Feeder equipment and method of intelligent monitoring power distribution system based on AI algorithm

Abstract

The invention discloses feeder equipment and method for an intelligent monitoring power distribution system based on an AI algorithm, and the method comprises the steps: 1, collecting the historical state data of the operation of the power distribution system, and transmitting the data to a data server; 2, preprocessing the historical state data, performing feature extraction to obtain model training data, dividing the model training data into a training set and a test set, and processing the training set to obtain training subsets; 3, constructing a monitoring model by using the marked training set, and evaluating the trained monitoring model by using the test set; and 4, inputting the data acquired in real time into the monitoring model, and monitoring the operation state of the power distribution system in real time. According to the method, the optimal wavelet decomposition threshold value is obtained based on the whale algorithm, so that the data purity is higher, the data set purification model is created by using the local abnormal factor algorithm to obtain the training subset, the monitoring model better fitting the actual operation state is created according to the training subset, and the efficiency and accuracy of fault monitoring are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of feeder monitoring of a power distribution system, and in particular to feeder equipment and a method for intelligently monitoring a power distribution system based on an AI algorithm. Background Art

[0002] Feeder equipment is a key equipment for realizing the power distribution system. In order to improve the intelligence of the distribution network, the existing feeder terminals are equipped with automatic monitoring equipment. Through real-time monitoring of the current, voltage, temperature, partial discharge and other parameters of the feeder equipment, abnormal conditions of the equipment, such as overload, overheating, insulation damage, etc., can be discovered in time, so as to take corresponding measures to avoid power outages caused by equipment failures. In the prior art, such as the invention patent with the publication number CN114726087B and the patent name "A distribution network feeder automatic fault processing method, system and electronic equipment", although a magnetically controlled switch is used as a connection control mechanism between multiple sections in the feeder line, and multiple magnetically controlled switches are used to form a multi-level difference, the feeder line fault can be efficiently and reliably processed and responded to under normal and abnormal communication conditions of the wireless communication network, but there is a lack of effective processing means for the feeder equipment monitoring signal data transmitted through the wireless communication network, and the value of the monitoring data cannot be fully utilized, which makes the reliability of the distribution network feeder automatic fault discrimination model lacking.

[0003] To this end, we have designed feeder equipment and methods for intelligent monitoring of power distribution systems based on AI algorithms to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art in that there is a lack of effective processing means for the monitoring signal data of the feeder equipment, the value of the monitoring data cannot be fully utilized, and the reliability of the distribution network feeder automation fault discrimination model is lacking. The feeder equipment and method of the intelligent monitoring distribution system based on AI algorithm are proposed, aiming to fully explore and utilize the collected operation data of the feeder equipment of the distribution system, and improve the efficiency and accuracy of the feeder equipment operation fault monitoring of the distribution system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm, the intelligent monitoring method comprising the following steps:

[0007] Step 1: Collect historical status data of the power distribution system operation through the feeder equipment, and transmit the collected historical status data to the data server of the monitoring center;

[0008] Step 2: Preprocess the historical state data, remove abnormal data and noise interference to obtain preprocessed data, perform feature extraction on the preprocessed data to obtain model training data, divide the model training data into a training set and a test set, and further process the training set to obtain a training subset;

[0009] Step 3: Label the training subset, label the operating status of the power distribution system, input the labeled training set into the algorithm training to build a monitoring model, and use the test set to evaluate the trained monitoring model;

[0010] Step 4: Input the real-time collected distribution system operating status parameter data into the monitoring model to monitor the operating status of the distribution system in real time.

[0011] Preferably, in step one, the collected historical status data is transmitted to the data server of the monitoring center through any one of the wireless communication methods of the LoRa gateway or the NB-IoT gateway, and the historical status data is transmitted using the TCP / IP transmission protocol with redundancy function. When an error occurs in the transmission of the historical status data, the redundant confirmation mechanism and retransmission mechanism of the TCP / IP transmission protocol are used to automatically retransmit the data.

