Automatic verification system and method for relay protection of power system

Abstract

The invention discloses an electric power system relay protection automatic verification system and method, and the system comprises an electrical characteristic collection module which is used for obtaining key electrical data of an electric power system in real time; the data processing module is used for preprocessing the collected original data and extracting features beneficial to fault detection; the deep belief network model is used for judging whether the operation state of the power grid is abnormal or not according to the electrical data flow data of the power grid, and immediately judging the abnormal fault type if the abnormality is found; and the automatic point checking module is used for adjusting the operation parameters of the relay, matching the response of the relay with the actual power grid state, restoring the power grid to the normal operation state and generating the adjusted protection parameter configuration according to the fault type identification result. The reliability and safety of the power system can be remarkably improved, equipment damage or power interruption caused by improper setting of the relay is reduced, and the method is suitable for further popularization and application.

Description

Technical Field

[0001] The present invention relates to the technical field of relay protection, and in particular to an automatic verification system and method for relay protection of a power system. Background Art

[0002] With the continuous development of power systems, the scale and complexity of power grids are also increasing. In order to ensure the stable operation of power systems, the application of relay protection equipment has become increasingly important. The main function of relay protection equipment is to monitor the operating status of the power system and act quickly and accurately when a fault occurs to reduce equipment damage and the impact of power outages. However, traditional relay protection equipment often relies on manually configured parameters, the setting of which requires a lot of experience and technical knowledge, and is easily affected by human factors, resulting in inaccurate parameter settings, which in turn affects the performance of relay protection equipment. Secondly, existing methods are often difficult to adapt to complex and uncertain power grid environments, and are prone to misjudgment or missed judgments. In addition, existing methods need to be improved in terms of real-time and accuracy, and cannot meet the application requirements of scenarios with high reliability and high security requirements for power grids.

[0003] At the same time, the development of artificial intelligence technology and deep learning technology has provided new ideas for solving the above problems. Deep learning technology can automatically extract features, make predictions and make decisions through a large amount of data training models. Therefore, applying deep learning technology to the automatic configuration and verification of relay protection in power systems can improve the accuracy and efficiency of parameter configuration of relay protection systems, thereby improving the reliability and safety of power systems and the automation level of relay protection systems. Therefore, how to develop a scientific, effective and automated automatic verification system and method for relay protection in power systems is an important issue facing current technology. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a system and method for automatic verification of relay protection in an electric power system. The system and method collect electrical characteristic data such as current, voltage and frequency waveforms of the electric power system for preprocessing, and then compose a data set for training a model. The system accurately evaluates the type of abnormality according to the real-time operating characteristics of the electric power system, analyzes the impact of faults by learning a large amount of historical data, and accurately verifies the relay configuration parameters. The automatic point verification can significantly improve the reliability and safety of the electric power system and reduce equipment damage or power outages caused by improper relay settings.

[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0006] An automatic verification system for relay protection of a power system, comprising:

[0007] Electrical feature collection module, used to obtain key electrical data of the power system in real time;

[0008] A data processing module is used to pre-process the collected raw data and extract features that are helpful for fault detection;

[0009] The deep belief network model is used to determine whether the grid operation status is abnormal based on the electrical data flow data of the grid, and immediately determine the abnormal fault type if an abnormality is found;

[0010] The automatic point verification module is used to adjust the operating parameters of the relay according to the fault type identification results, match the relay response with the actual power grid status, restore the power grid to normal operation, and generate the adjusted protection parameter configuration.

[0011] As a possible implementation, further, the electrical feature collection module collects current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time tag and amplitude value of each waveform, and stores the data as a time series to generate a real-time waveform record of the power grid;

[0012] In the real-time monitoring of current and voltage data, each collected data point records the amplitude and time information at the same time. Each waveform data point is equipped with a time tag and a corresponding amplitude value. The time and amplitude of data collection are recorded synchronously by the timer and sensor.

[0013] The time and amplitude of each data point are integrated to generate a time-amplitude pairing, which is expressed as follows:

[0014] P(t i )=(t i ,V i )

[0015] P(t i ) represents the time point t i Paired data points; t i is the time tag recorded by the data acquisition timer, indicating the sampling time of the i-th data point; V i At time point t i The collected amplitude value, i.e. the instantaneous value of current or voltage;

[0016] The integrated time-amplitude paired data are stored in a cloud database.

