Power distribution equipment intelligent fault early warning method and system based on Internet of Things

Abstract

The invention discloses a power distribution equipment intelligent fault early warning method and system based on the Internet of Things, and relates to the technical field of power distribution equipment fault early warning, and the method comprises the steps: collecting time sequence operation data of power distribution equipment, carrying out the cleaning and alignment, and analyzing the comprehensive characteristics of the power distribution equipment based on a sliding time window; analyzing the abnormal condition of the power distribution equipment based on the comprehensive characteristics of the power distribution equipment; determining an initial fault level according to the abnormal condition; when the power distribution equipment generates a new fault event, analyzing an upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence; according to the upgrading risk assessment value, final early warning grade confirmation is carried out on the new fault event of the power distribution equipment; and updating the historical alarm event sequence. The risk of alarm storm is reduced, and the operation and maintenance efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment fault early warning technology, specifically to a method and system for intelligent fault early warning of power distribution equipment based on the Internet of Things. Background Technology

[0002] Internet of Things (IoT)-based power equipment condition monitoring and early warning systems have significant application value in modern power systems. Through sensor data acquisition, wireless communication transmission, condition assessment, and intelligent early warning, the system can achieve comprehensive and real-time monitoring of power equipment, effectively reducing equipment failure rates and maintenance costs, and improving the safety and reliability of power grid operation.

[0003] Existing early warning level management technologies often lack flexibility and foresight. Once an alarm is triggered, its level is usually fixed after the initial judgment, or it can only be mechanically upgraded based on simple time accumulation rules, such as "automatically upgrading the level after the same alarm persists for a certain period of time." This method ignores the spatiotemporal correlation and evolutionary characteristics of faults. It cannot distinguish between an occasional, independent event and the beginning of a continuously worsening fault trend, nor can it identify the "recurring" characteristics of similar faults repeatedly occurring on the same device. Therefore, the system cannot intelligently suppress occasional event alarms without subsequent harm, nor can it proactively and promptly upgrade the early warning level to attract sufficient attention when the fault shows clear signs of deterioration. The result is often that, on the one hand, a large number of low-value alarms interfere with normal operation and maintenance; on the other hand, for truly significant fault risks, the upgrade of early warnings appears lagging and blind, losing the best opportunity to intervene before the accident escalates. The accuracy and timeliness of early warnings are difficult to meet the needs of modern intelligent operation and maintenance.

[0004] Therefore, this invention discloses an intelligent fault early warning method and system for power distribution equipment based on the Internet of Things to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent fault early warning method and system for power distribution equipment based on the Internet of Things, so as to solve the problems raised in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent fault early warning method for power distribution equipment based on the Internet of Things, the method comprising the following steps: S1: Collect time-series operation data of power distribution equipment, clean and align it, and analyze the comprehensive characteristics of power distribution equipment based on a sliding time window; S2: Based on the comprehensive characteristics of the power distribution equipment, analyze the abnormal conditions of the power distribution equipment; determine the initial fault level based on the abnormal conditions; S3: When a new fault event occurs in the power distribution equipment, analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence; S4: Based on the upgraded risk assessment value, confirm the final warning level for new fault events of power distribution equipment; and update the historical alarm event sequence.

[0007] S1 includes the following: S101: By using an acquisition interface adapted to several industrial communication protocols, synchronously acquire the raw readings of voltage, current, temperature, power quality parameters, and insulation resistance of the target power distribution equipment; add a timestamp and device identifier to each raw reading to generate a raw data sequence with a spatiotemporal tag; the industrial communication protocols include, but are not limited to, Modbus, Profibus, OPC UA, and IEC 61850; the power quality parameters include, but are not limited to, total harmonic distortion and harmonic content; The original data sequence is cleaned to generate a cleaned data sequence, including missing value imputation and noise filtering. S102: Based on a preset sliding time window, the cleaned data sequence is processed by sliding; for each physical quantity within the sliding time window, the statistical distribution characteristics, temporal evolution characteristics, and physical coupling-derived characteristics are analyzed in parallel. The statistical distribution characteristics include mean, standard deviation, skewness, and kurtosis; the temporal evolution characteristics include the mean and standard deviation of the first-order difference sequence of the cleaned data sequence, denoted as difference mean and difference standard deviation, and linear regression is performed on the cleaned data sequence within the sliding time window, with the slope of the resulting regression line denoted as the trend index; the physical coupling derived characteristics include, but are not limited to, power temperature rise index, current temperature rise index, and stability index. The power-induced temperature rise index is equal to the average ratio of apparent power to the temperature of the device body; the current-induced temperature rise index is equal to the correlation coefficient between the time series of the current square and the corresponding temperature rise time series; the correlation coefficient is the Pearson correlation coefficient. The stability index is equal to the product of voltage unbalance and negative sequence current. S103: For each sliding time window, after normalizing the features of all physical quantities, they are spliced ​​and fused in a predetermined order to generate a comprehensive feature vector.

