time period.

[0129] The thermal state deduction module combines the load prediction value, uses a transient thermal circuit equivalent model, and solves by a finite difference iteration algorithm according to the thermal characteristic parameters and the limit tolerance temperature of the cable insulation material to calculate the remaining thermal tolerance time to reach the limit tolerance temperature.

[0130] The intelligent control decision module inputs the data confidence factor, the current rate of change, the temperature rise rate, the load prediction value and the residual thermal tolerance time into an adaptive load control decision model adopting a radial basis function neural network architecture, and adopts a particle swarm optimization algorithm for training and optimization to calculate and output a load control adjustment coefficient.

[0131] The hierarchical response module determines the current state interval according to the load control adjustment coefficient, and controls the execution end to execute the hierarchical response strategies of maintaining through-flow, short-time overload monitoring and early warning, and disconnection current limiting in combination with the residual thermal tolerance time.

Claims

1. An adaptive load control method for intelligent cable branch boxes, characterized in that, Includes the following steps: Real-time acquisition and preprocessing of current data in branch circuits and temperature data at cable joints; electrothermal coupling reliability verification of preprocessed current and temperature data; acquisition of data confidence factor; and extraction of current change rate and temperature rise rate. The preprocessed current data is input into the load trend prediction model to obtain the load prediction value for a future set time period. Combining load forecast values, using the transient thermal circuit equivalent model, and based on the thermal characteristic parameters and ultimate withstand temperature of the cable insulation material, the remaining thermal withstand time to reach the ultimate withstand temperature is calculated using the finite difference iterative algorithm. An adaptive load control decision model is constructed using a radial basis function neural network architecture, taking data confidence factor, current change rate, temperature rise rate, load forecast, and remaining heat tolerance time as inputs. This model is then trained and optimized using a particle swarm optimization algorithm to calculate the output load control regulation coefficient. The process includes: The data confidence factor, current change rate, temperature rise rate, load prediction value, and remaining heat tolerance time are normalized and combined to construct the input feature vector. An adaptive load control decision model is constructed, which includes a decision input layer, a decision hidden layer, and a decision output layer. The decision hidden layer uses the Gaussian radial basis function as the activation function and obtains a historical operation dataset containing the input feature vectors of historical moments and the corresponding ideal load control adjustment coefficients as model training samples. The particle swarm optimization algorithm is used to optimize the network parameters of the adaptive load control decision model, resulting in an optimized adaptive load control decision model. The network parameters include the basis function center vector of the decision hidden layer, the basis function width vector of the decision hidden layer, and the output weights from the decision hidden layer to the decision output layer. Initialize the particle swarm, set the particle swarm size, learning factor, inertia weight and maximum number of iterations, and encode the network parameters as particle position vectors; Perform particle fitness evaluation, input historical operation dataset into adaptive load control decision model, calculate model output value using network parameters corresponding to the current particle position vector, calculate mean square error between model output value and ideal load control adjustment coefficient, and use the reciprocal of mean square error as fitness function value; Update the optimal extreme value by comparing the fitness function value of the current particle with the individual historical best fitness function value of the current particle and the global historical best fitness function value of the population. If the fitness function value of the current particle is larger, then use the position of the current particle to replace and update the corresponding individual historical best position and the global historical best position respectively. Perform iterative optimization, adjust the particle's flight speed and spatial position based on the individual historical best position and the global historical best position, and repeatedly execute the above particle fitness evaluation and update optimization extreme value process until the maximum number of iterations or the fitness function value satisfies one of the convergence conditions. The final global historical best position is decoded into the optimized basis function center vector, basis function width vector and output weight. The input feature vector at the current moment is input into the optimized adaptive load control decision model. The output of the Gaussian radial basis function is calculated through the decision hidden layer, and the load control adjustment coefficient is calculated by linear weighted summation through the decision output layer. The current state range is determined based on the load control adjustment coefficient, and the remaining heat tolerance time is combined with the control actuator to implement a graded response strategy of maintaining flow, short-term overload monitoring and early warning, and disconnection and current limiting.