[0017] As a possible implementation, further, the data processing module preprocesses the data stored in the cloud database, and then normalizes it, and marks the data according to known abnormal events to generate data labels, and then divides the data into a training set and a test set after the preprocessing.

[0018] As a possible implementation, further, the preprocessing includes cleaning, removing or filling missing values, deleting outliers or erroneous data points, and extracting features from the original data;

[0019] The features extracted from the original data include the normal range of electrical characteristics such as current and voltage waveforms in the electrical data and the abnormal values ​​of current and voltage during short circuit.

[0020] As a possible implementation, further, the deep belief network model is trained by preprocessed historical electrical data until the model converges;

[0021] The historical electrical data includes the current, voltage and frequency variation range during daily operation and any known abnormal short circuit events.

[0022] As a possible implementation method, further, the deep confidence network model uses the latest collected operating electrical characteristic data of the power system to evaluate the system operating status in real time. The model will automatically determine the current operating status of the operating area based on the real-time electrical parameter information. When the operating status sends an abnormality, it will determine the type of abnormality and determine whether to further update the protection parameter configuration.

[0023] As a possible implementation mode, further, the automatic point calibration module adjusts the operating parameters of the relay including the setting of the trigger time and the protection threshold;

[0024] The automatic point check module analyzes the impact of the fault on the operation status of the power grid, including frequency anomaly or voltage drop, based on the fault type identification result, and sets the initial value W of the protection threshold according to the impact. The initial setting is based on the fault severity and historical data statistics. The calculation formula is:

[0025] W=u*L

[0026] Among them, u is an adjustment coefficient related to the fault type, and L is a statistic of the fault impact;

[0027] The trigger time is set according to the current load of the power grid and the urgency of fault recovery, and is calculated using a dynamic adjustment formula, which is:

[0028]

[0029]

[0030] Where T is the adjusted trigger time; T 基础 is the basic triggering time of the relay under standard conditions; k is the adjustment coefficient, which is used to adjust the triggering time according to the grid load; P is the actual load of the grid at present; P 额定is the rated load of the power grid, which is used as a reference benchmark; e is the base of the natural logarithm; a is the waveform adjustment coefficient, which can be used to control the impact of load changes on the k value; b is the offset, which is used to adjust the reference point of the load; E is the fault urgency index, which can be a quantitative representation of the fault type, fault duration or fault impact range;

[0031] The adjusted protection parameter configuration is output in combination with the adjusted protection threshold W and trigger time T, and the protection parameter configuration is applied to the relay, the response matches the grid state, and the grid is restored to a normal operating state.

[0032] The present invention also provides a method for automatic verification of relay protection in a power system, which is applied to the above-mentioned automatic verification system for relay protection in a power system, and the method comprises the following steps:

[0033] S1, the electrical characteristics collection module collects the current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time label and amplitude value of each waveform, and stores the data as a time series to generate real-time waveform records of the power grid;

[0034] S2. Calculate the change value based on the current and voltage data at two adjacent moments;

[0035] S3, extracting feature data that is helpful for fault detection from the preprocessed data;

[0036] S4. According to the existing fault data, the change values ​​are marked with normal as 0 and abnormal as 1, and the fault type labels are annotated, and then the data set is divided into a training set and a test set;

[0037] S5. Use the training set data to train the deep belief network model, learn the characteristics of normal and abnormal data, and use the test set to evaluate the accuracy and effect of the model;

[0038] S6. Use the trained model and the latest collected electrical characteristic data to monitor the operation status of the power grid. If an abnormal fault is found, determine the type of abnormal fault;

[0039] S7, recording the abnormal waveform state of the power grid when an abnormal event occurs, and updating the fixed parameters and protection thresholds of the relay according to the waveform state;

[0040] S8. Regularly retrain the model using the latest collected data, and use the newly trained model to monitor the operating status of the power grid.

