Adaptive monitoring method for integrated power cabinet

Abstract

This invention relates to the field of power equipment monitoring and adaptive control technology, and discloses an adaptive monitoring method and system for integrated power cabinets. The method includes collecting multimodal sensor data and cleaning and calibrating it to obtain environmental time-series data; analyzing state parameters to identify abnormal fluctuations and overlapping sections; analyzing the risk propagation range to classify risk levels; matching a rule base to generate a control scheme and correcting it to obtain adaptive parameters; encapsulating instruction transmission verification; archiving logs and evaluating the operation recovery effect; and determining the adaptive stable state. This method achieves accurate perception of power cabinet operating parameters, rapid identification of abnormal operating conditions, and adaptive control, solving the problems of fixed thresholds, delayed response, and poor control adaptability in traditional monitoring, thus improving the operational stability and safety assurance capabilities of integrated power cabinets.

Description

Technical Field

[0001] This invention relates to the field of integrated power cabinet operation monitoring and intelligent control technology, and in particular to an adaptive monitoring method and system for integrated power cabinets. Background Technology

[0002] As the core power supply equipment in scenarios such as power systems, communication base stations, and industrial control, the operational stability of integrated power cabinets directly affects the safe and reliable operation of the entire supporting system. With the continuous upgrading of intelligent power management needs, the requirements for the efficiency and accuracy of real-time status monitoring, accurate anomaly identification, and adaptive control of integrated power cabinets are also increasing.

[0003] In existing technologies, the monitoring of integrated power cabinets mostly adopts the traditional fixed threshold monitoring mode, relying on a single type of sensor to collect basic operating parameters. The supporting data acquisition and control system can only realize simple parameter acquisition, transmission and threshold alarm functions. It cannot effectively filter environmental interference and perform time synchronization calibration on multi-modal sensor data, and it is difficult to accurately capture the dynamic characteristics of parameters such as load fluctuations and temperature and humidity changes. At the same time, existing monitoring methods lack in-depth analysis of operating deviations and correlation with historical data, and cannot make targeted adjustments based on the degree and spread of abnormal fluctuations. When complex operating conditions with multiple abnormal parameters intertwined occur, problems such as monitoring lag and poor adaptability of control parameters are prone to occur, making it impossible to avoid operational risks in a timely manner, thus affecting the long-term stable operation of the integrated power cabinet.

[0004] Therefore, existing technologies suffer from problems such as low monitoring accuracy and untimely abnormal response of integrated power cabinets due to imperfect data acquisition and control system functions and a lack of adaptive analysis and precise control capabilities in monitoring methods, making it difficult to achieve full-process adaptive and stable monitoring. Summary of the Invention

[0005] This invention provides an adaptive monitoring method and system for integrated power cabinets to solve the problems in the prior art, such as low monitoring accuracy, delayed abnormal response, and difficulty in achieving stable and reliable adaptive monitoring, caused by the imperfect functions of the data acquisition and control system, insufficient identification of the intertwined characteristics of load and temperature and humidity abnormal fluctuations, and lack of adaptive and precise control capabilities.

[0006] In a first aspect, to solve the above-mentioned technical problems, the present invention provides an adaptive monitoring method for an integrated power supply cabinet, comprising:

[0007] Collect multimodal sensor data including temperature, humidity, and load fluctuations, and preprocess the multimodal sensor data to obtain environmental state information;

[0008] Based on the environmental status information, the load and temperature and humidity parameters are analyzed to generate fluctuation trend characteristics. The fluctuation trend characteristics are compared with the preset fluctuation threshold to obtain the complete operating deviation.

[0009] The complete operational deviation is scanned node by node in time sequence to obtain the multi-parameter abnormal superposition section. The fluctuation amplitude and deviation degree of the multi-parameter abnormal superposition section are analyzed to obtain the location of the operational abnormality.

[0010] The abnormal time-series features are extracted from the environmental state records of the abnormal operation location. Based on the preset data hierarchical storage and the abnormal time-series features, the deviation degree score is calculated.

[0011] Based on the deviation degree score, the operational state segment of the complete operational deviation is classified to obtain risk classification features. The risk classification features are then matched with a preset abnormal transmission path to obtain the risk propagation range.

[0012] Based on the risk propagation range, a preset adjustment rule base is retrieved to obtain a parameter configuration scheme. The parameter configuration scheme is then matched and verified dimension by dimension with the abnormal operating conditions of the risk propagation range to obtain a corrected parameter sequence.

