A method for analyzing power system faults based on artificial intelligence

Abstract

The present invention relates to the technical field of power system fault analysis, and more specifically, to a method for analyzing power system faults based on artificial intelligence. The method comprises the following steps: collecting electrical data, grid topology operation data, and protection and control equipment data from the power system; using fractal geometry theory to calculate the multi-scale fractal dimension characteristics of the electrical data, combining space-time curvature to capture the local dynamic characteristics of the electrical waveform and construct a joint feature space, and identifying fault modes through a machine learning model; defining the grid operation state based on the grid topology operation data and protection and control equipment data and constructing a multi-layer association graph model, using logical deduction and dynamic rules to determine the final set of fault sources; and generating different fault warning signals. This method for analyzing power system faults based on artificial intelligence captures the dynamic characteristics of power system faults by combining a joint feature space based on fractal geometry and space-time curvature, and accurately locates the fault source through a multi-layer association graph model.

Description

Technical Field

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

[0002] The method of analyzing power system faults based on artificial intelligence aims to improve the accuracy of fault diagnosis and enhance the response speed of fault detection. By combining the joint feature space construction of fractal geometry and space-time curvature, as well as multi-layer association graph models and dynamic rule reasoning, it controls the identification of power system fault modes and the location of fault sources, thereby achieving rapid and accurate identification of abnormal events in the power system, reducing system downtime and improving the stability of power supply.

[0003] Existing methods for analyzing power system faults often struggle to effectively capture nonlinear and complex dynamic fault modes. Furthermore, because traditional fault diagnosis relies on fixed rules and manual experience, it cannot accurately and timely analyze power system faults when faced with large-scale, complex power grids and new multi-dimensional fault modes involving topology, current flow, power load, and equipment status. Therefore, a method for analyzing power system faults based on artificial intelligence is provided. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for analyzing power system faults based on artificial intelligence to solve the problem raised in the above background technology that traditional fault diagnosis relies on fixed rules and manual experience, which leads to the inability to timely and accurately analyze power system faults when facing large-scale complex power grids and multi-dimensional new fault modes involving topology, current flow, power load and equipment status.

[0005] To achieve the above object, the present invention provides a method for analyzing power system faults based on artificial intelligence, comprising the following steps:

[0006] S1. Collect electrical data of the power system, grid topology operation data, and protection and control equipment data, and pre-process the data;

[0007] S2. Use fractal geometry theory to calculate the multi-scale fractal dimension characteristics of electrical data, combine space-time curvature to capture the local dynamic characteristics of electrical waveforms, construct a joint feature space, and identify fault modes through machine learning models;

[0008] S3. Define the grid operation status based on the grid topology operation data and protection and control equipment data and build a multi-layer association graph model, and use logical deduction and dynamic rules to determine the final fault source set;

[0009] S4. Generate different fault warning signals according to the final fault source set and the grid operation status.

[0010] In S1, the electrical data of the power system, the grid topology operation data and the protection and control equipment data are collected and pre-processed. The specific data are as follows:

[0011] S1.1. Collect electrical data of the power system, including current ,Voltage ,frequency ,power and harmonic current amplitude ;in Indicates the nodes;

[0012] S1.2. Collect grid topology operation data of the power system, including power flow and load distribution , obtain the topology of the power grid and construct the topological adjacency matrix ;

[0013] S1.3. Collect data on protection and control equipment of the power system, including relay status and circuit breaker status , and set the device fault indication ;

[0014] S1.4. Use Kalman filtering to filter the electrical data, and use maximum and minimum value normalization to normalize the electrical data.

[0015] As a further improvement of this technical solution, in S2, fractal geometry theory is used to calculate the multi-scale fractal dimension characteristics of electrical data, and the local dynamic characteristics of the electrical waveform are captured by combining time and space curvature to construct a joint feature space. The fault mode is identified through a machine learning model. The specific steps are as follows:

[0016] S2.1. Use wavelet transform to decompose the current and voltage signals and obtain frequency bands respectively. Current subsequence and voltage subsequence , is the number of frequency bands;

[0017] S2.2. For each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension ;

[0018] S2.3, through current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage;

[0019] S2.4, the integrated current fractal dimension , comprehensive voltage fractal dimension , the space-time curvature of current and the space-time curvature of voltage are combined to construct a joint feature space, and the support vector machine model is used to train and analyze the joint feature space to identify fault modes in the power system.