[0041] As a possible implementation, further, step S2 specifically includes:

[0042] Preprocessing the collected data, wherein the preprocessing steps include: denoising, calibration, and calculating the change value of the current and voltage waveforms and the waveform adjustment coefficient according to the current and voltage data at two adjacent moments;

[0043] The change value includes an instantaneous change value, a cumulative change value, and a percentage change value;

[0044] The instantaneous change value is calculated by calculating the difference between the current or voltage values ​​at two or more consecutive time points. The expression is as follows:

[0045] ΔI=I(tn)-I(tn-1)

[0046] ΔV=V(tn)-V(tn-1)

[0047] Where I(tn) and V(tn) are the current and voltage values ​​at the nth time point, respectively;

[0048] The cumulative change value is the integration of the change of current or voltage over time to calculate the total change over a period of time. The expression is as follows:

[0049]

[0050] The percentage change value is the percentage of the calculated current or voltage change, and the expression is as follows:

[0051]

[0052] The waveform adjustment coefficient is a dynamic adjustment of the time axis of the real-time waveform data, including scaling and translation, to align with the standard waveform template. The expression is as follows:

[0053] Smodel=Sreal*α+b

[0054] Smodel is the standard waveform template, Sreal is the real-time waveform data, α and b are adjustment coefficients, which control the scaling and translation respectively.

[0055] As a possible implementation, further, the features helpful for fault detection in step S3 include sudden change, sudden drop, etc.;

[0056] The fault type labels in step S4 include: short circuit fault, overload fault, ground fault, overvoltage / undervoltage fault, harmonic interference, contactor or relay fault, etc.

[0057] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0058] 1) The present invention adopts a deep belief network model, collects the operating electrical characteristic data of the power system to form a data set, and trains the model. It accurately evaluates the type of abnormality according to the real-time operating characteristics of the power grid, can analyze the impact of faults by learning a large amount of historical data, and accurately verify the relay configuration parameters, thereby realizing the automatic configuration and verification of the relay protection of the power system, greatly improving work efficiency, reducing manual intervention and configuration time, and enhancing the reliability and safety of the power system.

[0059] 2) The deep belief network model of the present invention can automatically extract features and make decisions, avoiding the influence of human factors and improving the accuracy and reliability of parameter configuration. At the same time, through automatic point verification, the reliability and safety of the system can be significantly improved, and equipment damage or power outages caused by improper relay settings can be reduced.

[0060] 3) The deep belief network model of the present invention can automatically adjust and optimize parameter configuration when facing a complex and uncertain power grid environment. It has strong adaptability, can effectively avoid misjudgment or missed judgment, and improves the robustness of the system.

[0061] 4) The deep belief network model of the present invention can realize real-time monitoring and fault diagnosis of the power grid, and perform parameter verification and adjustment in a timely manner, thereby improving the real-time and accuracy of the system and meeting the power grid's requirements for high reliability and high security.

[0062] 5) The power system relay protection automatic verification system and method of the present invention are not only applicable to power grids in a single scenario, but can also be extended and applied to other types of power grid systems, and have broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 This is a flow chart of the automatic verification system for relay protection in the power system;

[0065] Figure 2 This is a schematic diagram of the framework of the automatic verification system for relay protection in the power system;

[0066] Figure 3 This is a schematic diagram of the time-amplitude pairing process;

[0067] Figure 4 Flowchart for fault assessment of deep belief network model;

[0068] Figure 5 This is a schematic diagram of the automatic point calibration module parameter configuration process. DETAILED DESCRIPTION

[0069] The present invention will be further described in detail below in conjunction with the accompanying drawings and examples. It is particularly noted that the following examples are only used to illustrate the present invention, but are not intended to limit the scope of the present invention. Similarly, the following examples are only partial embodiments of the present invention rather than all embodiments, and all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0070] See attached Figures 1-2 As shown, this embodiment provides a power system relay protection automatic verification system, including:

[0071] The electrical characteristics collection module is used to obtain the key electrical data of the power system in real time, and transmit it to the data processing module for preprocessing in real time. All preprocessed electrical parameter data form a comprehensive and reliable data set and are stored in a secure database for subsequent analysis. Specifically, the electrical characteristics collection module obtains the electrical characteristics monitoring flow data of the target power system operation area. The electrical characteristics monitoring flow data includes the electrical characteristics waveform monitoring flow data corresponding to the set abnormal data points in the power system during the operation period. The electrical characteristics monitoring flow data of the power system can be the waveform data stream of the target power system operation at a certain time, including time series data such as current and voltage waveforms. All electrical characteristics data will be preprocessed and composed of electrical characteristics data sets stored in a secure database for subsequent analysis.

