Power distribution network fault research and judgment analysis method and system based on artificial intelligence

Abstract

The invention provides a power distribution network fault research and judgment analysis method and system based on artificial intelligence, aiming at solving the problems that fault line selection of a single criterion is easy to cause line selection misjudgment, single-phase grounding load characteristics cannot be known, and fault analysis and fault positioning work cannot be performed in a targeted manner. Multi-source fault information feature data in a power distribution network of a monitoring area are obtained through an RPA robot, single-phase earth fault multi-source data are analyzed through an artificial intelligence algorithm formed by optimizing and fusing a self-adaptive genetic algorithm and a BP neural network, and according to an analysis result of the fault multi-source data through the artificial intelligence algorithm, the fault multi-source data is analyzed. Various electric power data characteristics existing in the single-phase earth fault are identified, then the single-phase earth fault of the power distribution network is automatically positioned, fault line selection and fault positioning power supply services are provided for the single-phase earth fault based on an analysis result and in combination with the actual condition of the power distribution network, and a scientific operation mode optimization strategy is formulated. And safe and stable operation of a power grid is ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence of power systems, and in particular to a distribution network fault analysis method and system based on artificial intelligence. Background Art

[0002] In my country, in the 10KV-35KV low-voltage distribution network system, single-phase short-circuit grounding faults can account for about 80% to 90% of the total number of system faults. The power system of this type of distribution network generally adopts a small current grounding system. With the complexity of the grid structure and the increase in the number of feeders, the probability of single-phase grounding faults in this type of distribution network is very high. When a single-phase grounding fault occurs, the grounding current is very small and the system can continue to operate for 1-2 hours under fault conditions. However, the fault point must be found as soon as possible. The single-phase grounding fault location and fault line selection problems of the small current grounding system have always been difficulties in engineering practice and on-site operation.

[0003] The fault location and line selection problems of distribution network have not been well solved for a long time. The manual line inspection method is often used on site. Manual line inspection not only consumes a lot of manpower and material resources, but also prolongs the power outage time and affects the power supply safety. At present, the more mature fault line selection methods can be divided into the external signal method and the fault signal method. The fault signal method can be divided into the fault steady-state signal method and the fault transient signal method. However, the above methods only give a single fault criterion for line selection, which often leads to misjudgment of line selection. Therefore, the line selection method based on the fusion of multiple fault criteria for single-phase grounding is an effective way to solve the misjudgment. As for fault location, there are currently three methods for automatic location on site. The first method is to inject high-frequency signals from PT and detect the signals along the line to determine the fault location. However, since the distributed capacitance of the line forms a path for the high-frequency signal, the positioning is inaccurate when grounded by resistance. The second method is to use the fault indicator method, but the accuracy of locating single-phase grounding faults is very low. The third method is to install an intelligent switch with a built-in CT. Although this method can measure the zero-sequence current, the terminal and master station algorithms running on the market are simple and only determine whether the steady-state zero-sequence current exceeds the customization. The positioning accuracy of the neutral point grounding system through the arc suppression coil is very low.

[0004] The Chinese patent publication number is CN111474442B, and the publication date is May 24, 2022. The name of the invention is a method and system for locating a single-phase grounding fault in a distribution network. The scheme discloses a method and system for locating a single-phase grounding fault in a distribution network. Its shortcoming is that it can only perform simple statistical analysis on the single-phase grounding fault signal, and cannot understand the single-phase grounding load characteristics existing in different power data. It cannot carry out targeted fault analysis work in combination with the user's portrait of single-phase grounding fault data, and cannot formulate a scientific, reasonable and personalized single-phase grounding fault handling strategy. Summary of the invention

[0005] The present invention solves the problems that fault line selection based on a single criterion is prone to misjudgment of line selection, cannot understand the characteristics of single-phase grounded load, and cannot perform fault analysis and fault location work in a targeted manner, and provides a distribution network fault assessment and analysis method and system based on artificial intelligence.