[0012] Preferably, the historical status data is the collected operating parameters H of the power distribution system Ljc , H Ljc =[I i (t),U i (t),P i (t),T i (t)],

[0013] I i (t) is the current data of the distribution system collected at time t, U i (t) is the voltage data of the power distribution system collected at time t, P i (t) is the power data of the power distribution system collected at time t, T j (t) is the temperature data of the power distribution system collected at time t.

[0014] Preferably, in step 2, the historical state data is preprocessed in the following manner to obtain preprocessed data after removing abnormal data and noise interference:

[0015] Calculate the operating parameter H of historical status data Ljc The mean and standard deviation of

[0016]

[0017]

[0018]

[0019] In the above formula, μ is the mean value of the distribution system operating parameters, σ is the standard deviation of the distribution system operating parameters, and H Ljc represents the operating parameters of the historical state data, n is the set of operating parameters of the historical state data; Z represents an indicator for evaluating the degree of deviation of a single operating parameter data value in the historical state data relative to the average value. If the operating parameter data value is less than the absolute value of the indicator value, the operating parameter data value is determined to be abnormal data;

[0020] The historical state data after removing abnormal data is subjected to wavelet decomposition, the threshold of the high-frequency coefficients of the wavelet decomposition is quantized, and one-dimensional wavelet reconstruction is performed according to the Nth layer low-frequency coefficients of the wavelet decomposition and the first to Nth layers high-frequency coefficients to obtain the denoised historical state data.

[0021] Preferably, in the threshold quantization process, the selected threshold directly affects the quality of historical state data preprocessing. Considering the characteristics of the whale algorithm (WOA) such as fast convergence speed, strong global search capability, and easy implementation, the whale algorithm is used to optimize the threshold selection.

[0022] Define the initial position of the whale population, X i =c+ψ(bc)

[0023] In the formula, X i is the position of whale individual i, c and b represent the lower and upper limits of the whale's search threshold space, and ψ is the search control factor;

[0024] The optimal position in the search threshold space is unknown. The current threshold best candidate solution position is defined as the target prey position. After the target prey position is defined, other whales will try to surround the target prey position, which is expressed by the following formula:

[0025]

[0026] Among them, t is the current iteration number, X represents the location of the current solution, and X * represents the position of the current optimal solution, Y and W are conversion parameters, and a represents the total number of iterations;

[0027] The following formula is used to simulate the spiral search behavior of whales:

[0028]

[0029] Among them, W * represents the distance between the i-th whale individual and the optimal individual, d is a constant, and l is A random number between

[0030] To simultaneously simulate the contraction and enclosure mechanism and the spiral update mechanism of whales, it is assumed that the execution probabilities of the contraction and enclosure mechanism and the spiral update mechanism are equal, and their expressions are as follows:

[0031]

[0032] Using the above formula, the optimal threshold of wavelet decomposition in the preprocessing process of historical state data is searched and selected as follows:

[0033] Step 2.1, initialize the number of whales N and the maximum number of iterations a, and initialize the whale population: X i (i = 1, 2,..., N);

[0034] Step 2.2, use a for loop with an initial value of 1 and a termination value of N. During the loop execution, update the transformation parameters Y and W and the random number l, update the position of each whale. If ψ < 0.5, adopt the contraction and enclosure strategy; otherwise, adopt the spiral forward search strategy;

[0035] Step 2.3, check whether any whale exceeds the search space and make modifications, calculate the hunger value of each whale, and find the whale X * ;

[0036] Step 2.4, while t < a, after the loop ends, return the whale X * , which is the optimal threshold of wavelet decomposition.

[0037] Preferably, in step two, the convolutional neural network algorithm is used for feature extraction of the preprocessed data. In the convolutional layer, the weight matrix of the Gaussian kernel is used to process the local area of the preprocessed data to extract time series data. In the activation layer, the transformation function is used to perform non-linear transformation on the time series data; in the pooling layer, the feature dimension of the transformed time series data is reduced. In the normalization layer, the normalization factor is used to standardize the output of the convolutional layer. Finally, through the output layer, the final feature vector is output according to the features obtained from the convolutional layer, activation layer, and pooling layer to obtain the model training data.