[0020] As a further improvement of this technical solution, in S2.2, for each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension , the specific steps are as follows:

[0021] S2.2.1. For each frequency band , using the basic formula of the box counting method to calculate the frequency band The fractal dimension of the current and voltage fractal dimension :

[0022] ;

[0023] ;

[0024] in, is the box size of the current signal; The current signal in the frequency band The number of boxes required; is the current fractal dimension; is the box size of the voltage signal; The voltage signal in the frequency band The number of boxes required; is the voltage fractal dimension;

[0025] S2.2.2. Obtaining the comprehensive current fractal dimension by weighted average method and integrated voltage fractal dimension :

[0026] ;

[0027] ;

[0028] in, Frequency band The weight coefficient of is the total number of frequency bands.

[0029] As a further improvement of this technical solution, in S2.3, the current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage. The specific method is as follows:

[0030] ;

[0031] ;

[0032] in, For time; is the space-time curvature of the current; is the voltage space-time curvature.

[0033] As a further improvement of this technical solution, in S2.4, the comprehensive current fractal dimension , comprehensive voltage fractal dimension , current space-time curvature and voltage space-time curvature Combined with the construction of a joint feature space, and using the support vector machine model to train and analyze the joint feature space, the fault mode in the power system is identified. The specific steps are as follows:

[0034] S2.4.1. Based on the comprehensive current fractal dimension , comprehensive voltage fractal dimension , current space-time curvature and voltage space-time curvature Combine to construct a joint feature space:

[0035] ;

[0036] in, To construct the joint feature space;

[0037] S2.4.2. Build a support vector machine model:

[0038] Objective function:

[0039] ;

[0040] Constraints:

[0041] ;

[0042] in, is the weight vector; is the transpose of the weight vector; For the slack variables in the joint feature space; is the regularization parameter; For the Failure mode categories in the joint feature space; is the bias term; For the joint feature space; is the total number of joint feature spaces; is the joint feature space label, ;

[0043] S2.4.3. Constructing a training dataset , and train the support vector machine model to obtain the optimal and Complete support vector machine model training;

[0044] S2.4.4. Use the support vector machine model to analyze the new joint feature space Perform analysis to identify failure modes in power systems:

[0045] ;

[0046] in, is the new joint feature space; New failure mode categories;

[0047] when When , it means the power system is normal;

[0048] when When , it indicates a power system failure.

[0049] As a further improvement of this technical solution, the multi-layer association graph model includes a topology layer, a flow direction layer, a load distribution layer, and an equipment status layer, which is used to locate the fault source of the power system in combination with dynamic logical reasoning;

[0050] In S3, the grid operation status is defined based on the grid topology operation data and the protection and control equipment data, and a multi-layer association graph model is constructed. The final fault source set is determined using logical deduction and dynamic rules. The specific method is as follows:

[0051] S3.1. Define the grid operation status based on the grid topology operation data and protection and control equipment data and construct a multi-layer association graph model;

[0052] S3.2. Based on the multi-layer association graph model, the fault source is located through logical reasoning and dynamic rules.

[0053] As a further improvement of the present technical solution, in S3.1, the grid operation state is defined based on the grid topology operation data and the protection and control equipment data and a multi-layer association graph model is constructed. The specific method steps are as follows:

[0054] Multi-layer association graph model:

[0055] Topology layer:

[0056] ;

[0057] Flow direction layer:

[0058] ;

[0059] Load distribution layer:

[0060] ;

[0061] in, For nodes The set of connected adjacent nodes;

[0062] Device status layer:

[0063] ;

[0064] in, For the device status layer expression, Indicates that the equipment power system is normal; Indicates that the equipment power system is normal; Indicates a fault in the device's power system.

[0065] As a further improvement of this technical solution, in S3.2, based on a multi-layer association graph model, the fault source is located through logical reasoning and dynamic rules. The specific method steps are as follows:

[0066] S3.2.1. Analyze the power system by combining the topology layer and the device status layer:

[0067] ;

[0068] in, is the set of equipment failure nodes; is the grid edge set;

[0069] S3.2.2. Analyze the power system based on the flow layer:

[0070] ;

[0071] in, is the set of power flow abnormal nodes;

[0072] S3.2.3. Analyze the power system based on the load distribution layer:

[0073] ;

[0074] in, is the set of nodes with abnormal load distribution; is the maximum load value; is the minimum load value;

[0075] S3.2.4, combined with the fault screening of each layer, the equipment fault node set , power flow abnormal node set and the set of abnormal load distribution nodes Find the intersection and get the final fault source set:

[0076] ;

[0077] in, is the final fault source set.