[0006] In order to achieve the above-mentioned invention object, this scheme adopts the following technical scheme: A distribution network fault analysis method based on artificial intelligence includes the following steps: S1. Obtaining characteristic data of multi-source fault information of power system; S2. Use artificial intelligence algorithm to analyze multi-source data of single-phase grounding fault; S3. Identify the power data characteristics of the single-phase grounding fault according to the analysis results of the multi-source data of the single-phase grounding fault; S4, automatically locate the single-phase grounding fault in the distribution network. The distribution network fault analysis method also includes step S5, customizing the fault handling strategy in combination with the actual situation of the distribution network. This solution obtains multi-source fault information feature data in the distribution network in the monitoring area through the RPA robot. On this basis, an artificial intelligence algorithm is used to analyze the multi-source data of the single-phase grounding fault. According to the analysis results of the multi-source data of the fault by the artificial intelligence algorithm, the various power data characteristics existing in the single-phase grounding fault are identified, and then the single-phase grounding fault of the distribution network in the monitoring area of ​​the RPA robot is automatically located. According to the analysis results and automatic fault location, combined with the actual situation of the distribution network in the monitoring area, the single-phase grounding fault is provided with fault line selection and fault location power supply services, and a scientific operation mode optimization strategy is formulated to ensure the safe and stable operation of the power grid.

[0007] Preferably, in step S1, multi-source fault information feature data is acquired by means of an open interface. The multi-source fault information feature data includes distribution network model information, line topology information, line reactive information, small current line selection device information, fault indicator information and distribution transformer information. Considering the speed and stability of data acquisition, an open interface is adopted to acquire internal data through the RPA robot, and the data is stored, and the data is desensitized to ensure the security and availability of the data acquired by the RPA robot.

[0008] At the same time, due to the diversity of types of multi-source fault information feature data, there are differences in the data sources and acquisition methods. Specifically, the distribution network model information, line topology information, line reactive information, and small current line selection device information are all acquired by interface change update; the fault indicator information and distribution transformer information are acquired by manual regular import update; the data sources of the distribution network model information, line topology information, fault indicator information, and distribution transformer information are all from the distribution automation system; the data sources of the line reactive information and small current line selection device information are all from the dispatching automation system.

[0009] Preferably, in step S2, an artificial intelligence algorithm is used to analyze the single-phase grounding fault load data, and the power load data from different sources are aggregated into a single-phase grounding fault portrait. The artificial intelligence algorithm is composed of an adaptive genetic algorithm and a BP neural network optimization fusion.

[0010] Genetic algorithm is an algorithm that finds the most suitable solution set for the corresponding problem through iterative calculation. It has the advantages of global search, not being affected by the gradient value of each iteration, and not falling into the local dead zone. Using genetic algorithm to optimize the initial weight threshold of BP neural network can improve the progress and convergence speed of the network. The adaptive genetic algorithm proposed in this scheme has been improved and optimized to solve the problem that genetic algorithm is prone to fall into the local optimal solution when seeking the global optimal solution.

[0011] The formula of the crossover operator of the adaptive genetic algorithm is as follows: The formula of the mutation operator of the adaptive genetic algorithm is as follows: Among them, P c , P m are the crossover operator and mutation operator respectively; A is the coefficient of the sigmoid function; f is the fitness value of the individual to be mutated; f′ is the maximum fitness value of the two crossover individuals; f avg is the average fitness value of the population in each generation; f max is the maximum fitness value of each generation of population; P cmax , P cmin , P mmax , P mmin They are the maximum crossover operator, minimum crossover operator, maximum mutation operator, and minimum mutation operator of each generation of population respectively.

[0012] The adaptive genetic algorithm is integrated with the BP neural network, and the BP neural network is optimized using the good global search ability of the adaptive genetic algorithm to obtain the optimal network initial value. Through this optimization combination, the advantage of the strong global search ability of the adaptive genetic algorithm is brought into play, and the original self-learning ability of the network is retained, which greatly improves the performance of distribution network fault location.

[0013] The BP neural network, namely the back propagation neural network, is a traditional multi-layer feedforward neural network, comprising an input layer, a hidden layer and an output layer. It is trained according to the error back propagation algorithm and is one of the artificial neural network models currently used in various fields.