[0008] This invention utilizes statistical distribution features (such as skewness and kurtosis) to characterize the shape of data distribution, helping to discover abnormal patterns that traditional mean-standard deviation methods cannot detect. Temporal evolution features (difference and trend slope) focus on the data's changing process, capturing slow degradation trends earlier than absolute values ​​at a single moment, enabling early warning. The introduction of physically coupled derivative features encodes domain knowledge (electrical and thermal principles) into the features, allowing the model to understand the inherent physical relationships between parameters, significantly improving the feature's representational power and diagnostic interpretability. The power-temperature rise index directly reflects the temperature rise effect per unit power of the equipment, serving as a direct indicator for assessing heat dissipation efficiency and overload risk. The current-temperature rise index quantifies the correlation strength between Joule heating and temperature changes by calculating the correlation coefficient between the square of the current and the temperature rise, accurately identifying abnormal heating caused by increased resistance, poor contact, etc. The stability index combines voltage imbalance with negative sequence current, effectively characterizing the additional stress caused by grid power quality on the equipment. These features quantify expert experience, giving the system expert-like physical insight and significantly improving its ability to identify complex composite faults.

[0009] S2 includes the following: S201: Synchronously input the comprehensive feature vector into the threshold determination engine, statistical anomaly engine, and trend analysis engine; The threshold determination engine compares each feature value in the comprehensive feature vector with a preset safe operating threshold range. If the limit is exceeded, it generates a feature value abnormal evidence vector containing an evidence type identifier of the feature value abnormality, the corresponding feature parameter, the direction of exceeding the limit, and an initial severity score calculated based on the deviation percentage. The statistical anomaly engine is used to analyze the Z-score of physical quantities. If the absolute value of the Z-score exceeds a preset significance threshold, a statistical anomaly evidence vector is generated, which includes the evidence type identifier corresponding to the statistical anomaly, the corresponding feature parameters, the Z-score value, and an initial severity score that is positively correlated with the absolute value of the Z-score. The trend analysis engine is used to generate a trend anomaly evidence vector if the trend indicator continues to be in the same direction or the cumulative change exceeds the cumulative change threshold within a number of consecutive sliding time windows. This vector includes the evidence type identifier corresponding to the cumulative change anomaly, the corresponding feature parameters, the trend indicator, the cumulative change, and an initial severity score calculated based on the weighted average of the trend indicator and the cumulative change. The parallel output consists of a comprehensive evidence set composed of all generated evidence vectors; S202: Input the comprehensive evidence set into a preset rule level mapper, and map the initial fault level according to the preset decision rules; The device identifier, timestamp, initial fault level, and comprehensive evidence set are combined to generate preliminary diagnostic results; This invention employs a parallel engine architecture, enabling simultaneous and non-blocking threshold, statistical, and trend analysis methods, thus improving the efficiency of the diagnostic process. The threshold engine's deviation percentage calculation makes severity scores continuously quantifiable, rather than a simple Boolean judgment, preserving information on the degree of limit violation. The statistical anomaly engine's Z-score determination, based on the device's historical operating data, is an adaptive and personalized anomaly detection method, avoiding the problem of fixed thresholds being unsuitable for all operating conditions. The trend analysis engine's continuous window and cumulative change judgment effectively distinguish between occasional fluctuations and genuine deterioration trends, reducing false alarms. The combination of these three elements forms a three-in-one evidence system encompassing instantaneous limit violations, statistical deviations, and trend deterioration, resulting in a comprehensive and information-rich evidence set that provides raw materials for in-depth analysis in subsequent steps.

[0010] S3 includes the following: S301: Maintain a historical alarm event sequence for each monitored power distribution device, where each event record includes the initial fault level, fault type, occurrence time, and comprehensive evidence set; When a new fault event of the power distribution equipment is received, based on the current time, all historical alarm events that are within a preset recent time window and have the same fault type as the new fault event are retrieved from the historical alarm event sequence to form a relevant historical event subset. S302: Based on the aforementioned subset of relevant historical events, analyze the recurrence risk index and the deterioration trend index; S303: Normalize the relapse risk index and the deterioration trend index, and then perform weighted fusion of the normalized relapse risk index and the deterioration trend index to generate an upgrade risk assessment value.

