A reliability evaluation and optimization system for relay protection systems

Abstract

The present invention relates to the technical field of relay protection system evaluation, and discloses a relay protection system reliability evaluation optimization system, including a data acquisition and preprocessing module, an equipment status analysis module, a region division and score aggregation module, and an optimization strategy generation module: the data acquisition and preprocessing module is used to collect equipment characteristic parameters of the relay protection device, and represent them through a first characteristic sequence, and preprocess them to obtain a second characteristic sequence; the equipment status analysis module is used to analyze the second characteristic sequence according to a preset equipment health evaluation model to obtain an equipment status score of the relay protection device in a second preset time period; the region division and score aggregation module is used to divide the first preset area into regions and summarize the scores to obtain an overall score; the optimization strategy generation module is used to generate an optimization strategy based on the score. The present invention significantly improves the accuracy, reliability and optimization efficiency of relay protection device status evaluation.

Description

Technical Field

[0001] The present invention belongs to the technical field of relay protection system evaluation, and in particular relates to a reliability evaluation and optimization system for a relay protection system. Background Art

[0002] As modern power grids continue to expand in scale and complexity, the safe and stable operation of power systems faces unprecedented challenges. Relay protection devices, as a crucial safeguard for grid security, shoulder the critical functions of rapidly identifying and isolating faults, protecting power equipment, and maintaining stable grid operation. Currently, relay protection devices face a variety of abnormal conditions during actual operation, including short circuits, overloads, overvoltages, and other faults, as well as malfunctions and refusal to operate. Improving the real-time performance, accuracy, and intelligence of relay protection devices is a core technical challenge in ensuring the efficient operation of power grids.

[0003] Existing technologies primarily rely on rule-based models or fixed thresholds to assess whether a device is operating normally. However, this technology has significant limitations. It lacks the ability to adapt to the dynamic characteristics of grid operation and the ability to sensitively capture device operating status, making it difficult to identify complex patterns and long-term trends in device operation. Furthermore, due to its limited intelligence, existing methods struggle to meet the scalability and real-time requirements of large-scale power grids. Summary of the Invention

[0004] The present invention provides a reliability evaluation and optimization system for a relay protection system, which solves the technical problems in related technologies that lack dynamic adaptability and sensitivity and cannot meet the real-time, reliability and accuracy requirements of modern complex power grids.

[0005] The present invention provides a reliability evaluation and optimization system for a relay protection system, including a data acquisition and preprocessing module, an equipment status analysis module, a region division and scoring summary module, and an optimization strategy generation module:

[0006] a data acquisition and preprocessing module, configured to acquire device characteristic parameters of relay protection devices in a first preset area within a first preset time period at a first preset time interval, represent the parameters through a first characteristic sequence, and preprocess the first characteristic sequence to obtain a second characteristic sequence;

[0007] Among them, the equipment characteristic parameters include: current, voltage, frequency, number of operations, communication delay and load utilization;

[0008] The i-th sequence unit of the second characteristic sequence represents the equipment characteristic parameters of the relay protection device at the i-th time point after preprocessing;

[0009] an equipment status analysis module, configured to analyze the second characteristic sequence according to a preset equipment health assessment model to obtain an equipment status score of the relay protection device for a second preset time period;

[0010] The regional division and scoring aggregation module is used to divide the first preset area into K first-level sub-areas according to the functional type of the power grid, and each first-level sub-area is divided into G second-level sub-areas. The equipment status scores of the relay protection devices in the second-level sub-areas are aggregated to generate the corresponding second-level sub-area scores. The scores of all second-level sub-areas are then aggregated to generate the first-level sub-area scores. Finally, the overall score is calculated by aggregating the first-level sub-area scores.

[0011] The optimization strategy generation module is used to generate an optimization strategy based on the overall score, the first-level sub-area score, the second-level sub-area score and the device status score.