[0013] The execution verification features are obtained by performing instruction transmission and integrity verification based on the modified parameter sequence, and the execution verification features are stored hierarchically to obtain data update features. The stability of the power cabinet's operating state is determined based on the data update features to obtain an adaptive stable state.

[0014] Secondly, the present invention provides an adaptive monitoring system for an integrated power supply cabinet, comprising:

[0015] The data acquisition module is used to collect multimodal sensor data including temperature, humidity, and load fluctuations, and to preprocess the multimodal sensor data to obtain environmental state information.

[0016] The status analysis module is used to analyze the load and temperature and humidity parameter characteristics based on the environmental status information, generate fluctuation trend characteristics, compare the fluctuation trend characteristics with the preset fluctuation threshold, and obtain the complete operating deviation.

[0017] An anomaly identification module is used to perform a node-by-node time-series scan of the complete operational deviation to obtain a multi-parameter anomaly superposition segment, analyze the fluctuation amplitude and deviation degree of the multi-parameter anomaly superposition segment, and obtain the location of the operational anomaly.

[0018] The deviation comparison module is used to detect the environmental state records of the abnormal operation location, extract abnormal time-series features, and calculate the deviation degree score based on the preset data hierarchical storage and the abnormal time-series features.

[0019] The risk identification module is used to classify the operational state segments of the complete operational deviation according to the deviation degree score to obtain risk classification features, and match the risk classification features with a preset abnormal transmission path to obtain the risk propagation range;

[0020] The parameter correction module is used to retrieve a preset adjustment rule base based on the risk propagation range to obtain a parameter configuration scheme, and to match and verify the parameter configuration scheme with the abnormal working conditions of the risk propagation range dimension by dimension to obtain a correction parameter sequence.

[0021] The stability determination module is used to perform instruction transmission and integrity verification according to the modified parameter sequence to obtain execution verification characteristics, record operation logs according to the execution verification characteristics, update data hierarchical storage according to the log recording characteristics to obtain data update characteristics, and determine the stability of the power cabinet operation state according to the data update characteristics to obtain an adaptive stable state.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) By building a dedicated data acquisition and control system, the present invention performs environmental interference filtering and time synchronization calibration on multimodal sensor data, which breaks through the limitations of traditional monitoring that relies on a single sensor and is prone to data distortion. It realizes accurate acquisition and standardized transmission of sensor data, and effectively improves the accuracy and reliability of integrated power cabinet operation parameter monitoring.

[0024] (2) This invention analyzes environmental state information, generates fluctuation trend features and compares them with preset thresholds, and combines state traversal algorithm to accurately identify anomalies. It overcomes the problem of existing technologies not capturing anomalies with multiple intertwined parameters in a timely manner and not judging deviations accurately. It realizes rapid location and accurate identification of operational anomalies and greatly improves anomaly response efficiency.

[0025] (3) This invention combines historical similar states with real-time operating conditions, generates a precise control scheme through adaptability verification, and simultaneously links historical data updates and state determination to form a complete closed loop. It breaks through the limitations of traditional monitoring that lacks adaptive control capabilities and cannot dynamically adjust according to actual operating conditions, and realizes full-process adaptive and stable monitoring of integrated power cabinet, effectively avoids operational risks, and ensures long-term stable operation of equipment. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart of an adaptive monitoring method for an integrated power cabinet provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of an adaptive monitoring system for an integrated power cabinet provided in the second embodiment of the present invention. Detailed Implementation

[0028] 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.

[0029] Reference Figure 1 The first embodiment of the present invention provides an adaptive monitoring method for an integrated power supply cabinet, comprising the following steps:

[0030] S1, Collect multimodal sensor data including temperature, humidity, and load fluctuations, and preprocess the multimodal sensor data to obtain environmental state information;

[0031] S2, based on the environmental state information, analyze the load and temperature and humidity parameter characteristics, generate fluctuation trend characteristics, compare the fluctuation trend characteristics with the preset fluctuation threshold, and obtain the complete operating deviation;

[0032] S3, the complete operational deviation is scanned node by node in time sequence to obtain the multi-parameter abnormal superposition section, and the fluctuation amplitude and deviation degree of the multi-parameter abnormal superposition section are analyzed to obtain the location of the operational abnormality;

