Power grid anomaly detection system and method based on artificial intelligence

Abstract

The invention discloses a power grid anomaly detection system and method based on artificial intelligence, relates to the technical field of power grid detection, and solves the technical problems of low recognition precision and insufficient fault positioning capability generally existing in the existing power grid anomaly detection technology. The method comprises the following steps: analyzing an acquisition period of operation data of power grid equipment in different operation periods; based on historical fault information, analyzing the fault probability of each fault type caused by each operation data of the power grid equipment in different operation periods; analyzing contribution coefficients of the operation data to different fault types in different operation periods based on the fault probability; calculating a fault index of the power grid equipment based on the contribution coefficient; determining the fault type of the power grid equipment based on the fault index; and analyzing the fault level of the power grid equipment based on the fault type and the fault index. According to the method, the fault type identification accuracy is remarkably improved, and powerful support is provided for intelligent operation and maintenance management of a power grid.

Description

Technical Field

[0001] The present invention belongs to the field of power grid detection, and specifically relates to a power grid anomaly detection system and method based on artificial intelligence. Background Art

[0002] During the operation of the power grid, with the continuous expansion and complexity of the power system, abnormal events such as short circuits, circuit breaks, equipment failures, overloads, etc. occur frequently, posing huge challenges to the safe and stable operation of the power grid.

[0003] Traditional power grid anomaly detection uses threshold settings, logical analysis, and historical data analysis to identify potential risks or faults. However, power grid anomalies often exhibit complex, nonlinear, and uncertain characteristics. For example, voltage sags can be caused by a variety of factors, including lightning strikes, line short circuits, and sudden load changes. Traditional methods lack the ability to analyze these complex characteristics, making it difficult to accurately identify specific fault types and significantly complicating maintenance efforts. Therefore, the present invention provides an artificial intelligence-based power grid anomaly detection system and method. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an artificial intelligence-based power grid anomaly detection system and method, which are used to solve the technical problems of low recognition accuracy and insufficient fault location capability commonly existing in existing power grid anomaly detection technologies.

[0005] To achieve the above-mentioned object, a first aspect of the present invention provides an artificial intelligence-based power grid anomaly detection system, comprising an acquisition module, a fault analysis module, and an alarm module;

[0006] Acquisition module: used to analyze the collection cycle of operation data of power grid equipment at different operation periods; wherein the operation data is the state parameters of the power grid equipment during operation;

[0007] Fault analysis module: used to analyze the fault probability of each fault type caused by each operating data of power grid equipment in different operating periods based on historical fault information; and analyze the contribution coefficient of operating data to different fault types in different operating periods based on the fault probability;

[0008] Calculating a fault index of the power grid device based on the contribution coefficient; determining a fault type of the power grid device based on the fault index; and,

[0009] Analyze the fault level of power grid equipment based on fault type and fault index.

[0010] Preferably, the collection period of the operating data is obtained by the following methods, including:

[0011] Extracting a number of operating data of power grid equipment from historical data, and classifying the operating data according to different operating periods; wherein the operating period includes a peak electricity consumption period during severe weather, a low electricity consumption period during severe weather, a peak electricity consumption period during non-severe weather, and a low electricity consumption period during non-severe weather; the time period length of the operating data extracted in each operating period is the same;

[0012] Calculate the abnormal frequency of each operating data in each operating period, perform average processing on several abnormal frequencies of the same operating data in the same operating period, and obtain the collection period of each operating data in each operating period; among which, the abnormal frequency is the frequency of occurrence of abnormal operating data; abnormal operating data is the operating data that is not within the preset range.

[0013] By categorizing operational data into four categories, namely, inclement weather / non-inclement weather and peak / off-peak periods, this method can more meticulously reflect the actual operating status of equipment under different environmental and load conditions. This avoids the potential bias caused by traditional holistic analysis. For example, equipment is more likely to exhibit abnormalities during inclement weather, making it difficult to identify abnormal patterns in specific periods through a unified analysis. Furthermore, the data extracted from each operating period has the same time period length, eliminating data volume bias caused by differences in time spans and making the calculation of abnormality frequency more consistent and comparable.