[0012] Furthermore, the specific steps of preprocessing include:

[0013] Step S201: extract the electrical state feature according to the first feature sequence and append it to the first feature sequence. The calculation formula of the electrical state feature is: Among them, S ele Represents the electrical state characteristics, α1 represents the current-voltage product coefficient, j represents the time point index, I j represents the current at the jth time point, k represents the time point index, V k represents the voltage at the kth time point, β1 and β2 represent the current harmonic weight coefficient and voltage unbalance weight coefficient respectively, I harm Indicates the current harmonic distortion coefficient, which is determined by the proportion of harmonic components in the current, V imb represents the voltage unbalance coefficient, which is determined according to the three-phase voltage, γ represents the frequency stability coefficient, a and b represent the time point index, f a and f b They represent the frequencies of the ath time point and the bth time point, respectively, and N represents the number of time points;

[0014] Step S202: extract the action and load coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the action and load coupling feature is: Among them, A l Indicates the action and load coupling characteristics, A c and A c-1 represents the number of actions at the cth time point and the c-1th time point, t represents the first preset time interval, λ1 represents the network load weight coefficient, Lutil represents the load utilization, λ2 represents the communication delay weight coefficient, d, e and f represent the time point index, D d 、D e and D f denote the communication delays at the dth time point, the eth time point, and the fth time point, respectively;

[0015] Step S203: extract the motion and delay coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the motion and delay coupling feature is:

[0016]

[0017] Among them, A d represents the action and delay coupling feature, g, h, q, r and s represent the index of the time point, ξ1 and ξ2 represent the delay mean weight coefficient and delay jitter weight coefficient respectively, D g 、D h and D q denote the communication delay at the gth, hth, and qth time points, δ denotes the action performance adjustment coefficient, and T r and T s Represent the action time intervals at the rth and sth time points respectively;

[0018] Step S204 , using the maximum-minimum normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

[0019] Furthermore, the device health assessment model includes: an input layer, an anomaly accumulation layer, a gating mechanism layer, a hidden state update layer, and an output layer;

[0020] The input layer is used to input the second feature sequence;

[0021] The anomaly accumulation layer is used to calculate the cumulative anomaly weight at each time point in combination with the broken window effect;

[0022] The gating mechanism layer includes N first units, each of which is used to output the hidden state at the current time point according to the accumulated anomaly weight of the anomaly accumulation layer;

[0023] The hidden state update layer is used to update the hidden state according to the gating mechanism layer and the abnormality accumulation layer to obtain a first updated state;

[0024] The output layer includes a first classifier, which is used to input the first updated state output at the Nth time point into the first classifier, and the classification space of the first classifier represents the device state score of the relay protection device in the second preset time period.

[0025] Furthermore, the calculation formula of the abnormal accumulation layer is:

[0026]

[0027] in, represents the cumulative anomaly weight at the u-th time point, w1 represents the memory coefficient of historical anomalies, represents the cumulative anomaly weight at the u-1th time point, τ represents the anomaly growth coefficient, B represents the number of features in the second feature sequence, b1 represents the index of the number of features, and w b1 represents the weight coefficient of the b1th feature, represents the b1th eigenvalue at the uth time point, Indicates the first preset standard value of the b1th feature.

[0028] Furthermore, the calculation formulas for the gating mechanism layer and the hidden state update layer include:

[0029]

[0030] in, represents the first updated state output by the hidden state update layer at the c1th time point, represents the c1th sequence unit of the second characteristic sequence, and Represent the unified gate and hidden state at the c1th time point, represents the first updated state output by the hidden state update layer at the c1-1th time point, represents the cumulative abnormal weight at the c1th time point, W g 、U g and b g Respectively represent the first weight parameter, the second weight parameter and the first bias parameter corresponding to the unified gate, Represents the abnormal weight adjustment coefficient, W h 、U h and b h They represent the first weight parameter, second weight parameter, and first bias parameter corresponding to the hidden state, ζ represents the abnormality penalty coefficient, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

[0031] Furthermore, the first preset area is divided into K first-level sub-areas according to the functional type of the power grid, and each first-level sub-area corresponds to a functional type. For the relay protection devices in each first-level sub-area, the K-Means clustering algorithm is used to perform detailed division according to their longitude and latitude to obtain G second-level sub-areas.