[0072] In this embodiment, the electrical characteristics collection module collects the current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time tag and amplitude value of each waveform, and stores the data as a time series to generate a real-time waveform record of the power grid;

[0073] In the real-time monitoring of current and voltage data, each collected data point records the amplitude and time information at the same time. Each waveform data point is equipped with a time tag and a corresponding amplitude value. The time and amplitude of data collection are recorded synchronously by the timer and sensor.

[0074] The time and amplitude of each data point are integrated to generate a time-amplitude pairing, which is expressed as follows:

[0075] P(t i )=(t i ,V i )

[0076] P(t i) represents the time point t i Paired data points; t i is the time tag recorded by the data acquisition timer, indicating the sampling time of the i-th data point; V i At time point t i The collected amplitude value, that is, the instantaneous value of current or voltage (such as the attached Figure 3 The integrated time-amplitude paired data are stored in the cloud database for subsequent analysis.

[0077] The data processing module is used to pre-process the collected raw data and extract features that are helpful for fault detection, such as mutations and sudden drops. In this embodiment, the data processing module pre-processes the data stored in the cloud database, and then performs normalization processing, and generates data labels based on known abnormal event labels, for example, marking data within the normal operating range as 0 and abnormal events as 1. Then, data division is performed to divide the pre-processed data set into a training set and a test set.

[0078] Among them, preprocessing includes cleaning, removing or filling missing values, deleting outliers or erroneous data points, and extracting features from the original data; the extracted features from the original data include the normal range of electrical characteristics such as current and voltage waveforms in the electrical data and abnormal current and voltage values ​​during short circuit.

[0079] The deep belief network model is used to determine whether the operation status of the power grid is abnormal based on the electrical data flow data of the power grid. Specifically, the electrical characteristic monitoring flow data of the power system is loaded into the deep belief network model that meets the network deployment conditions to generate abnormal event waveform data. The power grid abnormal event waveform data includes at least one of the real-time abnormal event and past abnormal event corresponding to the target system operation area. The abnormal event must contain abnormal characteristic values ​​such as change values; if an abnormality is found, the abnormal fault type is immediately determined.

[0080] Among them, the deep belief network model is trained by preprocessed historical electrical data until the model converges; the historical electrical data includes the current, voltage and frequency variation range during daily operation and any known abnormal short-circuit events. In the present invention, the deep belief network model learns normal or abnormal data features based on historical data and labels. The network model is trained using the training data set. Finally, the accuracy and effect of the model are evaluated using the test set. The model parameters are adjusted based on the test set evaluation results to maximize the accuracy of evaluating the operating status of the power grid.

[0081] In this embodiment, the deep belief network model is trained as follows:

[0082] Initialize the model weight parameters of the initialized deep belief network model, generate the model weight parameters, and update the generated deep belief network model;

[0083] Obtaining a data stream of electrical characteristic waveform samples corresponding to a target power system, and loading these waveform sample data into a deep belief network evaluation model to generate sample learning results corresponding to a plurality of abnormal data;

[0084] Based on the sample learning results corresponding to all abnormal data, a deep confidence network evaluation model for updating the network weight parameters is analyzed and generated. The sample learning results corresponding to all abnormal knowledge points are used to evaluate whether the initialized deep confidence network evaluation model meets the sample learning termination requirements. If the deep confidence network evaluation model meets the sample learning termination requirements, a new deep confidence network evaluation model for updating the network weight parameters is generated based on the sample learning results corresponding to all abnormal knowledge points.

[0085] The target sample learning results corresponding to the target abnormal knowledge points are extracted from the sample learning results corresponding to all the abnormal knowledge points; the target network weight information is determined based on the target abnormal knowledge points and the target sample learning results; based on the target network weight information and the initialized power system abnormal state deep confidence network evaluation model, an evaluation network model for updating the network weight parameters is generated.