[0011] This invention enables the system to have memory capabilities by maintaining and querying historical alarm event sequences, allowing it to perform personalized risk assessments based on the equipment's own fault history. The analysis of the recurrence risk index enables the system to identify recurring faults, and for frequently occurring similar problems, even if the severity of a single incident is not high, its cumulative risk can be quantitatively assessed. The analysis of the deterioration trend index enables the system to identify progressive faults, and has the ability to provide early warning for fault chains that develop from low-level to high-level.

[0012] S4 includes the following: S401: Maintain a cooling timer for each equipment type-fault type combination to define the cooling period of the combination; whenever a fault type issues an alert, the cooling timer corresponding to the equipment type-fault type combination starts counting down; When a new fault event is received from the power distribution equipment, a cooling-off period determination is performed; If it is in the cooling-off period and the risk assessment value of the new fault event is lower than the preset suppression threshold, it is judged as a non-persistent anomaly, and the decision is not to issue an alert, but only to log it. S402: If a new fault event is not in the cooling-off period, or is in the cooling-off period but the upgrade risk assessment value is greater than or equal to the preset suppression threshold, an early warning upgrade decision will be made for the new fault event. In the early warning escalation decision, if the escalation risk assessment value of the new fault event is between a preset first threshold and a second threshold, then the final early warning level is determined to be equal to the initial fault level. If the escalation risk assessment value of the new fault event is greater than or equal to the second threshold, the warning level will be escalated. Warning information will be issued based on the determined final warning level; The complete information of this fault event, including its comprehensive feature vector, evidence vector, initial fault level, escalation risk assessment value, and final warning level or suppression status, is added as a new alarm event to the historical alarm event sequence of the corresponding device. After an alert is issued, the cooling timer corresponding to the device type-fault type combination is reset, and the cooling duration is preset according to the final alert level; the higher the final alert level, the longer the cooling duration.

[0013] This invention employs a dual-judgment system of cooling-off period and suppression threshold to form the first-level filter, effectively blocking repetitive and minor invalid alarms within a short period. This directly reduces the risk of alarm storms at the source and improves operational efficiency. The two-stage decision-making based on the upgrade risk assessment value (suppression / classification / upgrade) constitutes a precise response mechanism, ensuring that limited operational attention resources are guided to truly important and urgent alarms. The decision results are fed back to the historical sequence and the cooling-off timer is reset, forming a complete closed loop. This allows the system to continuously self-optimize based on actual operational feedback and adapt to the characteristics of different devices and the requirements of operational strategies.

[0014] Another aspect of this application provides an intelligent fault early warning system for power distribution equipment based on the Internet of Things (IoT). The system is applied to the above-mentioned intelligent fault early warning method for power distribution equipment based on the IoT. The system includes an equipment data feature analysis module, an equipment anomaly analysis module, a historical evolution analysis module, and an early warning confirmation module. The equipment data feature analysis module is used to collect the time-series operation data of the power distribution equipment, clean and align it, and then analyze the comprehensive characteristics of the power distribution equipment based on a sliding time window. The equipment anomaly analysis module is used to analyze the anomalies of the power distribution equipment based on its comprehensive characteristics; and to determine the initial fault level based on the anomalies. The historical evolution analysis module is used to analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence when a new fault event occurs in the power distribution equipment. The early warning confirmation module is used to confirm the final early warning level of new fault events of power distribution equipment based on the upgraded risk assessment value; and to update the historical alarm event sequence.

[0015] The equipment data feature analysis module includes an equipment data acquisition unit and a feature fusion unit; The equipment data acquisition unit is used to synchronously acquire the raw readings of voltage, current, temperature, power quality parameters and insulation resistance of the target power distribution equipment through an acquisition interface adapted to several industrial communication protocols; add a timestamp and equipment identifier to each raw reading to generate a raw data sequence with a spatiotemporal label; and perform data cleaning on the raw data sequence to generate a cleaned data sequence. The feature fusion unit performs sliding processing on the cleaned data sequence based on a preset sliding time window; for the data of each physical quantity within the sliding time window, it analyzes the statistical distribution characteristics, temporal evolution characteristics, and physical coupling-derived characteristics in parallel; for each sliding time window, after normalizing the features of all physical quantities, they are spliced ​​and fused in a predetermined order to generate a comprehensive feature vector.