[0078] As a further improvement of this technical solution, in S4, different fault warning signals are generated according to the final fault source set and the grid operation status, as follows:

[0079] S4.1. If , generating a “power system normal” signal;

[0080] S4.2, if , generate "power system fault, fault source is "Signal;

[0081] S4.3, if , generating an “unknown fault” signal.

[0082] Compared with the prior art, the present invention has the following beneficial effects:

[0083] 1. This method for analyzing power system faults based on artificial intelligence captures the multi-dimensional, nonlinear dynamic fault characteristics in the power system by constructing a joint feature space based on fractal geometry and space-time curvature.

[0084] 2. In this method of analyzing power system faults based on artificial intelligence, the rapid location and accurate diagnosis of the fault source are achieved through the combination of a multi-layer association graph model and dynamic rule reasoning. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 The figure is a flow chart of the overall method of the present invention. DETAILED DESCRIPTION

[0086] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0087] See also Figure 1 As shown, this embodiment provides a method for analyzing power system faults based on artificial intelligence, including the following steps:

[0088] S1. Collect electrical data of the power system, grid topology operation data, and protection and control equipment data, and pre-process the data;

[0089] In this embodiment S1, electrical data of the power system, grid topology operation data, and protection and control equipment data are collected and pre-processed. The specific data are as follows:

[0090] S1.1. Collect electrical data of the power system, including current ,Voltage ,frequency ,power and harmonic current amplitude ;in Indicates the nodes;

[0091] S1.2. Collect grid topology operation data of the power system, including power flow and load distribution , obtain the topology of the power grid and construct the topological adjacency matrix ;

[0092] S1.3. Collect data on protection and control equipment of the power system, including relay status and circuit breaker status , and set the device fault indication ;

[0093] S1.4. Use Kalman filtering to filter the electrical data, and use maximum and minimum value normalization to normalize the electrical data.

[0094] In this embodiment, For the The state of a relay, the value is 0 or 1, 0 means closed, 1 means open; For the The state of a circuit breaker, the value is 0 or 1, 0 means open, 1 means closed; For the The fault indication of a device, the value is 0 or 1, 0 means no fault, 1 means a fault occurs.

[0095] S2. Use fractal geometry theory to calculate the multi-scale fractal dimension characteristics of electrical data, combine space-time curvature to capture the local dynamic characteristics of electrical waveforms, construct a joint feature space, and identify fault modes through machine learning models;

[0096] In S2, fractal geometry theory is used to calculate the multi-scale fractal dimension characteristics of electrical data, and the local dynamic characteristics of electrical waveforms are captured by combining time and space curvature to construct a joint feature space. The fault mode is identified through a machine learning model. The specific steps are as follows:

[0097] S2.1. Use wavelet transform to decompose the current and voltage signals and obtain frequency bands respectively. Current subsequence and voltage subsequence , is the number of frequency bands;

[0098] S2.2. For each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension ;

[0099] S2.3, through current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage;

[0100] S2.4, the integrated current fractal dimension , comprehensive voltage fractal dimension , the space-time curvature of current and the space-time curvature of voltage are combined to construct a joint feature space, and the support vector machine model is used to train and analyze the joint feature space to identify fault modes in the power system.

[0101] In this embodiment S2.2, for each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension , the specific steps are as follows:

[0102] S2.2.1. For each frequency band , using the basic formula of the box counting method to calculate the frequency band The fractal dimension of the current and voltage fractal dimension :

[0103] ;

[0104] ;

[0105] in, is the box size of the current signal; The current signal in the frequency band The number of boxes required; is the current fractal dimension; is the box size of the voltage signal; The voltage signal in the frequency band The number of boxes required; is the voltage fractal dimension;

[0106] S2.2.2. Obtaining the comprehensive current fractal dimension by weighted average method and integrated voltage fractal dimension :

[0107] ;

[0108] ;

[0109] in, Frequency band The weight coefficient of is the total number of frequency bands.

[0110] In this embodiment S2.3, the current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage. The specific method is as follows:

[0111] ;

[0112] ;

[0113] in, For time; is the space-time curvature of the current; is the voltage space-time curvature.

[0114] In this embodiment, the unit of current is ampere, the unit of voltage is volt, and the unit of space-time curvature is dimensionless, reflecting the severity of the waveform change; if the current waveform changes sharply at a certain moment, the second-order derivative will increase significantly, and the space-time curvature will also reflect the sharp change at that moment.