[0086] The above-mentioned judgment on whether the deep confidence network evaluation model for abnormal state of the power system meets the sample learning termination requirements comprises the following specific steps: determining the target sample abnormal learning data corresponding to each of the abnormal data, and for each of the abnormal data, determining the learning effect value (prediction value) corresponding to the abnormal data based on the target sample abnormal learning data corresponding to the abnormal data and the corresponding sample learning result;

[0087] Based on the learning effect values ​​(prediction values) corresponding to all the abnormal data, determine whether there are one or more target abnormal data (labels) that meet the learning effect requirements in all the abnormal data;

[0088] When one or more of the target abnormal data (labels) exist in all the abnormal data, it is determined that the deep confidence network evaluation model of the abnormal state of the power system meets the sample learning termination requirements; when the target abnormal data (label) does not exist in all the abnormal data, it is determined that the initialized deep confidence network evaluation model does not meet the sample learning termination requirements.

[0089] The specific steps of updating the network model weights are as follows: extracting the target sample learning results (predicted values) corresponding to the target abnormal data from the sample learning results corresponding to all the abnormal data; determining the target network weight information based on the target abnormal data and the target sample learning results; and generating a new network evaluation model for updating the network weight parameters based on the target network weight information and the network evaluation model.

[0090] When the deep confidence network evaluation model does not meet the sample learning termination requirements, based on the sample learning results corresponding to all the abnormal data and the set network weight information update requirements, determine the weight information to be updated corresponding to the network model; based on the weight information to be updated, the sample learning results corresponding to all the abnormal data and the set network weight information optimization requirements, determine the weight information optimization state corresponding to the weight information to be updated; based on the weight information optimization state, perform weight information optimization operations on the deep confidence network evaluation model to generate a network model after weight information optimization, and based on the deep confidence network evaluation model after weight information optimization, load the power system electrical characteristic monitoring flow data into the optimized network evaluation model, generate multiple sample learning results corresponding to the abnormal data, and re-judge whether the sample learning termination requirements are met.

[0091] Once the network model is successfully generated and meets the sample learning termination requirements, the deep confidence network evaluation model is deployed to the corresponding power system operation area, and the power system electrical characteristic monitoring flow data is loaded into the successfully updated deep confidence network evaluation model to evaluate the operating waveform of the power grid.

[0092] In this embodiment, the deep belief network model uses the latest collected operating electrical characteristic data of the power system to evaluate the system operating status in real time. The model will automatically determine the current operating status of the operating area based on real-time electrical parameter information, such as voltage value, current frequency waveform and other characteristic data. When the operating status sends an abnormality, it determines the type of abnormality and determines whether to further update the protection parameter configuration (such as the attached Figure 4 shown).

[0093] The automatic point calibration module is used to adjust the operating parameters of the relay according to the fault type identification results, including the setting of the trigger time and protection threshold, match the relay response with the actual power grid state, restore the power grid to normal operation, and generate the adjusted protection parameter configuration (such as the attached Figure 5 shown).

[0094] In this embodiment, the automatic point check module analyzes the impact of the fault on the operation state of the power grid, including frequency anomaly or voltage drop, based on the fault type identification result, and sets the initial value W of the protection threshold according to the impact. The initial setting is performed based on the fault severity and historical data statistics. The calculation formula is:

[0095] W=u*L

[0096] In the above formula, u is an adjustment coefficient related to the fault type, and L is a statistic of the degree of fault impact; the above parameters need to be determined based on actual data statistics. Among them, the u value is mainly used to preliminarily set the protection threshold according to the fault type; it reflects the degree of impact of the fault type on the power grid, which needs to be determined based on actual data statistics and experience. For example, for severe faults, u may be set higher to increase the protection threshold, while for minor faults, u may be set lower; here is an example to illustrate that if the fault type is high-risk such as line break, it can be set to a higher value (such as 1.2), and low-risk faults such as small-scale short circuits can be set to a lower value (such as 0.8). The statistic L can be obtained by analyzing the actual impact of similar faults on the power grid in the past. For example: based on historical data, calculate the average voltage drop or frequency fluctuation caused by the past 5 similar faults.