[0016] The equipment anomaly analysis module includes an anomaly evidence analysis unit and an initial fault confirmation unit; The abnormal evidence analysis unit is used to synchronously input the comprehensive feature vector into the threshold determination engine, the statistical anomaly engine, and the trend analysis engine; and output a comprehensive evidence set consisting of all generated evidence vectors in parallel. The initial fault confirmation unit is used to input the comprehensive evidence set into a preset rule level mapper, map the initial fault level according to the preset decision rules, and generate a preliminary diagnosis result by splicing the device identifier, timestamp, initial fault level and comprehensive evidence set. The historical evolution analysis module includes a historical event extraction unit and an escalation risk assessment unit; The historical event extraction unit is used to, when a new fault event of the power distribution equipment is received, retrieve all historical alarm events within a preset recent time window and whose fault type is the same as the new fault event from the historical alarm event sequence based on the current time, and form a relevant historical event subset. The upgrade risk assessment unit is used to analyze the recurrence risk index and the deterioration trend index based on a subset of relevant historical events; normalize the recurrence risk index and the deterioration trend index; and perform weighted fusion of the normalized recurrence risk index and the deterioration trend index to generate an upgrade risk assessment value.

[0017] The early warning confirmation module includes a cooling confirmation unit and a final early warning confirmation unit; The cooling confirmation unit is used to maintain a cooling timer for each equipment type-fault type combination to define the cooling period of the combination; whenever a fault type issues an early warning, the cooling timer corresponding to the equipment type-fault type combination starts counting down; when a new fault event of the power distribution equipment is received, the cooling period is determined. The final warning confirmation unit is used to make a warning escalation decision for a new fault event if it is not in the cooling-off period, or if it is in the cooling-off period but the escalation risk assessment value is greater than or equal to the preset suppression threshold; and to issue warning information according to the determined final warning level. The complete information of this fault event, including its comprehensive feature vector, evidence vector, initial fault level, escalation risk assessment value, and final warning level or suppression status, is added as a new alarm event to the historical alarm event sequence of the corresponding device. After an alert is issued, the cooling timer corresponding to the device type-fault type combination is reset, and the cooling duration is preset according to the final alert level; the higher the final alert level, the longer the cooling duration.

[0018] Compared with existing technologies, the beneficial effects of this invention are as follows: The statistical distribution features of this invention can characterize the shape of data distribution, helping to discover abnormal patterns that traditional mean-standard deviation methods cannot detect; the temporal evolution features focus on the data change process, capturing the slow degradation trend of equipment earlier than the absolute value at a single moment, achieving early warning; the introduction of physically coupled derived features encodes domain knowledge into the features, enabling the model to understand the inherent physical relationship between parameters, significantly improving the feature representation ability and diagnostic interpretability; this invention adopts a parallel engine architecture, allowing threshold, statistical, and trend analysis methods to be performed simultaneously without blocking each other, improving the efficiency of the diagnostic process; the deviation percentage calculation of the threshold engine makes the severity score continuously quantifiable, rather than a simple Boolean judgment, preserving information on the degree of exceeding limits; the Z-score judgment of the statistical anomaly engine is based on the equipment's own historical operating data, which is an adaptive and personalized anomaly detection method, avoiding the problem that fixed thresholds are not suitable for all operating conditions; the continuous window and cumulative change judgment of the trend analysis engine can effectively distinguish between random fluctuations and real deterioration trends, reducing false alarms. This invention employs a dual-judgment system of cooling period and suppression threshold to form the first-level filter, effectively blocking repetitive and minor invalid alarms within a short period. This directly reduces the risk of alarm storms at the source and improves operational efficiency. The two-stage decision-making based on the upgrade risk assessment value constitutes a precise response mechanism, ensuring that limited operational attention resources are guided to truly important and urgent alarms. The decision results are fed back to the historical sequence and the cooling timer is reset, forming a complete closed loop. This allows the system to continuously self-optimize based on actual operational feedback and adapt to the characteristics of different devices and the requirements of operational strategies. Attached Figure Description