2. The intelligent cable branch box adaptive load control method according to claim 1, characterized in that, The process of real-time acquisition and preprocessing of current data from branch circuits and temperature data at cable joints includes: A Hall current sensor is placed at the wire of the branch circuit to sense the magnetic field generated when the wire of the branch circuit is energized, outputs a Hall voltage signal corresponding to the magnetic induction intensity, and obtains the current data at each discrete moment according to the ratio of the Hall voltage signal to the current. A fiber Bragg grating temperature sensor is arranged on the surface of the cable joint insulation layer to receive broadband optical signals and reflect narrowband light with a center wavelength. The center wavelength drifts due to changes in the grating period and effective refractive index caused by changes in the cable joint temperature. The amount of center wavelength drift is detected, and the temperature data at each discrete moment is calculated based on the amount of drift. A discrete linear dynamic system model containing system state equations and observation equations is constructed. The process noise covariance matrix and the measurement noise covariance matrix are defined. The Kalman filter algorithm is used to perform iterative calculations on the collected current data and temperature data for time update and measurement update, respectively. The optimal estimates of current and temperature after filtering out noise interference are output as preprocessed current data and temperature data, respectively.

3. The intelligent cable branch box adaptive load control method according to claim 2, characterized in that, The process of performing electrothermal coupling confidence verification on the preprocessed current and temperature data and obtaining the data confidence factor includes: Obtain the equivalent resistance parameters, heat dissipation coefficient of the connector, and ambient temperature parameters of the branch circuit. Construct an electrothermal coupling physical benchmark model based on Joule's law. Input the optimal current estimate into the electrothermal coupling physical benchmark model to calculate the theoretical temperature rise at the current moment. Calculate the absolute error between the optimal temperature estimate and the theoretical temperature rise, and use the absolute error as the residual index; Based on the residual index, the confidence factor of the data is calculated using the Gaussian membership function.

4. The intelligent cable branch box adaptive load control method according to claim 2, characterized in that, The process of extracting the rate of change of current and the rate of temperature rise includes: Obtain the optimal current and optimal temperature estimates at the current sampling time, as well as the optimal current and optimal temperature estimates at the previous sampling time; Calculate the current difference between the current at the current sampling time and the optimal current estimate at the previous sampling time, and divide the current difference by the sampling time interval to obtain the current change rate; Calculate the temperature difference between the current sampling time and the optimal temperature estimate of the previous sampling time, and divide the temperature difference by the sampling time interval to obtain the temperature rise rate.

5. The intelligent cable branch box adaptive load control method according to claim 2, characterized in that, The process of inputting preprocessed current data into a load trend prediction model to obtain load forecast values ​​for a future set time period includes: Set the length of the historical time window, use a sliding window mechanism to extract the best historical current estimate, and construct the input time series vector; A load trend prediction model is constructed, which includes an input layer, a long short-term memory network layer, and an output layer. The input time series vector is input into the long short-term memory network layer. The long short-term memory network layer includes a preset number of long short-term memory network units. Each long short-term memory network unit includes a forget gate, an input gate, an output gate, and a cell state. The forget gate is used to receive the input data at the current time step and the hidden layer state at the previous time step. The forget gate coefficient is calculated through the activation function to determine the degree of retention of the cell state at the previous time step. The input gate is used to receive the input data at the current time step and the hidden layer state at the previous time step. The input gating coefficient and the candidate cell state are calculated to determine the degree of update of the new information at the current time step. The current cell state is calculated by multiplying the cell state of the previous time step by the forgetting gating coefficient, and then adding the product of the candidate cell state and the input gating coefficient. The current input data and the hidden layer state of the previous time step are received by the output gate. The output gating coefficient is calculated, and the current cell state is processed by the activation function and multiplied by the output gating coefficient to obtain the current hidden layer state. The current hidden layer state is input into the output layer, and a linear transformation is performed through the weight matrix and bias term of the fully connected layer to calculate and output the current prediction sequence for a future set time period, which is used as the load prediction value.

6. The intelligent cable branch box adaptive load control method according to claim 5, characterized in that, The process of calculating the remaining thermal tolerance time to reach the ultimate tolerance temperature using the finite difference iterative algorithm includes: Thermal characteristic parameters include equivalent thermal resistance parameters and equivalent thermal capacity parameters. The equivalent thermal resistance parameter is calculated by combining the thermal conductivity of the cable insulation material with the geometric dimensions of the cable insulation layer. The equivalent thermal capacity parameter is calculated by combining the specific heat capacity parameter of the cable insulation material with the density and volume of the cable insulation layer. A transient thermal circuit equivalent model containing equivalent thermal resistance parameters and equivalent thermal capacity parameters is constructed. The optimal temperature estimate at the current moment is used as the initial temperature state for iterative calculation; Based on the current prediction sequence and transient thermal circuit equivalent model within a future time period, the finite difference iterative algorithm is used to deduce the cable temperature value at each discrete moment in the future point by point according to the time step. The calculated cable temperature values ​​at each discrete future moment are compared one by one with the ultimate tolerance temperature of the cable insulation material. When it is determined that the cable temperature value at a discrete moment exceeds the ultimate tolerance temperature for the first time, the time difference between the discrete moment and the current moment is calculated, and the time difference is used as the remaining thermal tolerance time.