[0115] In this embodiment S2.4, the integrated current fractal dimension , comprehensive voltage fractal dimension , current space-time curvature and voltage space-time curvature Combined with the construction of a joint feature space, and using the support vector machine model to train and analyze the joint feature space, the fault mode in the power system is identified. The specific steps are as follows:

[0116] S2.4.1. Based on the comprehensive current fractal dimension , comprehensive voltage fractal dimension , current space-time curvature and voltage space-time curvature Combine to construct a joint feature space:

[0117] ;

[0118] in, To construct the joint feature space;

[0119] S2.4.2. Build a support vector machine model:

[0120] Objective function:

[0121] ;

[0122] Constraints:

[0123] ;

[0124] in, is the weight vector; is the transpose of the weight vector; For the slack variables in the joint feature space; is the regularization parameter; For the Failure mode categories in the joint feature space; is the bias term; For the joint feature space; is the total number of joint feature spaces; is the joint feature space label, ;

[0125] S2.4.3. Constructing a training dataset , and train the support vector machine model to obtain the optimal and Complete support vector machine model training;

[0126] S2.4.4. Use the support vector machine model to analyze the new joint feature space Perform analysis to identify failure modes in power systems:

[0127] ;

[0128] in, is the new joint feature space; New failure mode categories;

[0129] when When , it means the power system is normal;

[0130] when , it indicates a power system failure.

[0131] S3. Define the grid operation status based on the grid topology operation data and protection and control equipment data and build a multi-layer association graph model, and use logical deduction and dynamic rules to determine the final fault source set;

[0132] The multi-layer association graph model includes a topology layer, a flow direction layer, a load distribution layer, and an equipment status layer, and is used to locate the fault source of the power system in combination with dynamic logical reasoning;

[0133] In this embodiment S3, the grid operation state is defined based on the grid topology operation data and the protection and control equipment data, and a multi-layer association graph model is constructed. The final fault source set is determined using logical deduction and dynamic rules. The specific method is as follows:

[0134] S3.1. Define the grid operation status based on the grid topology operation data and protection and control equipment data and construct a multi-layer association graph model;

[0135] S3.2. Based on the multi-layer association graph model, the fault source is located through logical reasoning and dynamic rules.

[0136] In this embodiment S3.1, the grid operation state is defined based on the grid topology operation data and the protection and control equipment data and a multi-layer association graph model is constructed. The specific method steps are as follows:

[0137] Multi-layer association graph model:

[0138] Topology layer:

[0139] ;

[0140] Flow direction layer:

[0141] ;

[0142] Load distribution layer:

[0143] ;

[0144] in, For nodes The set of connected adjacent nodes;

[0145] Device status layer:

[0146] ;

[0147] in, For the device status layer expression, Indicates that the equipment power system is normal; Indicates that the equipment power system is normal; Indicates a fault in the device's power system.

[0148] In S3.2 of this embodiment, based on the multi-layer association graph model, the fault source is located through logical reasoning and dynamic rules. The specific method steps are as follows:

[0149] S3.2.1. Analyze the power system by combining the topology layer and the device status layer:

[0150] ;

[0151] in, is the set of equipment failure nodes; is the grid edge set;

[0152] S3.2.2. Analyze the power system based on the flow layer:

[0153] ;

[0154] in, is the set of power flow abnormal nodes;

[0155] S3.2.3. Analyze the power system based on the load distribution layer:

[0156] ;

[0157] in, is the set of nodes with abnormal load distribution; is the maximum load value; is the minimum load value;

[0158] S3.2.4, combined with the fault screening of each layer, the equipment fault node set , power flow abnormal node set and the set of abnormal load distribution nodes Find the intersection and get the final fault source set:

[0159] ;

[0160] in, is the final fault source set.

[0161] S4. Generate different fault warning signals according to the final fault source set and the grid operation status;

[0162] In this embodiment S4, different fault warning signals are generated according to the final fault source set and the grid operation status, as follows:

[0163] S4.1. If , generating a “power system normal” signal;

[0164] S4.2, if , generate "power system fault, fault source is "Signal;

[0165] S4.3. If , generating an “unknown fault” signal.

[0166] The basic principles, main features, and advantages of the present invention are shown and described above. It should be understood by those skilled in the art that the present invention is not limited to the above-described embodiments. The above-described embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention claimed.

Claims

1. A method for analyzing power system faults based on artificial intelligence, characterized by: The following steps are involved: S1. Collect electrical data of the power system, grid topology operation data, and protection and control equipment data, and pre-process the data; S2. Use fractal geometry theory to calculate the multi-scale fractal dimension characteristics of electrical data, combine space-time curvature to capture the local dynamic characteristics of electrical waveforms, construct a joint feature space, and identify fault modes through machine learning models; S3. Define the grid operation status based on the grid topology operation data and protection and control equipment data and build a multi-layer association graph model, and use logical deduction and dynamic rules to determine the final fault source set; S4. Generate different fault warning signals according to the final fault source set and the grid operation status; In S2, fractal geometry theory is used to calculate the multi-scale fractal dimension characteristics of electrical data, and the local dynamic characteristics of electrical waveforms are captured by combining time and space curvature to construct a joint feature space. The fault mode is identified through a machine learning model. The specific steps are as follows: S2.