[0014] Preferably, the analyzing the failure probability of each failure type caused by each operation data of the power grid equipment in different operation periods includes:

[0015] Extracting fault information of power grid equipment from historical fault data; the fault information includes fault type k, operation period j, and abnormal operation data i; k is the label of the fault type, j is the label of the operation period, and i is the label of the operation data;

[0016] Calculate the ratio of the historical number of fault type k caused by abnormal operating data i during operation period j to the total number of historical faults to obtain the fault probability of fault type k caused by abnormal operating data i during operation period j. The total number of historical faults is the total number of historical times of each type of fault caused by abnormal operating data i during operation period j.

[0017] This method avoids the errors caused by traditional "one-size-fits-all" analysis by identifying the specific impact of a certain type of abnormal data (such as voltage over-limit) on a certain type of fault (such as tripping) under specific operating conditions (such as peak power consumption during inclement weather). Specifically, by statistically analyzing the number of historical faults and calculating their ratios, it converts abstract risks into specific fault probability values. It can also reveal that certain fault types are only prevalent during specific operating periods. Furthermore, by analyzing the correlation frequency between abnormal operating data and fault types, it can reveal potential fault-causing mechanisms.

[0018] Preferably, the analysis of contribution coefficients of operating data in different operating periods to different fault types includes:

[0019] Extracting a number of abnormal operation data i during operation period j from historical data, and dividing the abnormal operation data i during operation period j into a number of abnormal operation data sequences, wherein the number of data in each abnormal operation data sequence is the same;

[0020] and extracting a number of fault probabilities of fault type k caused by abnormal operation data i in operation period j, and dividing them into a number of fault probability sequences Pki;

[0021] By the formula Rki=|[(Xi-EXi)×(Pki-EPki)] / [sqrt(Xi-EXi) 2 ×sqrt(Yi-EYi) 2 ]|Calculate the correlation coefficient Ri between each operating data and each fault type at different operating periods; where Xi is the abnormal operating data sequence of abnormal operating data i, EXi is the mean of Xi, Pki is the fault probability sequence of fault type k caused by abnormal operating data i, EPkij is the mean of Pki; sqrt is the square root;

[0022] Based on several abnormal operation data sequences and corresponding several fault probability sequences of operation data i, several correlation coefficients Rij are calculated and the correlation coefficients Rij are averaged to obtain the contribution coefficient of operation data i to fault type k in operation period j.

[0023] Based on the abnormal operation data sequence Xi and the corresponding fault probability sequence Pki constructed above, the present invention quantitatively evaluates the degree of linear correlation between the two, and finally obtains the correlation coefficient Rki of each operation data i for each type of fault k in each operation period j. The coefficient values ​​of multiple operation periods are further averaged to obtain the contribution coefficient, which represents the comprehensive influence of the operation data on a certain type of fault. It can accurately identify which abnormal operation data has a significant impact on a specific fault type in which operation period, which helps to deeply understand the mechanism of power grid equipment failure.

[0024] Preferably, analyzing the fault type of the power grid equipment includes:

[0025] Compare the real-time collected operating data with the corresponding preset range to extract abnormal operating data; calculate the difference between the abnormal data and the near end point of the corresponding preset range, and multiply the difference by the contribution coefficient of the operating data to obtain the fault index of the power grid equipment; where the near end point is the left end point or the right end point of the preset range;

[0026] The fault index is compared with the fault threshold corresponding to the fault type in the same operation period in sequence, and the fault threshold that is less than or equal to the fault index is extracted; wherein the fault type corresponding to the largest fault threshold among the extracted fault thresholds is the fault type of the current power grid equipment.

[0027] The present invention can detect abnormal situations in a timely manner by collecting operating data in real time and comparing it with a preset range, realize dynamic monitoring of the operating status of power grid equipment, help to discover potential fault hazards in advance, avoid fault expansion, and ensure the stable operation of the power grid. The difference between the abnormal data and the near end point of the preset range is calculated and multiplied by the contribution coefficient of the operating data to obtain the fault index. This quantitative method can more accurately assess the severity of the fault. Different operating data have different degrees of influence on equipment operation. The introduction of the contribution coefficient fully takes this into account, making the fault index more scientific and accurate, and providing a reliable quantitative basis for subsequent fault judgment and processing; by selecting the fault type corresponding to the maximum fault threshold, it is possible to accurately identify the fault that currently poses the greatest threat to the operation of power grid equipment, which helps maintenance personnel quickly focus on the most critical issues and avoid being disturbed by some relatively minor faults, so as to take effective countermeasures in time to prevent the fault from further deteriorating.