[0033] S4, detect the environmental state record of the abnormal operation location, extract abnormal time sequence features, and calculate the deviation degree score based on the preset data hierarchical storage and the abnormal time sequence features;

[0034] S5, classify the operational state segments of the complete operational deviation according to the deviation degree score to obtain risk classification features, match the risk classification features with the preset abnormal transmission path to obtain the risk propagation range;

[0035] S6. Based on the risk propagation range, retrieve the preset adjustment rule base to obtain the parameter configuration scheme, and match and verify the parameter configuration scheme with the abnormal working conditions of the risk propagation range dimension by dimension to obtain the correction parameter sequence;

[0036] S7. Based on the modified parameter sequence, execute the instruction transmission and integrity verification to obtain the execution verification feature, store the execution verification feature in layers to obtain the data update feature, determine the stability of the power cabinet's operating state based on the data update feature, and obtain the adaptive stable state.

[0037] In step S1, multimodal sensor data including temperature, humidity, and load fluctuations are collected. The multimodal sensor data is preprocessed to obtain environmental state information, including:

[0038] Collect multimodal sensor data on temperature, humidity, and load fluctuations, and filter the multimodal sensor data for electromagnetic noise and remove physical extreme anomalies to obtain cleaning sensor data;

[0039] The cleaning sensor data is time-aligned and calibrated according to a unified timestamp to obtain time-series calibration features;

[0040] The time-series calibration features are processed to unify the dimensions to obtain environmental time-series data.

[0041] The topological relationship of the internal space of the power cabinet is obtained, and the location labels of the environmental time series data are completed according to the topological relationship. The environmental time series data and the location labels are fused in a spatiotemporal dimension to obtain environmental status information.

[0042] Temperature data is collected using temperature sensors located on the top and side panels and the surfaces of key heat-generating components inside the power cabinet. The sampling frequency is set to once per second, and the unit is °C. Humidity data is collected using humidity sensors located at the ventilation openings and in the middle of the power cabinet. The sampling frequency is set to once per second, and the unit is %. Load fluctuation data is collected using current transformers and voltage transformers installed in the main and branch circuits of the power cabinet. The power value is calculated using these sensors, and the sampling frequency is set to 10 times per second, and the unit is kW.

[0043] Electromagnetic noise filtering employs a moving average filtering method, setting the sliding window length to 10 sampling points to smooth the raw sensor data and eliminate instantaneous fluctuations caused by high-frequency electromagnetic interference. Physical extreme value anomaly removal uses a threshold judgment method, setting the physical extreme value ranges for temperature (-10℃ to 85℃), humidity (0% to 100%), and load power (0kW to 150% of rated power). Data points exceeding these ranges are judged as physical extreme value anomalies and removed, thus obtaining the cleaning sensor data.

[0044] A unified timestamp uses a network time protocol for clock synchronization to ensure the consistency of data acquisition time across all sensors. Timing alignment calibration employs a linear interpolation method. For time deviations caused by communication delays or differences in sampling frequencies, interpolation calculations are performed with a 1-second time granularity to ensure that temperature, humidity, and load data correspond at the same time point, thus obtaining timing calibration characteristics.

[0045] The Z-score standardization method was used to unify the dimensions of temperature, humidity, and load power. Historical means and standard deviations were calculated for each, converting the raw data into dimensionless standard fractions. The historical mean for temperature was set at 25℃, and the standard deviation at 5℃; the historical mean for humidity was set at 50%, and the standard deviation at 15%; the historical mean for load power was set at 60% of the rated power, and the standard deviation at 20% of the rated power. After Z-score standardization, the dimensions of each parameter were unified, resulting in environmental time-series data.

[0046] The topology includes the spatial coordinates of each functional area within the power cabinet and the sensor placement information. The spatial coordinates are represented using a three-dimensional Cartesian coordinate system, with the lower left front edge of the power cabinet as the origin, the positive X-axis pointing horizontally to the right, the positive Y-axis pointing horizontally backward, and the positive Z-axis pointing vertically upward, with the unit being meters. The sensor placement information records the installation coordinates of each temperature sensor, humidity sensor, and current transformer.