[0038] Preferably, in step two, a data set purification model is created using the local outlier factor algorithm, and the data in the training set is scanned and removed for local outliers using the data set purification model to form a training subset. The expression of the data set purification model is as follows:

[0039]

[0040] Among them, M k (p) represents the set composed of the k data points with the closest distance to each data point p in the training set among other data points in the training set; lrd k(p) represents the local reachable density of data point p in the training set; k is the control parameter; lrd k (q) represents the points in the k-neighborhood of the data point p;

[0041] κ represents the local outlier factor. If κ>1, it means that the density of the data point p in the training set is less than the density of its neighborhood points. The data point p is determined to be an outlier and is removed from the training set. If κ≤1, the data point p in the training set is retained to form a training subset.

[0042] in,

[0043] For data points p and q in the training set, the reachable distance from data point p to data point q is defined as:

[0044] reech_dist k (p,q)=max{d k (p),d(p,q)};

[0045] Among them, d k (p) is the distance from data point p to its k nearest neighbors, and d(p,q) represents the direct distance from data point p to data point q.

[0046] Preferably, in step 3, the labeled training subset is input into the algorithm training to construct a monitoring model, and the expression of the monitoring model is:

[0047]

[0048] In the formula, μ represents the mean value of the distribution system operating parameters obtained by collecting historical status data, σ represents the standard deviation of the distribution system operating parameters obtained by collecting historical status data, Optimize the model parameters, H Sjc Indicates the real-time collected distribution system operation status parameter data;

[0049] Use the test set to evaluate the monitoring model, input the divided test set into the monitoring model, verify the generalization ability and accuracy of the monitoring model, evaluate the performance indicators of the monitoring model on the test set, and adjust the model optimization parameters in the monitoring model according to the evaluation results of the test set. The monitoring model is adjusted and optimized to determine the final monitoring model for the intelligent monitoring method of the power distribution system.

[0050] Preferably, in step 4, the real-time collected distribution system operating status parameter data is input into the monitoring model: Real-time monitoring of the operating status of the power distribution system;

[0051] When the real-time collected distribution system operating status parameters are not in the monitoring model If it is within the range, it means that the distribution system is abnormal, a fault alarm is issued, and the pole switch of the feeder equipment is opened and closed.

[0052] A feeder device for an intelligent monitoring power distribution system based on an AI algorithm, applied to an intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm, the feeder device comprising:

[0053] The power management module includes a voltage transformer and a current transformer. It draws power from the voltage transformer to provide power to the MCU main control board, and provides energy storage power and opening and closing control power to the pole switch.

[0054] MCU main control board, including processor and output control module; the processor processes the received signal data to form a control command and then sends it to the execution terminal. The output control module is connected to the processor, and the processor issues an action command to open and close the pole switch;

[0055] The data acquisition module includes a temperature sensor for collecting temperature signals and an A / D conversion module. The A / D conversion module converts the strong electric signals of the current transformer and the voltage transformer into weak electric signals, and inputs the weak electric signals and the temperature signals into the MCU main control board;

[0056] The communication module is installed on the MCU main control board and establishes a communication connection with the data server of the monitoring center through wireless transmission.