[0097] The trigger time is set according to the current load of the power grid and the urgency of fault recovery, and is calculated using a dynamic adjustment formula, which is:

[0098]

[0099] Where T is the adjusted trigger time; Tbase is the basic trigger time of the relay under standard conditions; k is the adjustment coefficient, which is used to adjust the trigger time according to the grid load; P is the actual load of the grid at present; P 额定 is the rated load of the power grid, which serves as a reference benchmark; e is the base of the natural logarithm, which is approximately 2.71828; a is the waveform adjustment coefficient, which can be used to control the impact of load changes on the k value; b is the offset, which is used to adjust the reference point of the load; E is the fault urgency index, which can be a quantitative representation of the fault type, fault duration or fault impact range. Its value is usually determined according to the severity and urgency of the fault. Here is an example value: short circuit fault: E = 0.9 (very urgent) overload fault: E = 0.5 (medium urgency) ground fault: E = 0.7 (relatively urgent) overvoltage / undervoltage fault: E = 0.3 (lower urgency) harmonic interference: E = 0.2 (general urgency) contactor or relay fault: E = 0.4 (medium urgency). In actual applications, the actual value can be determined by expert systems, historical data analysis or real-time monitoring data in actual applications. For example, if a fault causes a large number of users to lose power, then E may be set to a higher value. If the failure occurred during a low-load period and affected only a few users, then E might be set to a lower value.

[0100] As P increases, the value of k increases according to the value of a, reflecting the need for more sensitive trigger time adjustment when the grid load increases. b acts as an offset to adjust the reference point of the load so that the change in k is smaller below a certain load value. P rated as the denominator ensures that the value of k is proportional to the rated load of the grid, so that the change in k is related to the design capacity of the grid.

[0101] The adjusted protection parameter configuration is output in combination with the adjusted protection threshold W and trigger time T, and the protection parameter configuration is applied to the relay, the response matches the grid state, and the grid is restored to a normal operating state.

[0102] This embodiment also provides a method for automatic verification of relay protection in a power system, which is applied to the above-mentioned automatic verification system for relay protection in a power system. The method comprises the following steps:

[0103] S1, the electrical characteristics collection module collects the current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time label and amplitude value of each waveform, and stores the data as a time series to generate real-time waveform records of the power grid;

[0104] S2. Calculate the change value based on the current and voltage data at two adjacent moments; specifically including:

[0105] The collected data is preprocessed, where the preprocessing steps include: denoising (using filters to remove noise from the data), calibration (ensuring that the measured values ​​are accurate and consistent with the actual values), etc., and calculating the change value of the current and voltage waveforms and the waveform adjustment coefficient based on the current and voltage data at two adjacent moments;

[0106] The change value includes instantaneous change value, cumulative change value, and percentage change value;

[0107] The instantaneous change value is the difference between the current or voltage values ​​at two or more consecutive time points. The expression is as follows:

[0108] ΔI=I(tn)-I(tn-1)

[0109] ΔV=V(tn)-V(tn-1)

[0110] Where I(tn) and V(tn) are the current and voltage values ​​at the nth time point, respectively;

[0111] The cumulative change value is the integration of the change of current or voltage over time to calculate the total change over a period of time. The expression is as follows:

[0112]

[0113] The percentage change value is the percentage of the calculated current or voltage change, and the expression is as follows:

[0114]

[0115] In the above formula, final represents the final value and init represents the initial value.

[0116] The waveform adjustment coefficient is a dynamic adjustment of the time axis of the real-time waveform data, including scaling and translation, to align with the standard waveform template. The expression is as follows:

[0117] Smodel=Sreal*a+b

[0118] Smodel is the standard waveform template, Sreal is the real-time waveform data, a and b are adjustment coefficients, which control the scaling and translation respectively.

[0119] S3. Extract characteristic data that is helpful for fault detection from the preprocessed data, such as mutations, sudden drops, etc. For example, a sudden change in current or a sudden drop in voltage will cause a significant increase in the instantaneous change value. If the sudden change in current or voltage is continuous, the cumulative change value will also show an abnormal pattern and may accumulate to a large value in a very short time. A sudden change in current or voltage may cause a significant increase in the percentage change value.

[0120] S4. According to the existing fault data, the change value is marked with normal as 0 and abnormal as 1, and the fault type label is marked. The fault type label includes: short circuit fault, overload fault, ground fault, overvoltage / undervoltage fault, harmonic interference, contactor or relay fault, etc. The corresponding data features are marked according to the abnormal data generated by the corresponding fault type, and then the data set is divided into a training set and a test set.

[0121] S5. Use the training set data to train the deep belief network model to learn the characteristics of normal and abnormal data, and use the test set to evaluate the accuracy and effect of the model. Adjust the model parameters according to the evaluation results to improve the prediction accuracy.