[0047] The location labels of the environmental time-series data are completed based on the topological relationships. Each location label includes the sensor number and its corresponding three-dimensional spatial coordinates, associating each sampled data point with the spatial coordinates of its acquiring sensor to form a data record with a location identifier. Spatiotemporal dimensional fusion employs a data stitching method, combining timestamps, spatial coordinates, and parameter values ​​into a five-dimensional data vector. The timestamp occupies one dimension, the spatial coordinates occupy three dimensions, and the parameter values ​​occupy one dimension, yielding environmental state information. This environmental state information simultaneously contains a complete description of both the temporal and spatial dimensions, providing a data foundation for subsequent analysis.

[0048] For example, an integrated power cabinet contains three temperature sensors installed at coordinates (0.2, 0.3, 1.5), (0.8, 0.3, 1.5), and (0.5, 0.8, 1.2), two humidity sensors installed at coordinates (0.5, 0.5, 1.8) and (0.5, 0.5, 0.3), and two current transformers installed in the main circuit and branch circuit, respectively. At a certain moment, the collected temperature data are 32℃, 35℃, and 30℃; the humidity data are 65% and 62%; and the load power data are 45kW and 12kW. After time-series alignment using the unified timestamp of January 15, 2024, at 14:30:00, and after Z-value standardization, the standard scores for temperature are 1.4, 2.0, and 1.0; the standard scores for humidity are 1.0 and 0.8; and the standard scores for load power are 1.5 and -1.8. After completing the location tags, an environmental status information record is generated, including timestamps, sensor numbers, three-dimensional coordinates, and standardized parameter values.

[0049] In step S2, the load and temperature / humidity parameters are analyzed based on the environmental state information to generate fluctuation trend characteristics. These fluctuation trend characteristics are then compared with a preset fluctuation threshold to obtain the complete operational deviation, including:

[0050] Based on the environmental state information, the load current, cabinet temperature, and ambient humidity parameters are analyzed to obtain a multidimensional state parameter set. Based on the multidimensional state parameter set, a multidimensional state vector at the same time section is constructed.

[0051] Based on the multidimensional state vector, the local variance and gradient of the sliding time window are calculated to obtain the parameter change feature set. Based on the parameter change feature set, the fluctuation trend of load and temperature and humidity is generated to obtain the fluctuation trend feature.

[0052] If the fluctuation trend characteristics exceed the preset fluctuation threshold, abnormal periods are selected; if the fluctuation trend characteristics do not exceed the preset fluctuation threshold, normal operation is determined.

[0053] Abnormal data segments of environmental state information are extracted based on the abnormal time period. The deviation value between the abnormal data segments and the preset standard operating baseline is calculated based on the abnormal data segments to obtain the deviation distribution characteristics. The abnormal segments and parameter deviation information are integrated based on the deviation distribution characteristics to obtain the complete operating deviation.

[0054] Standardized load power, temperature, and humidity values ​​are extracted from environmental condition information to form a multidimensional state parameter set. This multidimensional state parameter set contains a comprehensive description of the electrical load state and environmental state of the power cabinet at the same time.

[0055] The multidimensional state vector is represented in column vector form with a dimension of 3. The first dimension is the standard fraction of load power, the second dimension is the standard fraction of temperature, and the third dimension is the standard fraction of humidity. The vector is represented as X, which is equal to the transpose of the standard fractions of load power, temperature, and humidity.

[0056] The sliding time window is set to 60 seconds, containing 60 sampling points. Local variance is calculated using the unbiased variance formula, determining the dispersion of each parameter's value within the window to obtain the load power variance, temperature variance, and humidity variance. Gradient calculation employs the first-order difference method, calculating the rate of change between adjacent sampling points within the window to obtain the load power gradient, temperature gradient, and humidity gradient. The parameter variation feature set includes two types of features: local variance and gradient. Local variance reflects the drastic fluctuation of the parameter over a short period, while the gradient reflects the direction and rate of change of the parameter.

[0057] The fluctuation trend characteristics are represented by a combination of trend direction indicators and trend strength values. The trend direction indicators include three states: rising, falling, and stable. The trend strength value is the absolute value of the gradient. The trend determination rule is set as follows: if the absolute value of the gradient at three consecutive sampling points is greater than 0.1, then a trend is determined to have formed; a positive gradient indicates an upward trend, a negative gradient indicates a downward trend, and otherwise, it is determined to be a stable trend.