1. Use wavelet transform to decompose the current and voltage signals and obtain frequency bands respectively. Current subsequence and voltage subsequence ; is one of the nodes; S2.

2. For each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension ; S2.3, through current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage; S2.4, Based on the comprehensive current fractal dimension , comprehensive voltage fractal dimension , current space-time curvature and voltage space-time curvature Combined with constructing joint feature space, ,in, To construct a joint feature space and use the support vector machine model to train and analyze the joint feature space to identify fault modes in the power system; In S2.3, the current and voltage About time The second-order derivative is used to approximate the space-time curvature of current and voltage. The specific method is as follows: ； ； in, For time; is the space-time curvature of the current; is the voltage space-time curvature.

2. The method for analyzing power system faults based on artificial intelligence according to claim 1, characterized in that: In S1, the electrical data of the power system, the grid topology operation data and the protection and control equipment data are collected and pre-processed. The specific data are as follows: S1.

1. Collect electrical data of the power system, including current ,Voltage ,frequency ,power and harmonic current amplitude; S1.

2. Collect grid topology operation data of the power system, including power flow and load distribution , obtain the topology of the power grid and construct the topological adjacency matrix ,in, , They are two different nodes respectively; S1.

3. Collect data on protection and control equipment of the power system, including relay status and circuit breaker status , and set the device fault indication ; S1.

4. Use Kalman filtering to filter the electrical data, and use maximum and minimum value normalization to normalize the electrical data.

3. The method for analyzing power system faults based on artificial intelligence according to claim 1, characterized in that: In S2.2, for each frequency band , calculate the frequency bands by box counting method The fractal dimension of the current and voltage fractal dimension , and the comprehensive current fractal dimension is obtained by weighted average method and integrated voltage fractal dimension , the specific steps are as follows: S2.2.

1. For each frequency band , using the basic formula of the box counting method to calculate the frequency band The fractal dimension of the current and voltage fractal dimension : ； ； in, is the box size of the current signal; The current signal in the frequency band The number of boxes required; is the current fractal dimension; is the box size of the voltage signal; The voltage signal in the frequency band The number of boxes required; is the voltage fractal dimension; S2.2.

2. Obtaining the comprehensive current fractal dimension by weighted average method and integrated voltage fractal dimension : ； ； in, Frequency band The weight coefficient of is the total number of frequency bands.

4. The method for analyzing power system faults based on artificial intelligence according to claim 1, characterized in that: The support vector machine model in S2.4 includes Objective function: ； Constraints: ； in, is the weight vector; is the transpose of the weight vector; For the slack variables in the joint feature space; is the regularization parameter; For the Failure mode categories in the joint feature space; is the bias term; For the joint feature space; is the total number of joint feature spaces; is the joint feature space label, ; Building a training dataset , and train the support vector machine model to obtain the optimal and Complete support vector machine model training; The new joint feature space is trained using the support vector machine model Perform analysis to identify failure modes in power systems: ； in, is the new joint feature space; New failure mode categories; when When , it means the power system is normal; when When , it indicates a power system failure.

5. The method for analyzing power system faults based on artificial intelligence according to claim 1, characterized in that: The multi-layer association graph model includes a topology layer, a flow direction layer, a load distribution layer, and an equipment status layer, and is used to locate the fault source of the power system in combination with dynamic logical reasoning; In S3, the grid operation status is defined based on the grid topology operation data and the protection and control equipment data, and a multi-layer association graph model is constructed. The final fault source set is determined using logical deduction and dynamic rules. The specific method is as follows: S3.

1. Define the grid operation status based on the grid topology operation data and protection and control equipment data and construct a multi-layer association graph model; S3.

2. Based on the multi-layer association graph model, the fault source is located through logical reasoning and dynamic rules.

6. The method for analyzing power system faults based on artificial intelligence according to claim 1, characterized in that: In S4, different fault warning signals are generated according to the final fault source set and the grid operation status, as follows: S4.

1. If , generating a "power system normal" signal; S4.2, if , generate "power system fault, fault source is "Signal, is the final fault source set; S4.

3. If , generating an "unknown fault" signal.

Images