[0019] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating the intelligent fault early warning method for power distribution equipment based on the Internet of Things according to the present invention. Figure 2 This is a schematic diagram of the structure of the Internet of Things-based intelligent fault early warning system for power distribution equipment according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 The present invention provides a technical solution: an intelligent fault early warning method for power distribution equipment based on the Internet of Things, the method comprising the following steps: S1: Collect time-series operation data of power distribution equipment, clean and align it, and analyze the comprehensive characteristics of power distribution equipment based on a sliding time window; S1 includes the following: S101: Through an acquisition interface adapted to several industrial communication protocols, synchronously acquire the raw readings of voltage, current, temperature, power quality parameters, and insulation resistance of the target power distribution equipment; add a timestamp and device identifier to each raw reading to generate a raw data sequence with spatiotemporal tags; industrial communication protocols include but are not limited to Modbus, Profibus, OPCUA, and IEC 61850; power quality parameters include but are not limited to total harmonic distortion and harmonic content; Data cleaning is performed on the original data sequence to generate a cleaned data sequence, including missing value imputation and noise filtering; S102: Based on a preset sliding time window, the cleaned data sequence is processed by sliding; for each physical quantity within the sliding time window, the statistical distribution characteristics, temporal evolution characteristics, and physical coupling-derived characteristics are analyzed in parallel. Statistical distribution characteristics include mean, standard deviation, skewness, and kurtosis; temporal evolution characteristics include the mean and standard deviation of the first-order difference sequence of the cleaned data sequence, denoted as difference mean and difference standard deviation; linear regression is performed on the cleaned data sequence within the sliding time window, and the slope of the resulting regression line is denoted as the trend index; physical coupling-derived characteristics include, but are not limited to, power heating index, current heating index, and stability index. The power-induced temperature rise index is equal to the average ratio of apparent power to the temperature of the equipment body; the current-induced temperature rise index is equal to the correlation coefficient between the time series of the current square and the corresponding temperature rise time series; the stability index is equal to the product of voltage imbalance and negative sequence current. S103: For each sliding time window, after normalizing the features of all physical quantities, they are spliced ​​and fused in a predetermined order to generate a comprehensive feature vector.

[0022] S2: Based on the comprehensive characteristics of the power distribution equipment, analyze the abnormal conditions of the power distribution equipment; determine the initial fault level based on the abnormal conditions; S2 includes the following: S201: Synchronously input the comprehensive feature vector into the threshold determination engine, statistical anomaly engine, and trend analysis engine; The threshold determination engine compares each feature value in the comprehensive feature vector with the preset safe operation threshold range. If the limit is exceeded, it generates a feature value abnormal evidence vector containing an evidence type identifier of the feature value abnormality, the corresponding feature parameter, the direction of exceeding the limit, and an initial severity score calculated based on the deviation percentage. The statistical anomaly engine is used to analyze the Z-score of physical quantities. If the absolute value of the Z-score exceeds the preset significance threshold, a statistical anomaly evidence vector is generated, which includes the evidence type identifier corresponding to the statistical anomaly, the corresponding feature parameters, the Z-score value, and the initial severity score that is positively correlated with the absolute value of the Z-score. The trend analysis engine is used to generate a trend anomaly evidence vector if the trend indicator continues to be in the same direction or the cumulative change exceeds the cumulative change threshold within several consecutive sliding time windows. This vector includes the evidence type identifier corresponding to the cumulative change anomaly, the corresponding feature parameters, the trend indicator, the cumulative change, and an initial severity score calculated based on the weighted average of the trend indicator and the cumulative change. The parallel output consists of a comprehensive evidence set composed of all generated evidence vectors; S202: Input the comprehensive evidence set into the preset rule level mapper, and map the initial fault level according to the preset decision rules; Example 1: The decision rule includes at least the following: if the comprehensive evidence set is empty, then the initial fault level is mapped to the normal level; if the initial severity scores in the comprehensive evidence set are all lower than the severity threshold, then the initial fault level is mapped to the low-level alarm. If there is only one evidence vector in the comprehensive evidence set whose initial severity scores are all greater than or equal to the severity threshold, then the initial fault level is mapped to a medium alarm. If there is more than one evidence vector in the comprehensive evidence set whose initial severity scores are all greater than or equal to the severity threshold, then the initial fault level is mapped to a high-level alarm. The device identifier, timestamp, initial fault level, and comprehensive evidence set are combined to generate preliminary diagnostic results; S3: When a new fault event occurs in the power distribution equipment, analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence; S3 includes the following: S301: Maintain a historical alarm event sequence for each monitored power distribution device, where each event record includes the initial fault level, fault type, occurrence time, and comprehensive evidence set; When a new fault event of the power distribution equipment is received, based on the current time, all historical alarm events that are within a preset recent time window and have the same fault type as the new fault event are retrieved from the historical alarm event sequence to form a relevant historical event subset. S302: Based on a subset of relevant historical events, analyze the relapse risk index and the deterioration trend index; Example 2: In this example, ; Among them, R recur E represents the risk index of recurrence. i H represents a subset of relevant historical events. rel The i-th historical alarm event in the middle, W Li Indicates historical warning event E i The initial fault level corresponds to the severity weighting coefficient; λ represents the time decay coefficient; t now t represents the current time. i Indicates historical warning event E i The time of occurrence; Using the initial severity scores and corresponding occurrence times of each historical alarm event in the relevant historical event subset, a model for fitting the deterioration trend is established: y = α1 × x + α2; where α1 represents the deterioration trend index; α2 represents the fitting coefficient; x represents the independent variable of time; and y represents the dependent variable of the initial severity score. The least squares method is used to calculate and solve α1 and α2 in the model for fitting the deterioration trend. S303: Normalize the relapse risk index and the deterioration trend index, and then perform a weighted fusion of the normalized relapse risk index and the deterioration trend index to generate an upgrade risk assessment value.