[0058] The preset fluctuation thresholds include two types: variance thresholds and gradient thresholds. The load power variance threshold is set to 0.5, the temperature variance threshold to 0.3, and the humidity variance threshold to 0.4; the load power gradient threshold is set to 0.2 per second, the temperature gradient threshold to 0.15 per second, and the humidity gradient threshold to 0.1 per second. When the local variance or gradient of any parameter exceeds the corresponding threshold, it is determined that there is an abnormal fluctuation at that moment, and the time period consisting of consecutive abnormal moments is marked as an abnormal period.

[0059] If the fluctuation trend characteristics do not exceed the preset fluctuation threshold, the system is considered to be operating normally. Under normal operating conditions, the local variance and gradient of each parameter remain within the threshold range, indicating that the power cabinet is operating stably and there is no need to proceed to the subsequent anomaly analysis process.

[0060] The abnormal data segment contains multi-dimensional state vectors of all sampling points within the abnormal period, along with corresponding timestamps and location tags. The preset standard operating baseline is represented by the statistical average of historical normal operating data. The load power baseline is set to 60% of the rated power, the temperature baseline to 25℃, and the humidity baseline to 50%. Deviation values ​​are calculated using the absolute difference method, which is the absolute value of the difference between the value of each sampling point in the abnormal data segment and the baseline value, yielding the load power deviation, temperature deviation, and humidity deviation.

[0061] The deviation distribution characteristics include the deviation value sequence and the deviation duration. The deviation value sequence is the set of deviation values ​​of each sampling point arranged in chronological order, and the deviation duration is the start and end time interval of the abnormal period.

[0062] The complete operational deviation is recorded in a structured data format, including the abnormal section identifier, start time, end time, types of parameters involved, deviation sequence of each parameter, maximum deviation value, and average deviation value, providing a complete data description for subsequent anomaly localization analysis.

[0063] For example, on January 15, 2024, from 14:30:00 to 14:35:00, a power cabinet experienced significant fluctuations in load power, temperature, and humidity. Calculations using a sliding time window showed that the local variance and gradient of load power and temperature both exceeded the corresponding preset fluctuation thresholds, classifying this period as an abnormal period of 300 seconds. Deviations from the preset standard operating baseline were calculated, with a maximum load power deviation of 8kW, a maximum temperature deviation of 13℃, and a maximum humidity deviation of 12%. A complete operational deviation record was obtained, with the abnormal segment identified as Z20240115001, starting at 14:30:00 and ending at 14:35:00. The parameters involved were load power and temperature, with a maximum load power deviation of 8kW and an average deviation of 6kW, and a maximum temperature deviation of 13℃ and an average deviation of 10℃.

[0064] In step S3, the complete operational deviation is scanned node by node in a time sequence to obtain a multi-parameter anomaly superposition segment. The fluctuation amplitude and deviation degree of the multi-parameter anomaly superposition segment are analyzed to obtain the operational anomaly location, including:

[0065] The complete operational deviation is scanned node by node in time sequence. The abnormal frequency and parameter linkage features of nodes within the complete operational deviation are extracted to obtain a deviation feature set. Based on the deviation feature set, the abnormal fluctuations and intertwining of load and temperature and humidity are identified to obtain the fluctuation intertwining points.

[0066] Based on the analysis of the interlacing points of the fluctuation, the interlacing distribution characteristics within the continuous sampling period are analyzed. Based on the interlacing distribution characteristics, the time intervals in which the load and temperature and humidity parameters simultaneously exceed the preset safety threshold range are screened, and the multi-parameter abnormal superposition segment is located.

[0067] The fluctuation amplitude and deviation values ​​of the multi-parameter abnormal superposition sections are statistically analyzed item by item to obtain a quantitative statistical feature set. The abnormal superposition coefficient is calculated based on the quantitative statistical feature set, and the abnormal duration and comprehensive deviation level are determined according to the abnormal superposition coefficient.

[0068] Based on the duration of the anomaly and the overall deviation level, the core anomaly node with the highest overall deviation level is identified in the complete operational deviation, thus obtaining the location of the operational anomaly.

[0069] It should be noted that a node refers to a monitoring unit within the power cabinet, defined by its spatial location. The physical location of each temperature and humidity sensor is defined as a monitoring node, and the node number corresponds to the sensor number. The time-series scan traverses each sampling moment in the complete operational deviation record in chronological order, checking for any abnormal states at each node.