[0122] S6. Use the trained model and the latest collected electrical characteristic data to monitor the operation status of the power grid. If an abnormal fault is found, determine the type of abnormal fault;

[0123] S7, recording the abnormal waveform state of the power grid when an abnormal event occurs, and updating the fixed parameters and protection thresholds of the relay according to the waveform state;

[0124] S8. Regularly retrain the model using the latest collected data, and use the newly trained model to monitor the operating status of the power grid.

[0125] In summary, the present invention provides an automatic verification system and method for relay protection in a power system, which prevents power outages in the power system due to abnormal identification of relays, thereby causing equipment damage and reducing property losses, through accurate real-time monitoring of the power grid status, rapid response to fault identification, and timely parameter verification. This automated verification system and method can greatly improve the efficiency and accuracy of power grid management, reduce manual intervention, reduce operating costs, and improve the stability and reliability of the power system.

[0126] Secondly, in the field of artificial intelligence technology, the present invention uses a deep learning neural network model for data analysis and feature extraction. By learning a large amount of historical data, it can accurately evaluate the type of anomalies and accurately verify the relay configuration parameters. This deep learning-based relay protection automatic verification system and method can improve the accuracy and efficiency of parameter configuration, provide a more intelligent solution for power grid management, and promote the application and development of artificial intelligence technology in the field of power systems.

[0127] Finally, in the field of automatic control technology, the present invention can significantly improve the reliability and safety of the system through automatic point calibration, and reduce equipment damage or power outages caused by improper relay settings. This automated control method can achieve self-regulation and optimization of the power system, improve the stability and reliability of the system, and promote the application and development of automatic control technology in the field of power systems.

[0128] The above descriptions are only some embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Any equivalent device or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An automatic verification system for relay protection of power system, It is characterized in that include: Electrical feature collection module, used to obtain key electrical data of the power system in real time; A data processing module is used to pre-process the collected raw data and extract features that are helpful for fault detection; The deep belief network model is used to determine whether the grid operation status is abnormal based on the electrical data flow data of the grid, and immediately determine the abnormal fault type if an abnormality is found; The automatic point verification module is used to adjust the operating parameters of the relay according to the fault type identification results, match the relay response with the actual power grid status, restore the power grid to normal operation, and generate the adjusted protection parameter configuration.

2. The automatic verification system for relay protection of a power system according to claim 1, It is characterized in that The electrical characteristics collection module collects the current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time label and amplitude value of each waveform, and stores the data as a time series to generate a real-time waveform record of the power grid; In the real-time monitoring of current and voltage data, each collected data point records the amplitude and time information at the same time. Each waveform data point is equipped with a time tag and a corresponding amplitude value. The time and amplitude of data collection are recorded synchronously by the timer and sensor. The time and amplitude of each data point are integrated to generate a time-amplitude pairing, which is expressed as follows: P(t i )=(t i ,V i ) P(t i ) represents the time point t i Paired data points of t i is the time tag recorded by the data acquisition timer, indicating the sampling time of the i-th data point; V i At time point t i The collected amplitude value, i.e. the instantaneous value of current or voltage; The integrated time-amplitude paired data are stored in a cloud database.

3. The automatic verification system for relay protection of a power system according to claim 1, It is characterized in that The data processing module preprocesses the data stored in the cloud database, then performs normalization processing, and marks the data according to known abnormal events to generate data labels, and then divides the data into a training set and a test set after preprocessing.

4. The automatic verification system for relay protection of a power system according to claim 3, It is characterized in that The preprocessing includes cleaning, removing or filling missing values, deleting outliers or erroneous data points, and extracting features from the original data; The features extracted from the original data include the normal interval of the current and voltage waveform characteristics in the electrical data and the abnormal current and voltage values ​​during short circuit.

5. The automatic verification system for relay protection of a power system according to claim 4, It is characterized in that The deep belief network model is trained by preprocessed historical electrical data until the model converges; The historical electrical data includes the current, voltage and frequency variation range during daily operation and any known abnormal short circuit events.

6. The automatic verification system for relay protection of a power system according to claim 5, It is characterized in that The deep belief network model uses the latest collected operating electrical characteristic data of the power system to evaluate the system operating status in real time. The model will automatically determine the current operating status of the operating area based on the real-time electrical parameter information. When the operating status sends an abnormality, it will determine the type of abnormality and determine whether to further update the protection parameter configuration.