[0014] The BP neural network algorithm inputs fault feature information, and the fault feature information passes through the input layer, hidden layer and output layer of the BP neural network algorithm in sequence.

[0015] When locating a single-phase grounding fault in a distribution network, the input signal is various fault feature information, which passes through the input layer, hidden layer and output layer in sequence. The hidden layer performs internal calculation iterations on the input signal, and the calculation results are output by the next layer. If the final result is different from the expected result and has a certain degree of deviation, then it is necessary to propagate the deviation in the reverse direction and iterate and update the weights of the neural network according to the gradient descent algorithm so that the error meets the specified requirements.

[0016] The input signal net of the jth neuron in the hidden layer of the BP neural network is j for: The output signal o of the jth neuron in the hidden layer of the BP neural network is j for: The input signal net of the kth neuron node in the output layer of the BP neural network is k for: The output signal y of the kth neuron node in the output layer of the BP neural network k for: The overall error function of the BP neural network is: Where: N is the input layer node, M is the output layer node, X is the input vector, Y is the output vector, L is the hidden layer node, w ij and w jk is the connection weight between these three layers, θ jis the threshold of the hidden layer, θ k is the threshold of the output layer, is the activation function of the hidden layer, and φ(·) is the activation function of the output layer.

[0017] Preferably, in step S3, based on multi-source fault information and according to the analysis of multi-source data of single-phase grounding fault, the similarity between the fault data and the single-phase grounding fault characteristics is calculated. By calculating the similarity between each type of fault data and the single-phase grounding fault characteristics, the influence of different types of data on the single-phase grounding fault characteristics is obtained, and corresponding disposal strategies are formulated according to artificial intelligence technology to give full play to the advantages of artificial intelligence technology, improve the accuracy of single-phase grounding positioning of the power grid, and ensure the safe and stable operation of the power grid.

[0018] A distribution network fault analysis system based on artificial intelligence includes an RPA robot and a single-phase grounding fault automatic analysis and analysis software unit. The RPA robot is connected to the single-phase grounding fault automatic analysis and analysis software unit to obtain multi-source fault information feature data of electronic systems in a monitoring area and send the data to the single-phase grounding fault automatic analysis and analysis software unit.

[0019] The single-phase grounding fault automatic judgment and analysis software unit is provided with a distribution network graphic display subunit, and the distribution network graphic display subunit displays the distribution network topology data obtained by the back end on the front end page through graphics.

[0020] The beneficial effects of the present invention are: a distribution network fault assessment and analysis method and system based on artificial intelligence, which obtains multi-source fault information feature data in the distribution network of the monitoring area through the RPA robot, and analyzes the multi-source data of single-phase grounding faults using an artificial intelligence algorithm composed of an adaptive genetic algorithm and a BP neural network optimization fusion. According to the analysis results of the multi-source data of the fault using the artificial intelligence algorithm, various power data characteristics existing in the single-phase grounding fault are identified, and then the single-phase grounding fault of the distribution network is automatically located. Based on the analysis results and combined with the actual situation of the distribution network in the monitoring area, fault line selection and fault location power supply services are provided for single-phase grounding faults, and scientific operation mode optimization strategies are formulated to ensure the safe and stable operation of the power grid.

[0021] At the same time, by clustering single-phase grounding behavior with multi-source data, the fault quantity behavior characteristics of different grounding fault types can be identified. Based on the analysis results of artificial intelligence technology, a scientific, reasonable and personalized grounding fault handling strategy is formulated to reduce the power outage time of users and ensure the safe and stable operation of the power grid. The single-phase grounding fault automatic judgment and analysis software can be used to monitor, identify and analyze single-phase grounding faults in the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1It is a flow chart of the method for automatic determination and analysis of single-phase grounding fault in distribution network of the present invention; Figure 2 It is a block diagram of the automatic judgment and analysis method of single-phase grounding fault in distribution network of the present invention; Figure 3 It is a multi-source fault characteristic information data graph of the present invention; Figure 4 It is a schematic diagram of the change of the improved adaptive crossover operator of the adaptive genetic algorithm of the present invention; Figure 5 It is a schematic diagram of the change of the improved adaptive mutation operator of the adaptive genetic algorithm of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technology in the art without creative work are within the scope of protection of this application.