[0070] The node anomaly frequency statistics track the number of times each node exhibits parameter deviations during abnormal periods. The parameter linkage feature analysis examines the correlation between abnormal states of different parameters at the same time, including four linkage modes: simultaneous occurrence of load power and temperature anomalies, simultaneous occurrence of load power and humidity anomalies, simultaneous occurrence of temperature and humidity anomalies, and simultaneous anomalies of all three parameters. The deviation feature set includes a node anomaly frequency statistics table and a parameter linkage feature record table. The node anomaly frequency statistics table records the node number and the number of times it occurs, while the parameter linkage feature record table records the linkage mode identifier at each sampling time.

[0071] A fluctuation interleaving point refers to a point in time where at least two of the parameters—load power, temperature, and humidity—simultaneously exhibit abnormal fluctuations at the same monitoring node or adjacent monitoring nodes. The identification rule is as follows: at the same sampling time, if the deviation values ​​of at least two parameters of a monitoring node are both greater than 50% of their respective preset safety thresholds, or if the load power deviation of that node is greater than its preset safety threshold and the temperature deviation of any adjacent monitoring node is greater than its preset safety threshold, then that moment is determined to be a fluctuation interleaving point.

[0072] The continuous sampling period is set to 30 seconds. The distribution density of fluctuation interleaving points within this period is analyzed. The distribution density is calculated by dividing the number of fluctuation interleaving points by the total sampling period duration, in units of points per second. The preset safety threshold ranges are: load power deviation not exceeding 20% ​​of rated power, temperature deviation not exceeding 15℃, and humidity deviation not exceeding 20%. The screening rule is that if at least three fluctuation interleaving points exist within the continuous sampling period, and the deviation values ​​of all parameters exhibiting abnormal fluctuations within that time interval exceed their respective preset safety thresholds, then that time interval is marked as a multi-parameter abnormal superposition segment. Multi-parameter abnormal superposition segments are described using segment identifiers, start times, end times, involved node numbers, superimposed parameter types, and distribution density values.

[0073] It should be added that the upper limit of 2.0 for the distribution density coefficient is set based on the following: Statistical analysis of 100 actual abnormal events shows that when the distribution density of fluctuation interlacing points exceeds 0.2 per second, the anomaly is already in a rapid diffusion stage. Further increasing the coefficient's contribution to the anomaly superposition coefficient tends to saturate and will not change the overall deviation level determination. Analysis of the correlation between the anomaly superposition coefficient and the final risk level in historical data reveals that when the distribution density coefficient exceeds 2.0, the anomaly superposition coefficient is always greater than 2.5, and the corresponding overall deviation level is always severe, consistent with the determination result when the upper limit of 2.0 is used. Therefore, an upper limit of 2.0 is set to suppress overweighting caused by extreme fluctuation amplitudes.

[0074] The duration adjustment factor is determined by analyzing the correlation between the duration of an anomaly and the deviation score. It was found that when the duration of an anomaly exceeds 450 seconds (1.5 times the baseline of 300 seconds), the deviation score is already below 40, indicating a high-risk level. Further increasing the duration factor will not change the risk classification result. Furthermore, excessively long anomaly periods may contain multiple independent anomalies. To avoid over-amplifying a single anomaly, the upper limit of the duration adjustment factor is set at 1.5.

[0075] It is worth noting that the fluctuation amplitude is the difference between the maximum and minimum values ​​of each parameter within the range, while the deviation is the average and maximum values ​​of the deviations of each parameter. The quantitative statistical feature set includes the fluctuation amplitude, average deviation, maximum deviation, and standard deviation of each parameter. The anomaly superposition coefficient is calculated using a weighted summation method. The formula is: the anomaly superposition coefficient equals the load power fluctuation amplitude multiplied by a weight of 0.4, plus the temperature fluctuation amplitude multiplied by a weight of 0.35, plus the humidity fluctuation amplitude multiplied by a weight of 0.25, multiplied by the distribution density coefficient. The distribution density coefficient is the ratio of the actual distribution density to the baseline distribution density of 0.1 units per second, with an upper limit set at 2.0. The anomaly superposition coefficient is a dimensionless value.