[0023] S4: Based on the upgraded risk assessment value, confirm the final warning level for new fault events of power distribution equipment; and update the historical alarm event sequence.

[0024] S4 includes the following: S401: Maintain a cooling timer for each equipment type-fault type combination to define the cooling period of the combination; whenever a fault type issues an alert, the cooling timer corresponding to the equipment type-fault type combination starts counting down; When a new fault event is received from the power distribution equipment, a cooling-off period determination is performed; If it is in the cooling-off period and the risk assessment value of the new fault event is lower than the preset suppression threshold, it is judged as a non-persistent anomaly, and the decision is not to issue an alert, but only to log it. S402: If a new fault event is not in the cooling-off period, or is in the cooling-off period but the upgrade risk assessment value is greater than or equal to the preset suppression threshold, an early warning upgrade decision will be made for the new fault event. If the risk assessment value of the new fault event is between the preset first threshold and the second threshold in the early warning escalation decision, the final early warning level is determined to be equal to the initial fault level. If the escalation risk assessment value of a new fault event is greater than or equal to the second threshold, the warning level will be escalated. Warning information will be issued based on the determined final warning level; Example 3: In this example, the final warning level is based on formula L. final =min(L init +N, L max The calculation determines the value, where N is the preset upgrade step size, and L is the value. init Initial fault level; L max The highest warning level; The complete information of this fault event, including its comprehensive feature vector, evidence vector, initial fault level, escalation risk assessment value, and final warning level or suppression status, is added as a new alarm event to the historical alarm event sequence of the corresponding device. After an alert is issued, the cooling timer corresponding to the device type-fault type combination is reset. The cooling duration is preset according to the final alert level; the higher the final alert level, the longer the cooling duration.

[0025] Please see Figure 2 The present invention provides a technical solution: an intelligent fault early warning system for power distribution equipment based on the Internet of Things, which includes an equipment data feature analysis module, an equipment anomaly analysis module, a historical evolution analysis module, and an early warning confirmation module; The equipment data feature analysis module is used to collect time-series operation data of power distribution equipment, clean and align it, and then analyze the comprehensive characteristics of the power distribution equipment based on a sliding time window. The equipment anomaly analysis module is used to analyze the abnormal conditions of power distribution equipment based on its comprehensive characteristics; and to determine the initial fault level based on the anomaly conditions. The historical evolution analysis module is used to analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence when a new fault event occurs in the power distribution equipment. The early warning confirmation module is used to confirm the final early warning level of new fault events of power distribution equipment based on the upgraded risk assessment value; and to update the historical alarm event sequence.

[0026] The equipment data feature analysis module includes an equipment data acquisition unit and a feature fusion unit; The equipment data acquisition unit is used to synchronously acquire the raw readings of voltage, current, temperature, power quality parameters and insulation resistance of the target power distribution equipment through an acquisition interface adapted to several industrial communication protocols; it adds a timestamp and equipment identifier to each raw reading to generate a raw data sequence with spatiotemporal tags; and it cleans the raw data sequence to generate a cleaned data sequence. The feature fusion unit performs sliding processing on the cleaned data sequence based on a preset sliding time window; for each physical quantity within the sliding time window, it analyzes the statistical distribution characteristics, temporal evolution characteristics, and physical coupling-derived characteristics in parallel; for each sliding time window, after normalizing the features of all physical quantities, they are spliced ​​and fused in a predetermined order to generate a comprehensive feature vector.