[0024] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0025] In this embodiment, if Figure 1 and Figure 2 As shown in the figure, a distribution network fault analysis method based on artificial intelligence includes the following steps: S1. Obtaining characteristic data of multi-source fault information of power system; S2. Use artificial intelligence algorithm to analyze multi-source data of single-phase grounding fault; S3. Identify the power data characteristics of the single-phase grounding fault according to the analysis results of the multi-source data of the single-phase grounding fault; S4, automatically locate the single-phase grounding fault in the distribution network. The distribution network fault analysis method also includes step S5, customizing the fault handling strategy in combination with the actual situation of the distribution network. This solution obtains multi-source fault information feature data in the distribution network of the monitoring area through the RPA robot. On this basis, an artificial intelligence algorithm is used to analyze the multi-source data of the single-phase grounding fault. According to the analysis results of the multi-source data of the fault by the artificial intelligence algorithm, the various power data characteristics existing in the single-phase grounding fault are identified, and then the single-phase grounding fault of the distribution network in the monitoring area of ​​the RPA robot is automatically located. According to the analysis results and automatic fault location, combined with the actual situation of the distribution network in the monitoring area, the single-phase grounding fault is provided with fault line selection and fault location power supply services, and a scientific operation mode optimization strategy is formulated to ensure the safe and stable operation of the power grid.

[0026] In this embodiment, the various power data characteristics existing in a single-phase grounding fault include a single-phase grounding fault signal, reactive power data, fault indicator data, distribution transformer data, etc. The various power data characteristics in a single-phase grounding fault are analyzed and identified through an artificial intelligence algorithm, and the grounding fault characteristics under different multi-source data are mined, and the single-phase grounding fault behavior under different grounding faults is mastered.

[0027] In this embodiment, if Figure 3 As shown, in step S1, according to the RPA robot, the multi-source fault information characteristic data of the power system is obtained, and the multi-source fault information characteristic data adopts an open interface to obtain data. The multi-source fault information characteristic data includes distribution network model information, line topology information, line reactive information, small current line selection device information, fault indicator information and distribution transformer information. Considering the speed and stability of data acquisition, an open interface is adopted to obtain internal data through the RPA robot, and the data is stored, and the data is desensitized to ensure the security and availability of the data obtained by the RPA robot.

[0028] At the same time, due to the diversity of the types of multi-source fault information feature data, the data sources obtained are different and the acquisition methods are also different. Specifically, the distribution network model information, line topology information, line reactive information, and small current line selection device information are all obtained by interface change update; the fault indicator information and distribution transformer information are all obtained by manual regular import update; the data sources of the distribution network model information, line topology information, fault indicator information, and distribution transformer information are all from the distribution automation system; the data sources of the line reactive information and small current line selection device information are all from the dispatching automation system.

[0029] In step S2, an artificial intelligence algorithm is used to analyze the single-phase grounding fault load data; the power load data from different sources are aggregated into a single-phase grounding fault portrait. The artificial intelligence algorithm is composed of an adaptive genetic algorithm and a BP neural network optimization fusion.

[0030] Genetic algorithm is an algorithm that finds the most suitable solution set for the corresponding problem through iterative calculation. It has the advantages of global search, not being affected by the gradient value of each iteration, and not falling into the local dead zone. Using genetic algorithm to optimize the initial weight threshold of BP neural network can improve the progress and convergence speed of the network. The adaptive genetic algorithm proposed in this scheme has been improved and optimized to solve the problem that genetic algorithm is prone to fall into the local optimal solution when seeking the global optimal solution.