[0032] Furthermore, the steps of the K-Means clustering algorithm include:

[0033] Step S301, using the longitude and latitude of the relay protection device as clustering input features;

[0034] Step S302, iteratively determining cluster centers through a clustering algorithm;

[0035] Step S303: Divide all relay protection devices in the first-level sub-area into G second-level sub-areas according to a first preset distance threshold.

[0036] Furthermore, the overall score is calculated as follows:

[0037]

[0038] Among them, S total represents the overall score within the first preset area, K represents the number of first-level sub-areas, d1 represents the index of the first-level sub-area, represents the weight coefficient of the first-level sub-region, represents the number of secondary sub-regions within the d1th first-level sub-region, e1 represents the index of the secondary sub-region within the d1th first-level sub-region, represents the number of relay protection devices in the e1th secondary sub-area within the d1th primary sub-area, f1 represents the index of the relay protection device in the e1th secondary sub-area within the d1th primary sub-area, It represents the equipment status score of the f1-th relay protection device in the e1-th second-level sub-area in the d1-th first-level sub-area. represents the dynamic correction coefficient, represents the failure rate of the d1th first-level sub-region obtained based on historical data, and max(R) represents the maximum failure rate among the K first-level sub-regions.

[0039] The beneficial effects of the present invention are as follows: by introducing the broken window effect modeling method to construct an equipment health assessment model, the present invention strengthens its dynamic perception of equipment anomalies, can capture the cumulative effects of abnormalities in equipment operation, and improve the accuracy and reliability of health status assessment;

[0040] By dividing the power grid area into first-level sub-areas and second-level sub-areas step by step and summarizing the scores layer by layer, the present invention can achieve hierarchical management and precise positioning of the equipment operating status, and can quickly discover areas or equipment with low scores, providing a clear direction for optimization measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a module schematic diagram of a reliability evaluation and optimization system for a relay protection system of the present invention;

[0042] Figure 2 It is a flow chart of the preprocessing of the present invention. DETAILED DESCRIPTION

[0043] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. In addition, features described with respect to some examples may also be combined in other examples.

[0044] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, but do not exclude other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0045] like Figure 1-Figure 2 As shown, a reliability evaluation and optimization system for a relay protection system includes a data acquisition and preprocessing module 101, an equipment status analysis module 102, a region division and score aggregation module 103, and an optimization strategy generation module 104:

[0046] The data collection and preprocessing module 101 is configured to collect device characteristic parameters of relay protection devices in a first preset area within a first preset time period at a first preset time interval, represent the parameters using a first characteristic sequence, and preprocess the first characteristic sequence to obtain a second characteristic sequence;

[0047] Among them, the equipment characteristic parameters include: current, voltage, frequency, number of operations, communication delay and load utilization;

[0048] The i-th sequence unit of the second characteristic sequence represents the equipment characteristic parameters of the relay protection device at the i-th time point after preprocessing;

[0049] The device status analysis module 102 is configured to analyze the second characteristic sequence according to a preset device health assessment model to obtain a device status score of the relay protection device for a second preset time period;

[0050] The region division and score aggregation module 103 is configured to divide the first preset region into K first-level sub-regions according to the functional type of the power grid, and each first-level sub-region is divided into G second-level sub-regions, and the device status scores of the relay protection devices in the second-level sub-regions are aggregated to obtain a second-level sub-region score, and the second-level sub-region scores are aggregated to obtain a first-level sub-region score, and the first-level sub-region scores are aggregated to obtain an overall score;

[0051] The optimization strategy generation module 104 is configured to generate an optimization strategy based on the overall score, the first-level sub-area score, the second-level sub-area score, and the device status score.