[0076] The abnormal superposition coefficient is calculated using a weighted summation method. First, the fluctuation amplitude of each parameter is divided by its corresponding preset safety threshold upper limit to obtain the dimensionless relative fluctuation amplitude. The preset safety threshold upper limit for load power is 20% of the rated power, the preset safety threshold upper limit for temperature is 15℃, and the preset safety threshold upper limit for humidity is 20%. The relative fluctuation amplitude is calculated as follows: the relative fluctuation amplitude of load power equals the load power fluctuation amplitude minus the rated power multiplied by 20%; the relative fluctuation amplitude of temperature equals the temperature fluctuation amplitude divided by 15℃; and the relative fluctuation amplitude of humidity equals the humidity fluctuation amplitude divided by 20%.

[0077] Therefore, the formula for calculating the abnormal superposition coefficient is as follows: first, multiply the relative fluctuation amplitude of the load power by 0.4, the relative fluctuation amplitude of the temperature by 0.35, and the relative fluctuation amplitude of the humidity by 0.25. Then, add the three products together to get a sum. Finally, multiply this sum by the distribution density coefficient to obtain the abnormal superposition coefficient.

[0078] The distribution density coefficient is the ratio of the actual distribution density to the baseline distribution density of 0.1 units per second, with an upper limit set at 2.0. The anomaly superposition coefficient is a dimensionless value.

[0079] The duration of the anomaly is the actual time span of the superimposed segment of multiple parameter anomalies. The comprehensive deviation level is divided according to the anomaly superposition coefficient: an anomaly superposition coefficient of less than 1.0 is a mild level, 1.0 to 2.0 is a moderate level, and greater than 2.0 is a severe level.

[0080] For example, in the complete operational deviation record Z20240115001, a node-by-node time-series scan revealed 12 deviations in the temperature sensor at node 1, 15 deviations in the temperature sensor at node 2, and 20 deviations in the main circuit load power. Eight fluctuation interleaving points were identified, distributed between 14:31:30 and 14:33:00. Analysis of the interleaving distribution characteristics showed a fluctuation interleaving point density of 0.27 points per second within a 30-second sampling period. A multi-parameter abnormal superposition segment was identified, starting at 14:31:30 and ending at 14:33:00, lasting 90 seconds, involving node 1 and the main circuit, with superimposed parameters being temperature and load power. The fluctuation amplitude was statistically analyzed: the load power fluctuation amplitude was 18kW, and the temperature fluctuation amplitude was 8℃. Assuming the rated power of the power cabinet is 100kW, the preset safety threshold upper limit for load power is 20kW, and the preset safety threshold upper limit for temperature is 15℃. The relative fluctuation amplitude of load power = 18 / 20 = 0.9, and the relative fluctuation amplitude of temperature = 8 / 15 ≈ 0.533. The distribution density coefficient = 0.27 / 0.1 = 2.7, with an upper limit of 2.0. The anomaly superposition coefficient = (0.9 × 0.4 + 0.533 × 0.35) × 2.0 = (0.36 + 0.1866) × 2.0 = 0.5466 × 2.0 = 1.0932, which is judged as a moderate level. Node 2 is identified as the core abnormal node, and the abnormal operation location is recorded as node number 2, coordinates (0.8, 0.3, 1.5). The comprehensive deviation level is moderate, the anomaly duration is 90 seconds, and the anomaly superposition coefficient is 1.09.

[0081] The core anomaly node is selected by prioritizing the node with a severe overall deviation level and the longest anomaly duration. If multiple severe nodes exist, the node with the largest anomaly superposition coefficient is selected as the core anomaly node. The runtime anomaly location record includes the spatial coordinates, node number, overall deviation level, anomaly duration, and anomaly superposition coefficient of the core anomaly node.

[0082] In step S4, the environmental state records at the abnormal operation location are detected to extract abnormal time-series features. Based on the preset data hierarchical storage and the abnormal time-series features, a deviation score is calculated, including:

[0083] Detect the environmental status records at the location of the abnormal operation, and extract the time-series change sequences of load and temperature and humidity from the environmental status records;

[0084] Based on the time-series change sequence, retrieve historical abnormal data of the same working condition and parameter type stored in the preset historical data hierarchy, compare the similarity between the historical abnormal data and the time-series change sequence to obtain similar abnormal segments.

[0085] The environmental state records are matched point-by-point in time sequence with the similar abnormal segments, and the differences in numerical deviations and trends are statistically analyzed to obtain a sequence difference feature set.

[0086] The sequence difference feature set is weighted and calculated according to a preset weighting coefficient to obtain a time-series difference quantification value. The overall deviation degree is determined based on the time-series difference quantification value to obtain a deviation degree score.