[0028] Preferably, the method for obtaining the proximal endpoint includes:

[0029] Calculate the difference between the abnormal data and each endpoint in the preset range, and determine whether the difference between the abnormal data and each endpoint is less than the preset difference; if so, mark the endpoint corresponding to the smallest difference as the near endpoint; if not, do not mark it.

[0030] Preferably, the method for obtaining the fault threshold includes:

[0031] Extract several fault information of fault type k from historical fault data, calculate the fault index mean of fault type k, and obtain the fault threshold of fault type k.

[0032] Preferably, analyzing the fault level of the power grid equipment includes:

[0033] The fault index and fault type are input into the level assessment model, and the desensitization model outputs a fault level label; wherein the fault level label matches the preset fault level;

[0034] The grade assessment model is constructed by an artificial intelligence model, and the construction process is as follows:

[0035] Extract the fault index, fault type and fault level of several power grid devices from historical fault data;

[0036] Integrate the fault index, fault type and fault level labels into several groups of training data and test data; use the training data to train the artificial intelligence model; use the test data to test the trained artificial intelligence model, and adjust the artificial intelligence model according to the test results; finally obtain a level assessment model with the fault index and fault type as input and the fault level as output; wherein the artificial intelligence model is a BP neural network model or an RBF neural network model.

[0037] Preferably, the alarm module emits sound and light warnings of different colors according to the fault level.

[0038] This invention integrates both the fault index and fault type into a fault severity assessment model, comprehensively considering multiple key factors influencing fault severity. The fault index reflects the degree to which abnormal data deviates from a preset range, while the fault type identifies the specific manifestation of the fault. Combining these two provides a more comprehensive and accurate picture of the fault status of power grid equipment, resulting in a more realistic fault severity label and improved risk assessment accuracy.

[0039] A second aspect of the present invention provides an artificial intelligence-based power grid anomaly detection system, comprising:

[0040] Analyze the collection cycle of operating data of power grid equipment in different operating periods;

[0041] Based on historical fault information, analyze the fault probability of each fault type caused by each operating data of power grid equipment in different operating periods;

[0042] Analyze the contribution coefficient of operation data to different fault types in different operation periods based on the failure probability;

[0043] Calculate the fault index of power grid equipment based on the contribution coefficient;

[0044] Determine the fault type of the power grid equipment based on the fault index;

[0045] Analyze the fault level of power grid equipment based on fault type and fault index.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention determines the collection cycle for each operating data point during different operating periods and can flexibly adjust it based on the actual operating conditions of the equipment. For example, during peak power consumption periods, the equipment load increases significantly and the operating status changes rapidly. In this case, the collection cycle can be shortened to promptly capture subtle changes in equipment parameters and accurately reflect the real-time operating status of the equipment. During low power consumption periods, the equipment load is low and operation is relatively stable. Appropriately extending the collection cycle can ensure monitoring of the equipment status while reducing unnecessary data collection and storage. Furthermore, based on historical fault information, the probability of each operating data point causing each fault type during different operating periods can be analyzed, providing a deep understanding of the relationship between operating data and fault types, providing a strong basis for fault prediction and prevention. Determining the contribution coefficient based on the fault probability can quantify the degree of influence of different operating data on different fault types, laying the foundation for accurately analyzing the fault index of power grid equipment, improving the accuracy of fault judgment, and locating the fault type. The present invention helps to improve the operation and maintenance management level of power grid equipment, ensure the safe and stable operation of the power grid, and reduce the economic losses and social impact caused by equipment failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. 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 any creative work.

[0049] Figure 1 Schematic diagram of the contribution coefficient analysis method of the present invention;

[0050] Figure 2 This is a flow chart of a method for analyzing power grid equipment fault types according to the present invention;

[0051] Figure 3 Schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0052] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0053] The first embodiment of the present invention provides a mixed plastic sorting system based on infrared spectroscopy, comprising a collection module, a fault analysis module and an alarm module;

[0054] The acquisition module analyzes the collection cycle of the operating data of the power grid equipment at different operating periods;

[0055] Among them, the operating data refers to the state parameters of the power grid equipment during operation, such as current, voltage, resistance, temperature, power, etc.