[0052] In one embodiment of the present invention, characteristic parameters of the device are collected using sensors installed on the relay protection device. Specifically, current and voltage are collected using current transformers and voltage transformers. The collected voltage is a sine wave that varies with time. The frequency is obtained by solving the frequency-voltage formula. The frequency-voltage formula is: Where V represents voltage, V m represents the voltage amplitude, f represents the frequency, t represents the time point, Indicates phase; the number of actions indicates the number of times the relay protection device performs a predetermined control operation based on the detected abnormal situation. It is obtained through the trigger record of the relay protection device and is used to evaluate the usage frequency and wear of the relay protection device. The time from sending a data packet to receiving a response packet is recorded, and the communication delay is calculated. The network traffic is monitored through the network performance monitoring tool Nagios to calculate the load utilization.

[0053] In one embodiment of the present invention, the specific steps of pre-processing include:

[0054] Step S201: extract the electrical state feature according to the first feature sequence and append it to the first feature sequence. The calculation formula of the electrical state feature is: Among them, S ele Represents the electrical state characteristics, α1 represents the current-voltage product coefficient, j represents the time point index, I j represents the current at the jth time point, k represents the time point index, V k represents the voltage at the kth time point, β1 and β2 represent the current harmonic weight coefficient and voltage unbalance weight coefficient respectively, I harm Indicates the current harmonic distortion coefficient, which is determined by the proportion of harmonic components in the current, V imb represents the voltage unbalance coefficient, which is determined according to the three-phase voltage, γ represents the frequency stability coefficient, a and b represent the time point index, f a and f bThey represent the frequencies of the ath time point and the bth time point, respectively, and N represents the number of time points;

[0055] Step S202: extract the action and load coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the action and load coupling feature is: Among them, A l Indicates the action and load coupling characteristics, A c and A c-1 represents the number of actions at the cth time point and the c-1th time point, t represents the first preset time interval, λ1 represents the network load weight coefficient, L util represents the load utilization, λ2 represents the communication delay weight coefficient, d, e and f represent the time point index, D d 、D e and D f denote the communication delays at the dth time point, the eth time point, and the fth time point, respectively;

[0056] Step S203: extract the motion and delay coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the motion and delay coupling feature is:

[0057]

[0058] Among them, A d represents the action and delay coupling feature, g, h, q, r and s represent the index of the time point, ξ1 and ξ2 represent the delay mean weight coefficient and delay jitter weight coefficient respectively, D g 、D h and D q denote the communication delay at the gth, hth, and qth time points, δ denotes the action performance adjustment coefficient, and T r and T s They represent the action time intervals at the rth and sth time points respectively. The action time interval is calculated by the difference between the occurrence times of the two actions.

[0059] Step S204 , using the maximum-minimum normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

[0060] In one embodiment of the present invention, the above-mentioned preprocessing steps are completed locally in the relay protection device through edge computing technology, thereby reducing dependence on central control and reducing the amount of data transmission.

[0061] In one embodiment of the present invention, the device health assessment model includes: an input layer, an anomaly accumulation layer, a gating mechanism layer, a hidden state update layer, and an output layer;

[0062] The input layer is used to input the second feature sequence;

[0063] The anomaly accumulation layer is used to calculate the cumulative anomaly weight at each time point in combination with the broken window effect;

[0064] The gating mechanism layer includes N first units, each of which is used to output the hidden state at the current time point according to the accumulated anomaly weight of the anomaly accumulation layer;

[0065] The hidden state update layer is used to update the hidden state according to the gating mechanism layer and the abnormality accumulation layer to obtain a first updated state;

[0066] The output layer includes a first classifier, which is used to input the first updated state output at the Nth time point into the first classifier, and the classification space of the first classifier represents the device state score of the relay protection device in the second preset time period.

[0067] The broken window effect emphasizes the continuous accumulation and spread of small problems. During operation, equipment may experience subtle anomalies (such as current fluctuations, voltage imbalances, frequency offsets, etc.). If these problems are not detected and handled in a timely manner, they will gradually intensify and affect the overall health status. By introducing the broken window effect in the gated recurrent unit, small anomalies in the features can be dynamically tracked through the anomaly accumulation layer, and the long-term impact of anomalies on the health status can be integrated into the state update process of the gated recurrent unit. In addition, fusing the update gate and reset gate of the gated recurrent unit and introducing anomaly weights can enhance the dynamic characteristics of the gate.