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows: in the technical scheme of the present invention, a variety of intelligent algorithms are used to fully mine and utilize the collected operation big data of the feeder equipment of the distribution system. In the data processing process, the threshold of the optimal wavelet decomposition is obtained based on the whale algorithm, so that the data purity is higher. For the processing of the data training set, on the basis of the parent set of the divided training set, a local anomaly factor algorithm is used to create a data set purification model for further purification, and a training subset is obtained. According to the training subset, a monitoring model that is more in line with the actual operating status is created, thereby improving the efficiency and accuracy of the operation fault monitoring of the feeder equipment of the distribution system, thereby realizing effective monitoring and early warning of feeder equipment faults of the distribution system. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 This is a block diagram of the feeder equipment of the intelligent monitoring power distribution system based on the AI ​​algorithm proposed in the present invention;

[0059] Figure 2 The figure is a flow chart of an intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm in the implementation of the present invention. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0061] With the continuous development of power technology, the power industry has made continuous progress, while also putting forward higher requirements for power supply security, including the need for distribution networks to ensure the reliability and high quality of power supply. From a technical point of view, it is necessary to develop a high-performance and complete distribution network automation system. Therefore, it is necessary for us to study distribution network automation technology, especially feeder automation system technology, including its system structure, operation mechanism, control method and other related technologies. Feeder automation is the core content of distribution network technology. It is the most effective and direct way to improve the quality and reliability of power supply to users. Therefore, in the process of developing distribution network automation, we must first develop feeder equipment automation technology, whose automation technology depends on the use of artificial intelligence, and the advantage of artificial intelligence lies in the effective mining and utilization of big data to create artificial intelligence models.

[0062] like Figure 1 As shown, this embodiment proposes a feeder device for intelligently monitoring a power distribution system based on an AI algorithm. For the monitoring function of the power distribution system, the feeder device mainly includes: a power management module, an MCU main control board, a data acquisition module and a communication module.

[0063] Among them, the power management module includes a voltage transformer and a current transformer, and draws power through the voltage transformer to provide power to the MCU main control board, and provides energy storage power and opening and closing control power to the pole switch.

[0064] The MCU main control board includes a processor and an output control module. The processor processes the received signal data to form a control command and then sends it to the execution terminal. The output control module is connected to the processor, and the processor issues an action command to open and close the pole switch. When a fault is detected, the line can be cut off in time, and the power supply can be restored as soon as the fault is eliminated.

[0065] The data acquisition module includes a temperature sensor for collecting temperature signals and an A / D conversion module. The A / D conversion module converts the strong electric signals of the current transformer and the voltage transformer into weak electric signals, and transmits the weak electric signals and the temperature signal to the MCU main control board. The conversion of strong electric signals into weak electric signals is to convert the secondary analog data of the current transformer and the voltage transformer into digital signals.

[0066] The communication module is set on the MCU main control board, and establishes a communication connection with the data server of the monitoring center through wireless transmission technology. The communication module in this embodiment can include any one of the LoRa gateway module or the NB-IoT gateway module, which can reduce the cost and power consumption of the equipment and improve the compatibility of the equipment and the signal data transmission distance; when the MCU main control board transmits data signals with the data server of the monitoring center, the collected historical status data can be transmitted to the data server of the monitoring center through any wireless communication method of the LoRa gateway or the NB-IoT gateway. And when the signal data is transmitted, the TCP / IP transmission protocol with redundancy function is used to transmit the historical status data to the data server of the monitoring center. When an error occurs in the historical status data transmission, the redundant confirmation mechanism and retransmission mechanism of the TCP / IP transmission protocol can be used to automatically retransmit.

[0067] Based on the feeder equipment of the above-mentioned intelligent monitoring distribution system based on AI algorithm, this embodiment also proposes an intelligent monitoring method, which is applied to the intelligent monitoring distribution system based on AI algorithm.

[0068] Reference Figure 2 The intelligent monitoring method of the intelligent monitoring power distribution system based on AI algorithm mainly includes the following steps:

[0069] Step 1: Collect the historical status data of the power distribution system operation through the feeder equipment, and transmit the collected historical status data to the data server of the monitoring center.

[0070] Step 2: Preprocess the historical status data to obtain preprocessed data after removing abnormal data and noise interference, and perform feature extraction on the preprocessed data to obtain model training data. The model training data is divided into a training set and a test set, and the training set is further processed to obtain a training subset.

[0071] Step 3: Label the training subset, label the operating status of the distribution system, input the labeled training set into the algorithm training to build a monitoring model, and use the test set to evaluate the trained monitoring model.