[0056] Specifically, a plurality of operating data of power grid equipment is extracted from historical data, and the plurality of operating data are classified according to different operating periods; wherein the operating periods include a peak electricity consumption period during severe weather, a low electricity consumption period during severe weather, a peak electricity consumption period during non-severe weather, and a low electricity consumption period during non-severe weather; and the time period length of the operating data extracted in each operating period is the same;

[0057] It should be noted that severe weather includes typhoons, heavy rain, ice and snow, thunderstorms, etc.; non-severe weather includes weather other than severe weather; peak and low electricity consumption periods are divided by electricity load conditions and can be determined based on the electricity load data recorded by the power grid company;

[0058] The abnormal frequency of each operating data in each operating period is calculated using the formula abnormal frequency = (number of times abnormal operating data occurs / historical analysis period). Several abnormal frequencies of the same operating data in the same operating period are averaged to obtain the collection period of each operating data in each operating period. Among them, abnormal operating data refers to operating data that is not within the preset range; the analysis period is the period length of the collected historical operating data.

[0059] It should be noted that when the power grid equipment operates normally, the variation range of each operating data is within the corresponding preset range.

[0060] For example, consider the main transformer of a 110kV substation. The operating data collected per unit time for several consecutive months include load factor (normal range: 30% to 90%), top oil temperature (normal range: ≤85°C), and winding temperature (normal range: ≤95°C).

[0061] The following embodiment takes the load rate as an example to analyze its sampling period;

[0062] The extracted historical load rate data are divided into four operating periods according to weather and power load conditions. The historical time period extracted for each operating period is the same (30 days).

[0063] Assume that the peak hours of electricity consumption are 6:00-13:00 and 17:00-22:00, and the low hours of electricity consumption are 24:00-6:00 and 13:00-17:00;

[0064] For example, to count the number of abnormal load rates during peak electricity consumption periods in severe weather, the load rate collection period = Σ(number of abnormal load rates during peak electricity consumption periods in severe weather x / 30) / N; the unit of the analysis period can be modified according to the situation; x is the number of the extracted historical time period, x = 0, 1, ..., N, N is a positive integer, and Σ sums x.

[0065] The fault analysis module analyzes the fault type of the power grid equipment;

[0066] Specifically, based on historical fault information, the fault probability of each fault type caused by each operating data of the power grid equipment in different operating periods is analyzed;

[0067] Extracting fault information of power grid equipment from historical fault data; wherein the fault information includes the fault type, marked as k, the operation period, marked as j, and the abnormal operation data, marked as i; and k, j, and i can all be set as natural number labels;

[0068] The probability of fault type k occurring in power grid equipment due to abnormal operating data i during operation period j is calculated using the formula P(k|i,j) = historical number of fault type k caused by abnormal operating data i during operation period j / total number of historical faults caused by abnormal operating data i during operation period j. The total number of historical faults is the historical sum of the number of each type of fault caused by abnormal operating data i during operation period j.

[0069] It should be noted that the number of historical faults is the number of times each type of fault occurs in a period of extracted historical data, and the extracted historical data cycle is long enough.

[0070] For example, suppose you want to extract transformer failure records from a power grid company over the past year, including the following data:

[0071] Fault type (k): k = 1: winding short circuit, k = 2: insulation breakdown, k = 3: cooling system failure;

[0072] Operation period (j): j = 1: peak electricity consumption period in bad weather, j = 2: peak electricity consumption period in non-bad weather;

[0073] Abnormal operating data (i): i=1: current imbalance > 10%, i=2: oil temperature > 85°C, i=3: partial discharge pulse > 20 times / hour;

[0074] The following example analyzes the probability of various fault types caused by current imbalance in transformers during peak power consumption periods in severe weather;

[0075] Assume that the total number of historical faults is the total number of times k=1, k=2, and k=3 occurred. If the total number of historical faults = 8+5+3 = 16 times;

[0076] The probability of the fault type k=1 caused by the current imbalance is P(k=1|i=1,j=1)=8 / 16≈50%;

[0077] The probability of the fault type k=2 caused by the current imbalance is P(k=2|i=1,j=1)=5 / 16≈31%;

[0078] The probability of the fault type k=3 caused by the current unbalance is P(k=3|i=1,j=1)=3 / 16≈19%.