[0068] In one embodiment of the present invention, the calculation formula of the abnormal accumulation layer is:

[0069]

[0070] in, represents the cumulative anomaly weight at the u-th time point, w1 represents the memory coefficient of historical anomalies, represents the cumulative anomaly weight at the u-1th time point, τ represents the anomaly growth coefficient, B represents the number of features of the second feature sequence, preferably, in this embodiment, the number of features is 9, b1 represents the index of the number of features, represents the weight coefficient of the b1th feature, represents the b1th eigenvalue at the uth time point, Indicates the first preset standard value of the b1th feature.

[0071] In one embodiment of the present invention, the calculation formulas of the gating mechanism layer and the hidden state update layer include:

[0072]

[0073] in, represents the first updated state output by the hidden state update layer at the c1th time point, represents the c1th sequence unit of the second characteristic sequence, and Represent the unified gate and hidden state at the c1th time point, represents the first updated state output by the hidden state update layer at the c1-1th time point, represents the cumulative abnormal weight at the c1th time point, W g 、U g and b g Respectively represent the first weight parameter, the second weight parameter and the first bias parameter corresponding to the unified gate, Represents the abnormal weight adjustment coefficient, W h 、U h and b h They represent the first weight parameter, second weight parameter, and first bias parameter corresponding to the hidden state, ζ represents the abnormality penalty coefficient, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

[0074] The gating mechanism layer dynamically adjusts the degree of opening and closing of the gate by combining the cumulative anomaly weights output by the anomaly accumulation layer, enabling the model to perceive abnormal changes at each time point in real time. Time points with large anomaly weights have a more significant impact on the update of the hidden state, enhancing the model's ability to respond to sudden anomalies, capturing the cumulative effects of anomalies, and realizing the modeling of the broken window effect.

[0075] In one embodiment of the present invention, the first preset area is divided into K first-level sub-areas according to the functional type of the power grid, and each first-level sub-area corresponds to a functional type. Preferably, the functional types include: main substation protection, high-voltage transmission line protection, regional substation and downstream distribution line protection, industrial user power supply protection, distributed power supply protection and low-voltage distribution network protection; for the relay protection devices in each first-level sub-area, the K-Means clustering algorithm is used to perform detailed division according to its longitude and latitude to obtain G second-level sub-areas.

[0076] In one embodiment of the present invention, the steps of the K-Means clustering algorithm include:

[0077] Step S301, using the longitude and latitude of the relay protection device as clustering input features;

[0078] Step S302, determining the cluster center and device allocation results through algorithm iteration;

[0079] Step S303: Divide all relay protection devices in the first-level sub-area into G second-level sub-areas according to the first preset distance threshold. Specifically, for each relay protection device in the first-level sub-area, calculate the Euclidean distance between it and the relay protection devices in other first-level sub-areas. When the Euclidean distance is less than the first preset distance threshold, divide these devices into the same second-level sub-area.

[0080] By dividing the first pre-defined area into first-level and second-level sub-areas, a hierarchical management model is formed. Score aggregation at each level helps pinpoint the source of problems, quickly identify low-scoring areas or equipment, and provide clear guidance for optimization measures. Score aggregation at each level effectively identifies the severity and distribution of problems, providing a basis for tiered optimization and rational resource allocation. For example, prioritizing optimization of low-scoring sub-areas or equipment can improve resource utilization and reduce overall operational risks.

[0081] The calculation formula for the secondary sub-region score is:

[0082] The first-level sub-region scores are:

[0083] The overall score is calculated as follows:

[0084] Among them, S ss represents the secondary sub-region score of the e1th secondary sub-region within the d1th primary sub-region, S fls represents the first-level sub-region score of the d1th first-level sub-region, S total represents the overall score within the first preset area, d1 represents the index of the first-level sub-area, Represents the weight coefficient of the first-level sub-region, which is determined according to the importance of the grid function type. represents the number of secondary sub-regions within the d1th first-level sub-region, e1 represents the index of the secondary sub-region within the d1th first-level sub-region, represents the number of relay protection devices in the e1th secondary sub-area within the d1th primary sub-area, f1 represents the index of the relay protection device in the e1th secondary sub-area within the d1th primary sub-area, It represents the equipment status score of the f1-th relay protection device in the e1-th second-level sub-area in the d1-th first-level sub-area. represents the dynamic correction coefficient, represents the failure rate of the d1th first-level sub-region obtained based on historical data, and max(R) represents the maximum failure rate among the K first-level sub-regions.