[0031] In this implementation, if Figure 4 As shown, the formula of the crossover operator of the adaptive genetic algorithm is as follows: In this implementation, if Figure 5 As shown, the formula of the mutation operator of the adaptive genetic algorithm is as follows: Among them, P c , P m are the crossover operator and mutation operator respectively; A is the coefficient of the sigmoid function; f is the fitness value of the individual to be mutated; f′ is the maximum fitness value of the two crossover individuals; f avg is the average fitness value of the population in each generation; f max is the maximum fitness value of each generation of population; P cmax , P cmin , P mmax , P mmin They are the maximum crossover operator, minimum crossover operator, maximum mutation operator, and minimum mutation operator of each generation of population respectively.

[0032] The adaptive genetic algorithm is integrated with the BP neural network, and the BP neural network is optimized using the good global search ability of the adaptive genetic algorithm to obtain the optimal network initial value. Through this optimization combination, the advantage of the strong global search ability of the adaptive genetic algorithm is brought into play, and the original self-learning ability of the network is retained, which greatly improves the performance of distribution network fault location.

[0033] BP neural network, also known as back propagation neural network, is a traditional multi-layer feedforward neural network that includes an input layer, a hidden layer, and an output layer. It is trained based on the error back propagation algorithm and is one of the artificial neural network models that is widely used in various fields.

[0034] The BP neural network algorithm inputs fault feature information, and the fault feature information passes through the input layer, hidden layer and output layer of the BP neural network algorithm in sequence.

[0035] When locating a single-phase grounding fault in a distribution network, the input signal is various fault feature information, which passes through the input layer, hidden layer and output layer in sequence. The hidden layer performs internal calculation iterations on the input signal, and the calculation results are output by the next layer. If the final result is different from the expected result and has a certain degree of deviation, then it is necessary to propagate the deviation in the reverse direction and iterate and update the weights of the neural network according to the gradient descent algorithm so that the error meets the specified requirements.

[0036] The input signal of the jth neuron in the hidden layer of the BP neural network is j for: The output signal o of the jth neuron in the hidden layer of the BP neural network j for: The input signal of the kth neuron node in the output layer of the BP neural network is net k for: The output signal y of the kth neuron node in the output layer of the BP neural network k for: The overall error function of the BP neural network is: Where: N is the input layer node, M is the output layer node, X is the input vector, Y is the output vector, L is the hidden layer node, w ij and w jk is the connection weight between these three layers, θ j is the threshold of the hidden layer, θ k is the threshold of the output layer, is the activation function of the hidden layer, and φ(·) is the activation function of the output layer.

[0037] In step S3, based on multi-source fault information and according to the analysis of multi-source data of single-phase grounding fault, the similarity between the fault data and the single-phase grounding fault characteristics is calculated. By calculating the similarity between each type of fault data and the single-phase grounding fault characteristics, the influence of different types of data on the single-phase grounding fault characteristics is obtained, and corresponding disposal strategies are formulated according to artificial intelligence technology to give full play to the advantages of artificial intelligence technology, improve the accuracy of single-phase grounding positioning of the power grid, and ensure the safe and stable operation of the power grid.

[0038] A distribution network fault analysis system based on artificial intelligence includes an RPA robot and a single-phase grounding fault automatic analysis software unit. The RPA robot is connected to the single-phase grounding fault automatic analysis software unit to obtain multi-source fault information feature data of electronic systems in a monitoring area and send the data to the single-phase grounding fault automatic analysis software unit.

[0039] The single-phase grounding fault automatic analysis software unit is provided with a distribution network graphic display subunit. The distribution network graphic display subunit displays the distribution network topology data obtained from the back end on the front end page through graphics. The single-phase grounding fault automatic analysis software unit is provided with single-phase grounding fault automatic analysis software. The power supply company can use the single-phase grounding fault automatic analysis software to give the distribution network grounding line and grounding fault range according to the line reactive power changes, small current grounding signals, fault indicator information and distribution network phase loss information, minimize the impact of line pulling on user power supply, thereby promoting stable and orderly operation of the power grid and maximizing power supply capacity.