[0079] See also Figure 1 ,and ,analyze the contribution coefficients of operation data in different ,operation periods to different fault types;

[0080] Specifically, a number of abnormal operation data i in operation period j is extracted from the historical data, and the abnormal operation data i in operation period j is divided into a number of abnormal operation data sequences, and the number of data in each abnormal operation data sequence is the same;

[0081] Extract several fault probabilities of fault type k caused by abnormal operation data i in operation period j, and divide them into several fault probability sequences Pki;

[0082] By the formula Rki=|[(Xi-EXi)×(Pki-EPki)] / [sqrt(Xi-EXi) 2 ×sqrt(Yi-EYi) 2 ]|Calculate the correlation coefficient Rki between each operating data and each fault type at different operating periods; where Xi is the abnormal operating data sequence of abnormal operating data i, EXi is the mean of Xi, Pki is the fault probability sequence of fault type k caused by abnormal operating data i, EPkij is the mean of Pki; sqrt is the square root;

[0083] It should be noted that this formula, a method based on covariance and standard deviation, effectively measures the strength of the linear relationship between two variables, taking into account not only the numerical changes but also the consistency direction of such changes.

[0084] Based on several abnormal operation data sequences Xi and several corresponding fault probability sequences Pki, several correlation coefficients Rki of abnormal operation data i causing fault type k are calculated, and the correlation coefficients Rki are averaged to obtain the contribution coefficient of operation data i to fault type k in operation period j.

[0085] For example, based on the above embodiment, an abnormal operation data sequence and a fault probability sequence are constructed. For example, 10 pieces of abnormal operation data i=1 in the operation period j=1 are extracted to form an abnormal operation data sequence of abnormal operation data i=1, such as X1=[12%, 15%, 11%, 14%, 13%, 16%, 12%, 15%, 14%, 13%]; and 10 fault probabilities P(k=1|i=1, j=1) of i=1 are constructed, such as P11=[0.50, 0.48, 0.52, 0.49, 0.51, 0.47, 0.53, 0.50, 0.49, 0.51];

[0086] Calculate EX1=13.5, EP11=0.50;

[0087] [(Xi-EXi)(Pki-EPki)]=(12-13.5)(0.50-0.50)+(15-13.5)(0.48-0.50)+…=-0.03;

[0088] (Xi-EXi) 2 =(12-13.5)²+(15-13.5)²+…=3.72;

[0089] (Pki-EPki) 2 =(0.50-0.50)²+(0.48-0.50)²+…=0.03;

[0090] Correlation coefficient R11 = |-0.03 / (3.72×0.03)|≈0.27;

[0091] According to this method, using some historical data, several correlation coefficients between operating data i=1 and fault type k=1 are calculated, and the mean of the correlation coefficients is taken to obtain the contribution coefficient of operating data i=1 to fault type k=1 in operating period j=1.

[0092] See also Figure 2 , compare the real-time collected operating data with the corresponding preset range to extract abnormal operating data; calculate the difference between the abnormal operating data and the near end point of the corresponding preset range, and multiply the difference by the contribution coefficient of the operating data to obtain the fault index of the power grid equipment;

[0093] The near endpoint is the left endpoint or the right endpoint of the preset range, and the near endpoint is obtained as follows:

[0094] Calculate the difference between the abnormal data and each endpoint of the preset range, and determine whether the difference between the abnormal data and the endpoint is less than the preset difference; if so, mark the endpoint as a near endpoint; if not, do not mark it.

[0095] Compare the fault index with the fault thresholds corresponding to the fault types in the same operating period in sequence, and extract the fault thresholds that are less than or equal to the fault index;

[0096] The fault type corresponding to the extracted maximum fault threshold is the fault type of the current power grid equipment.

[0097] For example, based on the above embodiment, assuming that the fault threshold corresponding to k=1 is A, the fault threshold corresponding to k=2 is B, the fault threshold corresponding to k=3 is C, and A>B>C;

[0098] Now compare the fault index D of the transformer equipment obtained by real-time analysis with A, B, and C one by one. Assume that D<A,D> B, D>C, and C>B, then the fault type of the transformer equipment is C, and the corresponding fault type k=3.