[0072] Step 4: Input the real-time collected distribution system operating status parameter data into the monitoring model to monitor the operating status of the distribution system in real time.

[0073] Specifically, in step one, the collected historical status data is transmitted to the data server of the monitoring center through any one of the wireless communication methods of the LoRa gateway or the NB-IoT gateway, and the historical status data is transmitted using the TCP / IP transmission protocol with redundancy function. When an error occurs in the transmission of the historical status data, the redundant confirmation mechanism and retransmission mechanism of the TCP / IP transmission protocol are used to automatically retransmit it.

[0074] The historical status data referred to in this implementation is the collected operating parameters H of the distribution system. Ljc , H Ljc =[I i (t),U i (t),P i (t),T i (t)],

[0075] I i (t) is the current data of the distribution system collected at time t, U i (t) is the voltage data of the power distribution system collected at time t, P i (t) is the power data of the power distribution system collected at time t, T j (t) is the temperature data of the power distribution system collected at time t.

[0076] The historical status data is preprocessed in the following way to obtain preprocessed data after removing abnormal data and noise interference:

[0077] Calculate the operating parameter H of historical status data Ljc The mean and standard deviation of

[0078]

[0079]

[0080]

[0081] In the above formula, μ is the mean value of the distribution system operating parameters, σ is the standard deviation of the distribution system operating parameters, and H Ljc represents the operating parameters of the historical status data, n is the operating parameter set of the historical status data; Z represents an indicator for evaluating the degree of deviation of a single operating parameter data value in the historical status data relative to the average value. If the operating parameter data value is less than the absolute value of the indicator value, then this operating parameter data value is determined to be abnormal data.

[0082] The historical state data after removing abnormal data is subjected to wavelet decomposition, the threshold of the high-frequency coefficients of the wavelet decomposition is quantized, and one-dimensional wavelet reconstruction is performed according to the Nth layer low-frequency coefficients of the wavelet decomposition and the first to Nth layers high-frequency coefficients to obtain the denoised historical state data.

[0083] It is worth mentioning that in the threshold quantization process, the selected threshold directly affects the quality of historical state data preprocessing. Considering the characteristics of the whale algorithm (WOA) such as fast convergence speed, strong global search capability and easy implementation, the whale algorithm is used to optimize the threshold selection.

[0084] Define the initial position of the whale population, Xi =c+ψ(bc);

[0085] In the formula, X i is the position of whale individual i, c and b represent the lower and upper limits of the whale’s search threshold space, and ψ is the search control factor.

[0086] The optimal position in the search threshold space is unknown. The current threshold best candidate solution position is defined as the target prey position. After the target prey position is defined, other whales will try to surround the target prey position, which is expressed by the following formula:

[0087]

[0088] Among them, t is the current iteration number, X represents the location of the current solution, and X * represents the position of the current optimal solution, Y and W are conversion parameters, and a represents the total number of iterations.

[0089] The following formula is used to simulate the spiral search behavior of whales:

[0090]

[0091] Among them, W * represents the distance between the i-th whale individual and the optimal individual, d is a constant, and l is A random number between .

[0092] In order to simulate the shrinking and encircling mechanism and spiral updating mechanism of the whale at the same time, it is assumed that the execution probability of the shrinking and encircling mechanism and the spiral updating mechanism are equal, and the expression is as follows:

[0093]

[0094] Using the above formula, the optimal threshold of wavelet decomposition in the preprocessing of historical state data is searched and selected. The process is as follows:

[0095] Step 2.1, initialize the number of whales N and the maximum number of iterations a, and initialize the whale population: X i (i=1,2,...,N);

[0096] Step 2.2, use a for loop with an initial value of 1 and an end value of N. During the loop execution, update the transformation parameters Y and W and the random number l, and update the position of each whale. If ψ<0.5, use the shrinking and encircling strategy, otherwise use the spiral search strategy;

[0097] Step 2.3, check if any whales are out of the search space and make modifications, calculate the hunger value of each whale, and find the most well-fed whale X * ;

[0098] Step 2.4, while t < a, after the loop ends, return the fullest whale X * , which is the optimal threshold of wavelet decomposition.