[0027] The equipment anomaly analysis module includes an anomaly evidence analysis unit and an initial fault confirmation unit; The anomaly evidence analysis unit is used to synchronously input the comprehensive feature vector into the threshold determination engine, the statistical anomaly engine, and the trend analysis engine; and output a comprehensive evidence set consisting of all generated evidence vectors in parallel. The initial fault confirmation unit is used to input the comprehensive evidence set into the preset rule level mapper, map the initial fault level according to the preset decision rules, and generate a preliminary diagnosis result by splicing the device identifier, timestamp, initial fault level and comprehensive evidence set. The historical evolution analysis module includes a historical event extraction unit and an escalation risk assessment unit; The historical event extraction unit is used to retrieve, based on the current time, all historical alarm events within a preset recent time window that have the same fault type as the new fault event from the historical alarm event sequence when a new fault event is received from the power distribution equipment, and form a relevant historical event subset. The upgraded risk assessment unit is used to analyze the recurrence risk index and the deterioration trend index based on a subset of relevant historical events; the recurrence risk index and the deterioration trend index are normalized, and the normalized recurrence risk index and the deterioration trend index are weighted and fused to generate the upgraded risk assessment value.

[0028] The early warning confirmation module includes a cooling confirmation unit and a final early warning confirmation unit; The cooling confirmation unit is used to maintain a cooling timer for each equipment type-fault type combination to define the cooling period of the combination; whenever a fault type issues an early warning, the cooling timer corresponding to the equipment type-fault type combination starts counting down; when a new fault event of the power distribution equipment is received, the cooling period is determined. The final warning confirmation unit is used to make a warning escalation decision for a new fault event if it is not in the cooling-off period, or if it is in the cooling-off period but the escalation risk assessment value is greater than or equal to the preset suppression threshold; and to issue warning information according to the determined final warning level. The complete information of this fault event, including its comprehensive feature vector, evidence vector, initial fault level, escalation risk assessment value, and final warning level or suppression status, is added as a new alarm event to the historical alarm event sequence of the corresponding device. After an alert is issued, the cooling timer corresponding to the device type-fault type combination is reset. The cooling duration is preset according to the final alert level; the higher the final alert level, the longer the cooling duration.

[0029] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "include," "contain," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0030] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for intelligent fault early warning of power distribution equipment based on the Internet of Things, characterized in that, The method includes the following steps: S1: Collect time-series operation data of power distribution equipment, clean and align it, and analyze the comprehensive characteristics of power distribution equipment based on a sliding time window; S2: Based on the comprehensive characteristics of the power distribution equipment, analyze the abnormal conditions of the power distribution equipment; determine the initial fault level based on the abnormal conditions; S3: When a new fault event occurs in the power distribution equipment, analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence; S4: Based on the upgraded risk assessment value, confirm the final warning level for new fault events of power distribution equipment; And update the historical alarm event sequence.

2. The method for intelligent fault early warning of power distribution equipment based on the Internet of Things according to claim 1, characterized in that: S1 includes the following: S101: By using an acquisition interface adapted to several industrial communication protocols, synchronously acquire the raw readings of voltage, current, temperature, power quality parameters, and insulation resistance of the target power distribution equipment; add a timestamp and device identifier to each raw reading to generate a raw data sequence with a spatiotemporal label; perform data cleaning on the raw data sequence to generate a cleaned data sequence. S102: Based on a preset sliding time window, the cleaned data sequence is processed by sliding; for each physical quantity within the sliding time window, the statistical distribution characteristics, temporal evolution characteristics, and physical coupling-derived characteristics are analyzed in parallel. S103: For each sliding time window, after normalizing the features of all physical quantities, they are spliced ​​and fused in a predetermined order to generate a comprehensive feature vector.

3. The method for intelligent fault early warning of power distribution equipment based on the Internet of Things according to claim 2, characterized in that: The statistical distribution features include mean, standard deviation, skewness, and kurtosis; the temporal evolution features include the mean and standard deviation of the first-order difference sequence of the cleaned data sequence, denoted as difference mean and difference standard deviation, and linear regression is performed on the cleaned data sequence within the sliding time window, with the slope of the resulting regression line denoted as the trend index; the physical coupling derived features include, but are not limited to, power heating index, current heating index, and stability index.

4. The intelligent fault early warning method for power distribution equipment based on the Internet of Things according to claim 3, characterized in that: The power temperature rise index is equal to the average ratio of apparent power to the temperature of the device body; the current temperature rise index is equal to the correlation coefficient between the time series of the current square and the corresponding temperature rise time series; the stability index is equal to the product of voltage imbalance and negative sequence current.

5. The intelligent fault early warning method for power distribution equipment based on the Internet of Things according to claim 4, characterized in that: S2 includes the following: S201: Synchronously input the comprehensive feature vector into the threshold determination engine, statistical anomaly engine, and trend analysis engine; and output a comprehensive evidence set consisting of all generated evidence vectors in parallel. S202: Input the comprehensive evidence set into a preset rule level mapper, and map the initial fault level according to the preset decision rules; The device identifier, timestamp, initial fault level, and comprehensive evidence set are combined to generate preliminary diagnostic results.