[0099] It should be noted that if the minimum satisfied threshold is directly selected, it may be misjudged as a minor fault. The present invention prioritizes the unsatisfied fault types to ensure that the type closest to the current fault index is selected.

[0100] And, the method for obtaining the fault threshold is as follows:

[0101] Extract several fault information of fault type k from historical fault data, calculate the fault index mean of fault type k, and obtain the fault threshold of fault type k.

[0102] The fault index and fault type are input into the level assessment model, and the desensitization model outputs a fault level label; wherein the fault level label matches the preset fault level;

[0103] Among them, the grade assessment model is constructed by the artificial intelligence model, and the construction process is as follows:

[0104] Extract the fault index, fault type and fault level of several power grid devices from historical fault data;

[0105] Integrate the fault index, fault type and fault level labels into several groups of training data and test data; use the training data to train the artificial intelligence model; use the test data to test the trained artificial intelligence model, and adjust the artificial intelligence model according to the test results; finally obtain a level assessment model with the fault index and fault type as input and the fault level as output; wherein the artificial intelligence model is a BP neural network model or an RBF neural network model.

[0106] In addition, the alarm module emits sound and light warnings of different colors according to the fault level.

[0107] For example, suppose the fault levels are set as 1. Level 1 alarm (minor fault), Level 2 alarm (moderate fault), and Level 3 alarm (serious fault).

[0108] Level 1 alarm (minor fault): sound and light alarm (yellow warning);

[0109] Level 2 alarm (medium fault): sound and light alarm (orange warning);

[0110] Level 3 alarm (serious fault): high-frequency sound and light alarm (red warning).

[0111] See also Figure 3 The second aspect of the present invention provides an artificial intelligence-based power grid anomaly detection system, comprising:

[0112] Analyze the collection cycle of operating data of power grid equipment in different operating periods;

[0113] Based on historical fault information, analyze the fault probability of each fault type caused by each operating data of power grid equipment in different operating periods;

[0114] Analyze the contribution coefficient of operation data to different fault types in different operation periods based on the failure probability;

[0115] Calculate the fault index of power grid equipment based on the contribution coefficient;

[0116] Determine the fault type of the power grid equipment based on the fault index;

[0117] Analyze the fault level of power grid equipment based on fault type and fault index.

[0118] Some of the data in the above formula are calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0119] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A power grid anomaly detection method based on artificial intelligence, characterized in that: include: Analyze the collection cycle of operating data of power grid equipment in different operating periods; Based on historical fault information, analyze the fault probability of each fault type caused by each operating data of power grid equipment in different operating periods; Analyze the contribution coefficient of operation data to different fault types in different operation periods based on the failure probability; Calculate the fault index of power grid equipment based on the contribution coefficient; Determine the fault type of the power grid equipment based on the fault index; Analyze the fault level of power grid equipment based on fault type and fault index.

2. The method for detecting power grid anomalies based on artificial intelligence according to claim 1, characterized in that: The collection period of the operating data is obtained by the following methods, including: Extracting a number of operating data of power grid equipment from historical data, and classifying the operating data according to different operating periods; wherein the operating period includes a peak electricity consumption period during severe weather, a low electricity consumption period during severe weather, a peak electricity consumption period during non-severe weather, and a low electricity consumption period during non-severe weather; the time period length of the operating data extracted in each operating period is the same; Calculate the abnormal frequency of each operating data in each operating period, perform average processing on several abnormal frequencies of the same operating data in the same operating period, and obtain the collection period of each operating data in each operating period; among which, the abnormal frequency is the frequency of occurrence of abnormal operating data; abnormal operating data is the operating data that is not within the preset range.

3. The method for detecting power grid anomalies based on artificial intelligence according to claim 1, characterized in that: The analysis of the failure probability of each fault type caused by each operating data of the power grid equipment in different operating periods includes: Extracting several fault information of power grid equipment from historical fault data; Calculate the ratio of the historical number of fault type k caused by abnormal operating data i in operation period j to the total number of historical faults to obtain the fault probability of fault type k caused by abnormal operating data i in operation period j of power grid equipment. k is the label of the fault type, j is the label of the operation period, and i is the label of the operation data. The total number of historical faults is the total number of historical times of each type of fault caused by abnormal operating data i in operation period j.