[0099] In Step 2, the convolutional neural network algorithm is used to extract features from the preprocessed data. In the convolutional layer, the weight matrix of the Gaussian kernel is used to process the local area of the preprocessed data to extract time series data. In the activation layer, a transformation function is used to perform a non-linear transformation on the time series data; in the pooling layer, the feature dimension of the transformed time series data is reduced, and in the normalization layer, a normalization factor is used to standardize the output of the convolutional layer. Finally, through the output layer, the final feature vector is output according to the features obtained from the convolutional layer, activation layer, and pooling layer to obtain the model training data.

[0100] In addition, in Step 2, this embodiment uses the local outlier factor algorithm to create a data set purification model, and uses the data set purification model to perform local outlier scanning and removal on the data in the training set to form a training subset. The expression of the data set purification model is as follows:

[0101]

[0102] Among them, M k (p) represents the set composed of the k data points with the closest distance to each data point p in the training set among other data points in the training set; lrd k (p) represents the local reachability density of the data point p in the training set; k is a control parameter; lrd k (q) represents the points within the k-neighborhood of the data point p.

[0103] κ represents the local outlier factor. If κ > 1, it means that the density at the data point p in the training set is less than the density of its neighboring points, and it is determined that the data point p is an outlier, then the data point p is removed from the training set. If κ ≤ 1, then the data point p in the training set is retained to form a training subset.

[0104] In the above formula,

[0105] For the data points p and q in the training set, the reachable distance from the data point p to the data point q is defined as:

[0106] reech_dist k (p,q) = max{d k (p), d(p,q)};

[0107] Among them, d k (p) is the distance from the data point p to its k nearest neighbors, and d(p,q) represents the direct distance between the data point p and the data point q.

[0108] In step 3, the labeled training subset is input into the algorithm training to build a monitoring model. The expression of the monitoring model is:

[0109]

[0110] In the formula, μ represents the mean value of the distribution system operating parameters obtained by collecting historical status data, σ represents the standard deviation of the distribution system operating parameters obtained by collecting historical status data, Optimize the model parameters, H Sjc Indicates the distribution system operating status parameter data collected in real time.

[0111] Use the test set to evaluate the monitoring model, input the divided test set into the monitoring model, verify the generalization ability and accuracy of the monitoring model, evaluate the performance indicators of the monitoring model on the test set, and adjust the model optimization parameters in the monitoring model according to the evaluation results of the test set. The monitoring model is adjusted and optimized to determine the final monitoring model for the intelligent monitoring method of the power distribution system.

[0112] In step 4, the real-time collected distribution system operating status parameter data is input into the monitoring model: Real-time monitoring of the operating status of the power distribution system;

[0113] When the distribution system operating status parameters collected in real time are not in the monitoring model If it is within the range of , it means that the distribution system is abnormal, and a fault alarm is issued to the monitoring center, and the pole switch of the feeder equipment is opened and closed.

[0114] The opening and closing operations of the pole-mounted switch of the feeder equipment include: when a fault occurs in the feeder equipment, it is determined whether the feeder equipment is a local fault based on the previous fault status and the current electrical operation status. If it is a local fault, the feeder equipment will trip and isolate the fault; if it is not a local fault, the feeder equipment will temporarily lock the tripping outlet to avoid the cross-level tripping of this level of switch, so as to facilitate the rapid restoration of power supply.