6. The method for intelligent fault early warning of power distribution equipment based on the Internet of Things according to claim 5, characterized in that: The threshold determination engine compares each feature value in the comprehensive feature vector with a preset safe operation threshold range. If the limit is exceeded, it generates a feature value abnormal evidence vector containing an evidence type identifier of the feature value abnormality, the corresponding feature parameter, the direction of exceeding the limit, and an initial severity score calculated based on the deviation percentage. The statistical anomaly engine is used to analyze the Z-score of physical quantities. If the absolute value of the Z-score exceeds a preset significance threshold, a statistical anomaly evidence vector is generated, which includes the evidence type identifier corresponding to the statistical anomaly, the corresponding feature parameters, the Z-score value, and an initial severity score that is positively correlated with the absolute value of the Z-score. The trend analysis engine is used to generate a trend anomaly evidence vector if the trend indicator continues to be in the same direction or the cumulative change exceeds the cumulative change threshold within a number of consecutive sliding time windows. This vector includes the evidence type identifier corresponding to the abnormal cumulative change, the corresponding feature parameters, the trend indicator, the cumulative change, and an initial severity score calculated based on the weighted average of the trend indicator and the cumulative change.

7. The intelligent fault early warning method for power distribution equipment based on the Internet of Things according to claim 6, characterized in that: S3 includes the following: S301: Maintain a historical alarm event sequence for each monitored power distribution device, where each event record includes the initial fault level, fault type, occurrence time, and comprehensive evidence set; When a new fault event of the power distribution equipment is received, based on the current time, all historical alarm events that are within a preset recent time window and have the same fault type as the new fault event are retrieved from the historical alarm event sequence to form a relevant historical event subset. S302: Based on the aforementioned subset of relevant historical events, analyze the recurrence risk index and the deterioration trend index; S303: Normalize the relapse risk index and the deterioration trend index, and then perform weighted fusion of the normalized relapse risk index and the deterioration trend index to generate an upgrade risk assessment value.

8. The method for intelligent fault early warning of power distribution equipment based on the Internet of Things according to claim 7, characterized in that: S4 includes the following: S401: Maintain a cooling timer for each equipment type-failure type combination to define the cooling period of the combination; Whenever a warning is issued for a fault type, the cooling timer corresponding to the device type-fault type combination starts its countdown. When a new fault event is received from the power distribution equipment, a cooling-off period determination is performed; If it is in the cooling-off period and the escalation risk assessment value of the new fault event is lower than the preset suppression threshold, it is judged as a non-persistent anomaly, and the decision is not to issue an alert, but only to record it in the log. S402: If a new fault event is not in the cooling-off period, or is in the cooling-off period but the upgrade risk assessment value is greater than or equal to the preset suppression threshold, an early warning upgrade decision will be made for the new fault event. Warning information will be issued based on the determined final warning level; After an alert is issued, the cooling timer corresponding to the device type-fault type combination is reset, and the cooling duration is preset according to the final alert level. The complete information of this fault event, including its comprehensive feature vector, evidence vector, initial fault level, escalation risk assessment value, and final warning level or suppression status, is added as a new alarm event to the historical alarm event sequence of the corresponding device.

9. The method for intelligent fault early warning of power distribution equipment based on the Internet of Things according to claim 8, characterized in that: In the early warning escalation decision, if the escalation risk assessment value of the new fault event is between a preset first threshold and a second threshold, the final early warning level is determined to be equal to the initial fault level; if the escalation risk assessment value of the new fault event is greater than or equal to the second threshold, the early warning level is escalated.

10. An IoT-based intelligent fault early warning system for power distribution equipment, wherein the system is applied to the IoT-based intelligent fault early warning method for power distribution equipment as described in any one of claims 1-9, characterized in that, The system includes a device data feature analysis module, a device anomaly analysis module, a historical evolution analysis module, and an early warning confirmation module; The equipment data feature analysis module is used to collect the time-series operation data of the power distribution equipment, clean and align it, and then analyze the comprehensive characteristics of the power distribution equipment based on a sliding time window. The equipment anomaly analysis module is used to analyze the anomalies of the power distribution equipment based on its comprehensive characteristics; and to determine the initial fault level based on the anomalies. The historical evolution analysis module is used to analyze the upgrade risk assessment value of the power distribution equipment based on the corresponding historical alarm event sequence when a new fault event occurs in the power distribution equipment. The early warning confirmation module is used to confirm the final early warning level of new fault events of power distribution equipment based on the upgraded risk assessment value. And update the historical alarm event sequence.

Images