4. The method for detecting power grid anomalies based on artificial intelligence according to claim 1, characterized in that: The analysis of the contribution coefficients of the operating data to different fault types at different operating periods includes: Extracting a number of abnormal operation data i during operation period j from historical data, and dividing the abnormal operation data i during operation period j into a number of abnormal operation data sequences, wherein the number of data in each abnormal operation data sequence is the same; and extracting a number of fault probabilities of fault type k caused by abnormal operation data i in operation period j, and dividing them into a number of fault probability sequences Pki; By the formula Rki=|[(Xi-EXi)×(Pki-EPki)] / [sqrt(Xi-EXi) 2 ×sqrt(Yi-EYi) 2 ]|Calculate the correlation coefficient Ri between each operating data and each fault type at different operating periods; where Xi is the abnormal operating data sequence of abnormal operating data i, EXi is the mean of Xi, Pki is the fault probability sequence of fault type k caused by abnormal operating data i, EPkij is the mean of Pki; sqrt is the square root; Based on several abnormal operation data sequences and corresponding several fault probability sequences of operation data i, several correlation coefficients Rij are calculated and the correlation coefficients Rij are averaged to obtain the contribution coefficient of operation data i to fault type k in operation period j.

5. The power grid anomaly detection method based on artificial intelligence according to claim 1 or 4, characterized in that: The analyzing the fault type of the power grid equipment includes: Compare the real-time collected operating data with the corresponding preset range to extract abnormal operating data; calculate the difference between the abnormal data and the near end point of the corresponding preset range, and multiply the difference by the contribution coefficient of the operating data to obtain the fault index of the power grid equipment; where the near end point is the left end point or the right end point of the preset range; The fault index is compared with the fault threshold corresponding to the fault type in the same operation period in sequence, and the fault threshold that is less than or equal to the fault index is extracted; wherein the fault type corresponding to the largest fault threshold among the extracted fault thresholds is the fault type of the current power grid equipment.

6. The method for detecting power grid anomalies based on artificial intelligence according to claim 5, characterized in that: The method for obtaining the proximal endpoint includes: Calculate the difference between the abnormal data and each endpoint in the preset range, and determine whether the difference between the abnormal data and each endpoint is less than the preset difference; if so, mark the endpoint corresponding to the smallest difference as the near endpoint; if not, do not mark it.

7. The method for detecting power grid anomalies based on artificial intelligence according to claim 5, characterized in that: The method for obtaining the fault threshold includes: Extract several fault information of fault type k from historical fault data, calculate the fault index mean of fault type k, and obtain the fault threshold of fault type k.

8. The method for detecting power grid anomalies based on artificial intelligence according to claim 5, characterized in that: The analyzing the fault level of the power grid equipment includes: The fault index and fault type are input into the level assessment model, and the desensitization model outputs a fault level label; wherein the fault level label matches the preset fault level; The grade assessment model is constructed by an artificial intelligence model, and the construction process is as follows: Extract the fault index, fault type and fault level of several power grid devices from historical fault data; Integrate the fault index, fault type and fault level labels into several groups of training data and test data; use the training data to train the artificial intelligence model; use the test data to test the trained artificial intelligence model, and adjust the artificial intelligence model according to the test results; finally obtain a level assessment model with the fault index and fault type as input and the fault level as output; wherein the artificial intelligence model is a BP neural network model or an RBF neural network model.

9. An artificial intelligence-based power grid anomaly detection system, operating based on an artificial intelligence-based power grid anomaly detection method according to any one of claims 1 to 8, characterized in that: Including acquisition module and fault analysis module: Acquisition module: used to analyze the collection cycle of operation data of power grid equipment at different operation periods; wherein the operation data is the state parameters of the power grid equipment during operation; Fault analysis module: used to analyze the fault probability of each fault type caused by each operating data of power grid equipment in different operating periods based on historical fault information; and analyze the contribution coefficient of operating data to different fault types in different operating periods based on the fault probability; wherein the fault information includes fault type, operating period and abnormal operating data; Calculating a fault index of the power grid device based on the contribution coefficient; determining a fault type of the power grid device based on the fault index; and, Analyze the fault level of power grid equipment based on fault type and fault index.

10. The artificial intelligence-based power grid anomaly detection system according to claim 9, characterized in that: It also includes an alarm module, which emits sound and light warnings of different colors according to the fault level.

Images