[0115] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm, characterized in that: The intelligent monitoring method comprises the following steps: Step 1: Collect historical status data of the power distribution system operation through the feeder equipment, and transmit the collected historical status data to the data server of the monitoring center; Step 2: Preprocess the historical state data, remove abnormal data and noise interference to obtain preprocessed data, perform feature extraction on the preprocessed data to obtain model training data, divide the model training data into a training set and a test set, and further process the training set to obtain a training subset; Step 3: Label the training subset, label the operating status of the power distribution system, input the labeled training set into the algorithm training to build a monitoring model, and use the test set to evaluate the trained monitoring model; Step 4: Input the real-time collected distribution system operating status parameter data into the monitoring model to monitor the operating status of the distribution system in real time.

2. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1 is characterized in that: In step 1, the collected historical status data is transmitted to the data server of the monitoring center through any wireless communication method of LoRa gateway or NB-IoT gateway, and the historical status data is transmitted using the TCP / IP transmission protocol with redundancy function. When an error occurs in the transmission of historical status data, the redundant confirmation mechanism and retransmission mechanism of the TCP / IP transmission protocol are used to automatically retransmit it.

3. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1 is characterized in that: The historical status data is the collected operating parameters of the distribution system H Ljc , H Ljc =[I i (t),U i (t),P i (t),T i (t)], I i (t) is the current data of the distribution system collected at time t, U i (t) is the voltage data of the power distribution system collected at time t, P i (t) is the power data of the power distribution system collected at time t, T j (t) is the temperature data of the power distribution system collected at time t.

4. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1, characterized in that: In step 2, the historical status data is preprocessed in the following way: Calculate the operating parameter H of historical status data Ljc The mean and standard deviation of In the above formula, μ is the mean value of the distribution system operating parameters, σ is the standard deviation of the distribution system operating parameters, H Ljc represents the operating parameters of the historical state data, n is the set of operating parameters of the historical state data; Z represents an indicator for evaluating the degree of deviation of a single operating parameter data value in the historical state data relative to the average value. If the operating parameter data value is less than the absolute value of the indicator value, the operating parameter data value is determined to be abnormal data; The historical state data after removing abnormal data is subjected to wavelet decomposition, the threshold of the high-frequency coefficients of the wavelet decomposition is quantized, and one-dimensional wavelet reconstruction is performed according to the Nth layer low-frequency coefficients of the wavelet decomposition and the first to Nth layers high-frequency coefficients to obtain the denoised historical state data.

5. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 4 is characterized in that: In the process of threshold quantization, the selected threshold directly affects the quality of historical state data preprocessing. The whale algorithm is used to optimize the threshold selection. Define the initial position of the whale population, X i =c+ψ(bc); Where, X i is the position of whale individual i, c and b represent the lower and upper limits of the whale's search threshold space, and ψ is the search control factor; The optimal position in the search threshold space is unknown. The current threshold best candidate solution position is defined as the target prey position. After the target prey position is defined, other whales will try to surround the target prey position, which is expressed by the following formula: X(t+1)=X * (t)-Y×W; Y=2×a×ψ-a, Y∈[-a,a]; W=|2×ψ×X * (t)-X(t)|; Among them, t is the current iteration number, X represents the location of the current solution, and X * represents the position of the current optimal solution, Y and W are conversion parameters, and a represents the total number of iterations; The following formula is used to simulate the spiral search behavior of whales: Among them, W * represents the distance between the i-th whale individual and the optimal individual, d is a constant, and l is A random number between In order to simulate the shrinking and encircling mechanism and spiral updating mechanism of the whale at the same time, it is assumed that the execution probability of the shrinking and encircling mechanism and the spiral updating mechanism are equal, and the expression is as follows: Using the above formula, the optimal threshold of wavelet decomposition in the preprocessing of historical state data is searched and selected. The process is as follows: Step 2.1, initialize the number of whales N and the maximum number of iterations a, and initialize the whale population: X i (i=1,2,...,N); Step 2.2, use a for loop with an initial value of 1 and an end value of N. During the loop execution, update the transformation parameters Y and W and the random number l, and update the position of each whale. If ψ<0.5, use the shrinking and encircling strategy, otherwise use the spiral search strategy; Step 2.3, check if any whales are out of the search space and make modifications, calculate the hunger value of each whale, and find the most well-fed whale X * ; Step 2.4, while t < a, after the loop ends, return the fullest whale X * , which is the optimal threshold for wavelet decomposition.

6. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1, characterized in that: In step 2, a convolutional neural network algorithm is used to extract features from the preprocessed data. The weight matrix of the Gaussian kernel is used in the convolution layer to process the local area of ​​the preprocessed data and extract the time series data. The transformation function is used in the activation layer to perform nonlinear transformation on the time series data. The feature dimension of the transformed time series data is reduced in the pooling layer. The output of the convolution layer is standardized using a normalization factor in the normalization layer. Finally, the output layer outputs the final feature vector based on the features obtained from the convolution layer, activation layer, and pooling layer to obtain the model training data.

7. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1, characterized in that: In step 2, a data set purification model is created based on a local anomaly factor algorithm. The data set purification model is used to perform local anomaly scanning and elimination on the data in the training set to form a training subset. The expression of the data set purification model is as follows: Among them, M k (p) represents the set of k data points that are closest to each data point p in the training set. k (p) represents the local reachable density of data point p in the training set; k is the control parameter; lrd k (q) represents the points in the k-neighborhood of the data point p; κ represents the local outlier factor. If κ>1, it means that the density of the data point p in the training set is less than the density of its neighborhood points. The data point p is determined to be an outlier and is removed from the training set. If κ≤1, the data point p in the training set is retained to form a training subset. in, For data points p and q in the training set, the reachable distance from data point p to data point q is defined as: reach_dist k (p,q)=max{d k (p),d(p,q)}; Among them, d k (p) is the distance from data point p to its k nearest neighbors, and d(p,q) represents the direct distance from data point p to data point q.

8. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 1, characterized in that: In step 3, the labeled training subset is input into the algorithm training to build a monitoring model. The expression of the monitoring model is: In the formula, μ represents the mean value of the distribution system operating parameters obtained by collecting historical status data, σ represents the standard deviation of the distribution system operating parameters obtained by collecting historical status data, Optimize the model parameters, H Sjc Indicates the distribution system operating status parameter data collected in real time; Use the test set to evaluate the monitoring model, input the divided test set into the monitoring model, verify the generalization ability and accuracy of the monitoring model, evaluate the performance indicators of the monitoring model on the test set, and adjust the model optimization parameters in the monitoring model according to the evaluation results of the test set. The monitoring model is adjusted and optimized to determine the final monitoring model for the intelligent monitoring method of the power distribution system.

9. The intelligent monitoring method for an intelligent monitoring power distribution system based on an AI algorithm according to claim 8, characterized in that: In step 4, the real-time collected distribution system operating status parameter data is input into the monitoring model: Real-time monitoring of the operating status of the power distribution system; When the real-time collected distribution system operating status parameters are not in the monitoring model If it is within the range, it means that the distribution system is abnormal, a fault alarm is issued, and the pole switch of the feeder equipment is opened and closed.

10. A feeder device of an intelligent monitoring power distribution system based on an AI algorithm, applied to the intelligent monitoring method of an intelligent monitoring power distribution system based on an AI algorithm according to any one of claims 1 to 9, characterized in that: The feeder equipment includes: The power management module includes a voltage transformer and a current transformer. It draws power from the voltage transformer to provide power to the MCU main control board, and provides energy storage power and opening and closing control power to the pole switch. MCU main control board, including processor and output control module; the processor processes the received signal data to form a control command and then sends it to the execution terminal. The output control module is connected to the processor, and the processor issues an action command to open and close the pole switch; The data acquisition module includes a temperature sensor for collecting temperature signals and an A / D conversion module. The A / D conversion module converts the strong electric signals of the current transformer and the voltage transformer into weak electric signals, and inputs the weak electric signals and the temperature signals into the MCU main control board; The communication module is installed on the MCU main control board and establishes a communication connection with the data server of the monitoring center through wireless transmission.

Images