[0085] In one embodiment of the present invention, when the overall score is less than the first preset score threshold, global optimization is initiated. The optimization measures at this time are: check the first-level sub-area whose score is lower than the second preset score threshold, allocate maintenance resources to the area first, strengthen regional inspections and maintenance, and troubleshoot regional problems; when the second-level sub-area score is less than the third preset score threshold, initiate local optimization, focus on optimizing the problem equipment in the second-level sub-area with the lowest score, and repair or replace the problem equipment.

[0086] The above describes the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation methods. The above specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make many forms based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A reliability evaluation and optimization system for a relay protection system, characterized in that: It includes data acquisition and preprocessing module, equipment status analysis module, area division and scoring summary module and optimization strategy generation module: a data acquisition and preprocessing module, configured to acquire device characteristic parameters of relay protection devices in a first preset area within a first preset time period at a first preset time interval, represent the parameters through a first characteristic sequence, and preprocess the first characteristic sequence to obtain a second characteristic sequence; Among them, the equipment characteristic parameters include: current, voltage, frequency, number of operations, communication delay and load utilization; The i-th sequence unit of the second characteristic sequence represents the equipment characteristic parameters of the relay protection device at the i-th time point after preprocessing; an equipment status analysis module, configured to analyze the second characteristic sequence according to a preset equipment health assessment model to obtain an equipment status score of the relay protection device for a second preset time period; The regional division and scoring aggregation module is used to divide the first preset area into K first-level sub-areas according to the functional type of the power grid, and each first-level sub-area is divided into G second-level sub-areas. The equipment status scores of the relay protection devices in the second-level sub-areas are aggregated to generate the corresponding second-level sub-area scores. The scores of all second-level sub-areas are then aggregated to generate the first-level sub-area scores. Finally, the overall score is calculated by aggregating the first-level sub-area scores. An optimization strategy generation module is used to generate an optimization strategy based on the overall score, the first-level sub-area score, the second-level sub-area score, and the device status score; The device health assessment model includes: input layer, abnormality accumulation layer, gating mechanism layer, hidden state update layer and output layer; The input layer is used to input the second feature sequence; The anomaly accumulation layer is used to calculate the cumulative anomaly weight at each time point in combination with the broken window effect; The gating mechanism layer includes N first units, each of which is used to output the hidden state at the current time point according to the accumulated anomaly weight of the anomaly accumulation layer; The hidden state update layer is used to update the hidden state according to the gating mechanism layer and the abnormality accumulation layer to obtain a first updated state; The output layer includes a first classifier, which is used to input the first updated state output at the Nth time point into the first classifier, and the classification space of the first classifier represents the device state score of the relay protection device in the second preset time period; The calculation formula of the abnormal accumulation layer is: in, represents the cumulative anomaly weight at the u-th time point, w1 represents the memory coefficient of historical anomalies, g represents the cumulative anomaly weight at the u-1th time point, τ represents the anomaly growth coefficient, B represents the number of features in the second feature sequence, b1 represents the index of the number of features, represents the weight coefficient of the b1th feature, represents the b1th eigenvalue at the uth time point, represents the first preset standard value of the b1th feature; The calculation formulas for the gating mechanism layer and the hidden state update layer include: in, represents the first updated state output by the hidden state update layer at the c1th time point, represents the c1th sequence unit of the second characteristic sequence, and Represent the unified gate and hidden state at the c1th time point, represents the first updated state output by the hidden state update layer at the c1-1th time point, represents the cumulative abnormal weight at the c1th time point, W g 、U g and b g They represent the first weight parameter, the second weight parameter, and the first bias parameter corresponding to the unified gate, φ represents the abnormal weight adjustment coefficient, and W h 、U h and b h They represent the first weight parameter, second weight parameter, and first bias parameter corresponding to the hidden state, ζ represents the abnormality penalty coefficient, sigmoid represents the sigmoid activation function, and tanh represents the tanh activation function.