7. The intelligent cable branch box adaptive load control method according to claim 2, characterized in that, The process by which the control execution end implements a graded response strategy of maintaining current flow, short-term overload monitoring and early warning, and current limiting includes: Based on the calculated load control adjustment coefficient and remaining heat tolerance time, a preset yellow warning threshold and a preset red action threshold are set, and the load control adjustment coefficient is numerically compared with the yellow warning threshold and the red action threshold respectively. When the load control adjustment coefficient is less than or equal to the yellow warning threshold, the current state is determined to be in the normal range, and the control execution end maintains the current flow state. When the load control adjustment coefficient is greater than the yellow warning threshold and less than the red action threshold, it is determined that the current state is in the short-term overload range, a warning signal is generated, the control execution end maintains the current flow state and enters the short-term overload monitoring warning state, and uses the remaining heat tolerance time as the initial duration to start countdown monitoring. When the load control adjustment coefficient is greater than or equal to the red action threshold, and when the countdown monitoring returns to zero in the short-term overload monitoring and early warning state, it is determined that the current state is in the danger zone, and the control execution terminal executes the disconnection and current limiting command to cut off and limit the current of the branch circuit accordingly. The control response feedback data from the execution end is collected in real time. The control response feedback data, along with the optimal current estimate, optimal temperature estimate, load forecast, remaining heat tolerance time, and load control adjustment coefficient, are packaged and sent to the monitoring terminal for storage and recording.

8. The intelligent cable branch box adaptive load control method according to claim 7, characterized in that, It also includes generating a visual load state map based on the load control adjustment coefficient: On the display interface of the monitoring terminal, a load status monitoring coordinate system for visualizing the load status map is constructed, with the horizontal axis set as the time dimension and the vertical axis as the load control adjustment coefficient dimension. In the load status monitoring coordinate system, draw the first boundary line corresponding to the yellow warning threshold and the second boundary line corresponding to the red action threshold. Use the first boundary line and the second boundary line to divide the load status monitoring coordinate system into the normal operation area, the short-term overload warning area and the dangerous action area. The historical sequence and current value of the load control adjustment coefficient are mapped to the load status monitoring coordinate system, and a continuous load status change trajectory curve is generated by spline interpolation. Determine the region where each data point on the load state change trajectory curve is located, and configure the corresponding color display attribute for the load state change trajectory curve according to the region it is located in; When the load control adjustment coefficient is in the short-term overload warning area, a countdown dynamic icon is generated in the load status monitoring coordinate system, and the real-time value of the countdown monitoring is mapped to the countdown dynamic icon for superimposed display.

9. An intelligent cable branch box adaptive load control system, used to implement the intelligent cable branch box adaptive load control method according to any one of claims 1-8, characterized in that, include: The sensing and verification module collects and preprocesses the current data of the branch circuit and the temperature data at the cable joint in real time. It performs electrothermal coupling reliability verification on the preprocessed current and temperature data, obtains the data confidence factor, and extracts the current change rate and temperature rise rate. The trend prediction module inputs the preprocessed current data into the load trend prediction model to obtain the load prediction value for a future set time period. The thermal simulation module combines load prediction values ​​and uses a transient thermal circuit equivalent model. Based on the thermal characteristic parameters and ultimate withstand temperature of the cable insulation material, it solves the remaining thermal withstand time to reach the ultimate withstand temperature using a finite difference iterative algorithm. The intelligent control decision module uses an adaptive load control decision model with a radial basis function neural network architecture to input data confidence factor, current change rate, temperature rise rate, load forecast value and remaining heat tolerance time, and uses particle swarm optimization algorithm for training and optimization to calculate the output load control regulation coefficient. The graded response module determines the current state range based on the load control adjustment coefficient and, in conjunction with the remaining heat tolerance time, controls the actuator to execute a graded response strategy that includes maintaining flow, short-term overload monitoring and early warning, and disconnection and current limiting.

Images