7. The automatic verification system for relay protection of a power system according to claim 1, It is characterized in that The automatic point calibration module adjusts the operating parameters of the relay including the setting of the trigger time and the protection threshold; The automatic point verification module analyzes the impact of the fault on the operation status of the power grid according to the fault type identification result, sets the initial value W of the protection threshold according to the impact, and performs preliminary setting according to the fault severity and historical data statistics. The calculation formula is: W=u*L Among them, u is an adjustment coefficient related to the fault type, and L is a statistic of the fault impact; The trigger time is set according to the current load of the power grid and the urgency of fault recovery, and is calculated using a dynamic adjustment formula, which is: Where T is the adjusted trigger time; T 基础 is the basic triggering time of the relay under standard conditions; k is the adjustment coefficient, which is used to adjust the triggering time according to the grid load; P is the actual load of the grid at present; P 额定 is the rated load of the power grid, which is used as a reference benchmark; e is the base of the natural logarithm; a is the waveform adjustment coefficient, which can be used to control the impact of load changes on the k value; b is the offset, which is used to adjust the reference point of the load; E is the fault urgency index, which can be a quantitative representation of the fault type, fault duration or fault impact range; The adjusted protection parameter configuration is output in combination with the adjusted protection threshold W and trigger time T, and the protection parameter configuration is applied to the relay, the response matches the grid state, and the grid is restored to a normal operating state.

8. A method for automatic verification of relay protection in a power system. It is characterized in that Applied in the power system relay protection automatic verification system as claimed in any one of claims 1 to 7, the method comprises the following steps: S1, the electrical characteristics collection module collects the current and voltage waveform data of the power grid by real-time monitoring of the power grid, records the time label and amplitude value of each waveform, and stores the data as a time series to generate real-time waveform records of the power grid; S2. Calculate the change value based on the current and voltage data at two adjacent moments; S3, extracting feature data that is helpful for fault detection from the preprocessed data; S4. According to the existing fault data, the change values ​​are marked with normal as 0 and abnormal as 1, and the fault type labels are annotated, and then the data set is divided into a training set and a test set; S5. Use the training set data to train the deep belief network model, learn the characteristics of normal and abnormal data, and use the test set to evaluate the accuracy and effect of the model; S6. Use the trained model and the latest collected electrical characteristic data to monitor the operation status of the power grid. If an abnormal fault is found, determine the type of abnormal fault; S7, recording the abnormal waveform state of the power grid when an abnormal event occurs, and updating the fixed parameters and protection thresholds of the relay according to the waveform state; S8. Regularly retrain the model using the latest collected data, and use the newly trained model to monitor the operating status of the power grid.

9. The automatic verification method for relay protection of a power system according to claim 8, It is characterized in that Step S2 specifically includes: Preprocessing the collected data, wherein the preprocessing steps include: denoising, calibration, and calculating the change value of the current and voltage waveforms and the waveform adjustment coefficient according to the current and voltage data at two adjacent moments; The change value includes an instantaneous change value, a cumulative change value, and a percentage change value; The instantaneous change value is calculated by calculating the difference between the current or voltage values ​​at two or more consecutive time points. The expression is as follows: ΔI=I(tn)-I(tn-1) ΔV=V(tn)-V(tn-1) Where I(tn) and V(tn) are the current and voltage values ​​at the nth time point, respectively; The cumulative change value is the integration of the change of current or voltage over time to calculate the total change over a period of time. The expression is as follows: The percentage change value is the percentage of the calculated current or voltage change, and the expression is as follows: The waveform adjustment coefficient is a dynamic adjustment of the time axis of the real-time waveform data, including scaling and translation, to align with the standard waveform template. The expression is as follows: Smodel=Sreal*a+b Smodel is the standard waveform template, Sreal is the real-time waveform data, a and b are adjustment coefficients, which control the scaling and translation respectively.

10. The automatic verification method for relay protection of a power system according to claim 8, It is characterized in that The features that are helpful for fault detection in step S3 include mutation and sudden drop; The fault type labels in step S4 include: short circuit fault, overload fault, ground fault, overvoltage / undervoltage fault, harmonic interference, contactor or relay fault.

Images