2. A relay protection system reliability evaluation and optimization system according to claim 1, characterized in that: The specific steps of preprocessing include: Step S201: extract the electrical state feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the electrical state feature is: Among them, S ele Represents the electrical state characteristics, α1 represents the current-voltage product coefficient, j represents the time point index, I j represents the current at the jth time point, k represents the time point index, V k represents the voltage at the kth time point, β1 and β2 represent the current harmonic weight coefficient and voltage unbalance weight coefficient respectively, I harm Indicates the current harmonic distortion coefficient, which is determined by the proportion of harmonic components in the current, V imb represents the voltage unbalance coefficient, which is determined according to the three-phase voltage, γ represents the frequency stability coefficient, a and b represent the time point index, f a and f b They represent the frequencies of the ath time point and the bth time point, respectively, and N represents the number of time points; Step S202: extract the action and load coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the action and load coupling feature is: Among them, A l Indicates the action and load coupling characteristics, A c and A c-1 represents the number of actions at the cth time point and the c-1th time point, t represents the first preset time interval, λ1 represents the network load weight coefficient, L util represents the load utilization, λ2 represents the communication delay weight coefficient, d, e and f represent the time point index, D d 、D e and D f denote the communication delays at the dth time point, the eth time point, and the fth time point, respectively; Step S203: extract the motion and delay coupling feature according to the first feature sequence and concatenate it after the first feature sequence. The calculation formula of the motion and delay coupling feature is: Among them, A d represents the action and delay coupling feature, g, h, q, r and s represent the index of the time point, ξ1 and ξ2 represent the delay mean weight coefficient and delay jitter weight coefficient respectively, D g 、D h and D q denote the communication delay at the gth, hth, and qth time points, δ denotes the action performance adjustment coefficient, and T r and T s Represent the action time intervals at the rth and sth time points respectively; Step S204 , using the maximum-minimum normalization method to normalize the concatenated first feature sequence to obtain a second feature sequence.

3. A relay protection system reliability evaluation and optimization system according to claim 1, characterized in that: The first preset area is divided into K first-level sub-areas according to the functional type of the power grid. Each first-level sub-area corresponds to a functional type. For the relay protection devices in each first-level sub-area, the K-Means clustering algorithm is used to perform detailed division according to their longitude and latitude to obtain G second-level sub-areas.

4. A relay protection system reliability evaluation and optimization system according to claim 3, characterized in that: The steps of the K-Means clustering algorithm include: Step S301, using the longitude and latitude of the relay protection device as clustering input features; Step S302, iteratively determining cluster centers through a clustering algorithm; Step S303: Divide all relay protection devices in the first-level sub-area into G second-level sub-areas according to a first preset distance threshold.

5. A relay protection system reliability evaluation and optimization system according to claim 1, characterized in that: The overall score is calculated as follows: Among them, S total represents the overall score within the first preset area, K represents the number of first-level sub-areas, d1 represents the index of the first-level sub-area, represents the weight coefficient of the first-level sub-region, represents the number of secondary sub-regions within the d1th first-level sub-region, e1 represents the index of the secondary sub-region within the d1th first-level sub-region, represents the number of relay protection devices in the e1th secondary sub-area within the d1th primary sub-area, f1 represents the index of the relay protection device in the e1th secondary sub-area within the d1th primary sub-area, It represents the equipment status score of the f1-th relay protection device in the e1-th second-level sub-area in the d1-th first-level sub-area. represents the dynamic correction coefficient, represents the failure rate of the d1th first-level sub-region obtained based on historical data, and max(R) represents the maximum failure rate among the K first-level sub-regions.

Images