[0040] In this embodiment, the single-phase grounding fault automatic analysis software is based on big data mining and analysis technology. The front-end page uses Ajax to obtain the back-end power grid topology data, and uses components such as Echarts to display it as a power grid topology diagram or formatted list. The ID parameter is passed to the back-end in operations such as clicking and right-clicking. According to various operations, the front-end contact relationship, historical data, historical operation mode, etc. are returned to the front-end for displaying curves, bar charts and other graphical displays on the front end. The generation of the optimization plan is to use Python's mature algorithm library to establish a matching power grid mathematical model on the back end, calculate the optimal solution under various constraints, and then display the final plan in the interface table. Develop a single-phase grounding fault auxiliary analysis software that integrates data collection, storage, analysis application and visual display, complete product development and testing work according to the development plan, and write an analysis software operation manual.

Claims

1. A distribution network fault analysis method based on artificial intelligence, characterized in that: The following steps are involved: S1. Obtaining characteristic data of multi-source fault information of power system; S2. Use artificial intelligence algorithm to analyze multi-source data of single-phase grounding fault; S3. Identify the power data characteristics of the single-phase grounding fault according to the analysis results of the multi-source data of the single-phase grounding fault; S4. Automatically locate single-phase grounding faults in the distribution network.

2. According to the artificial intelligence-based distribution network fault analysis method of claim 1, it is characterized in that: The distribution network fault analysis method further includes step S5, customizing a fault handling strategy based on the actual situation of the distribution network.

3. According to the artificial intelligence-based distribution network fault analysis method of claim 1 or 2, it is characterized in that: In the step S1, multi-source fault information feature data is acquired by using an open interface.

4. According to the artificial intelligence-based distribution network fault analysis method of claim 1, it is characterized in that: In step S2, an artificial intelligence algorithm is used to analyze the single-phase grounding fault load data; and the power load data from different sources are aggregated into a single-phase grounding fault portrait.

5. A distribution network fault analysis method based on artificial intelligence according to claim 1 or 4, characterized in that: The artificial intelligence algorithm is composed of an adaptive genetic algorithm and a BP neural network optimization fusion.

6. According to the artificial intelligence-based distribution network fault analysis method of claim 5, it is characterized in that: The formula of the crossover operator of the adaptive genetic algorithm is as follows: The formula of the mutation operator of the adaptive genetic algorithm is as follows: Among them, P c , P m are the crossover operator and mutation operator respectively; A is the coefficient of the sigmoid function; f is the fitness value of the individual to be mutated; f′ is the maximum fitness value of the two crossover individuals; f avg is the average fitness value of the population in each generation; f max is the maximum fitness value of each generation of population; P cmax , P cmin , P mmax , P mmin They are the maximum crossover operator, minimum crossover operator, maximum mutation operator, and minimum mutation operator of each generation of population respectively.

7. According to the artificial intelligence-based distribution network fault analysis method of claim 5, it is characterized in that: The BP neural network algorithm inputs fault feature information, and the fault feature information passes through the input layer, hidden layer and output layer of the BP neural network algorithm in sequence.

8. According to the artificial intelligence-based distribution network fault analysis method of claim 1, it is characterized in that: In step S3, based on the multi-source fault information and according to the analysis of the multi-source data of the single-phase to ground fault, the similarity between the fault data and the single-phase to ground fault characteristics is calculated.

9. A distribution network fault judgment and analysis system based on artificial intelligence, applicable to the distribution network fault judgment and analysis method according to any one of claims 1 to 8, characterized in that: It includes an RPA robot and a single-phase grounding fault automatic judgment and analysis software unit. The RPA robot is connected to the single-phase grounding fault automatic judgment and analysis software unit to obtain multi-source fault information feature data of the electronic system in the monitoring area and send it to the single-phase grounding fault automatic judgment and analysis software unit.

10. The distribution network fault analysis system based on artificial intelligence according to claim 9 is characterized in that: The single-phase grounding fault automatic judgment and analysis software unit is provided with a distribution network graphic display subunit, and the distribution network graphic display subunit displays the distribution network topology data obtained by the back end on the front end page through graphics.

Images