[0087] It should be noted that, based on the node number recorded at the location of the operational anomaly, all historical data records of that node during the abnormal period are extracted from the environmental status information database. The time-series change sequences include numerical sequences of load power changing over time, temperature changing over time, and humidity changing over time. Each sequence contains the values ​​of all sampling points within the start and end time of the abnormal period, with a sampling interval of 1 second.

[0088] The pre-defined historical data tiered storage adopts a three-tiered structure. The first tier is categorized by power cabinet model; the second tier is categorized by load condition (light load, medium load, heavy load); and the third tier is categorized by abnormal parameter type (load abnormality, temperature abnormality, humidity abnormality, and combined abnormality). During retrieval, the system first matches the power cabinet model, then determines the load condition category based on the current load power, and finally matches the corresponding historical abnormal data record based on the abnormal parameter type.

[0089] Similarity comparison employs a dynamic time warping algorithm to calculate the warped path distance between the current time-series change sequence and historical anomaly data sequences; a smaller distance indicates higher similarity. A similarity threshold is set, and historical anomaly data with a warped path distance less than the threshold are identified as similar anomaly segments. These segments are then sorted by distance from smallest to largest, and the top 5 are selected as the set of similar anomaly segments.

[0090] Time-series point-by-point matching employs an equal-time-interval alignment method, mapping the sampling points of the current abnormal time period one-to-one with the sampling points of similar abnormal segments at corresponding times. Numerical deviation and trend difference are statistically analyzed. Numerical deviation is calculated as the absolute value of the difference between the values ​​of the current sequence and similar segments at corresponding times; trend difference is calculated as the absolute value of the difference in the gradient of change between the current sequence and similar segments at corresponding times. The sequence difference feature set includes the numerical deviation sequence and trend difference sequence for each matching point, as well as the statistics for the average numerical deviation, maximum numerical deviation, average trend difference, and maximum trend difference over the entire time period.

[0091] The preset weighting coefficients are set as follows: numerical deviation weight 0.6, trend difference weight 0.4. The weighted calculation formula is: the time series difference quantification value equals the average numerical deviation multiplied by 0.6 plus the average trend difference multiplied by 0.4, and then multiplied by the duration adjustment coefficient. The duration adjustment coefficient is the ratio of the actual abnormal duration to the baseline duration of 300 seconds, with an upper limit set at 1.5.

[0092] The deviation score is expressed on a 100-point scale. The calculation formula is: deviation score = 100 minus the time series difference quantification value multiplied by 10. The lower the score, the greater the deviation from the historical normal pattern and the more severe the anomaly. A deviation score below 60 is considered high risk, 60 to 80 is considered medium risk, and above 80 is considered low risk.

[0093] For example, the time-series change sequence of node 2 at a certain abnormal operation location contains 90 sampling points, with the load power increasing from 45kW to 68kW and the temperature increasing from 32℃ to 40℃. A preset hierarchical storage of historical data was retrieved, matching 120 historical records of the same power cabinet model, heavy load condition, and temperature-load combination anomaly type. A dynamic time warping algorithm was used to calculate similarity, and the 5 records with the smallest warped path distance were selected as similar anomaly segments. Time-series point-by-point matching was performed, and the average numerical deviations were found to be 3.5kW for load power and 2.1℃ for temperature, with average trend differences of 0.05 seconds for load power and 0.03 seconds for temperature. The time-series difference quantification value is calculated by weighting according to the preset weighting coefficients. It equals (3.5 x 0.6 + 2.1 x 0.6 + 0.05 x 0.4 + 0.03 x 0.4) x (90 divided by 300, but with an upper limit of 1.5, the result is 1.5) = (2.1 x 1.26 + 0.02 + 0.012) x 1.5 = 5.088. The deviation score is 100 minus 5.088 x 10 = 49.12 points, which is classified as a high-risk level.

[0094] In step S5, risk classification features are obtained by classifying the operational state segments of the complete operational deviation according to the deviation degree score. These risk classification features are then matched with a preset anomaly propagation path to obtain the risk propagation range, including:

[0095] By comparing the deviation score with the preset multi-level risk threshold, the risk level of each operating state segment is divided to obtain risk classification features;

[0096] Based on the risk classification features, the same level and type of abnormal working condition records in the historical database are associated with each other, and a time-series association of historical similar states is constructed in chronological order.

[0097] Based on the historical similarity state chain, the abnormal transmission path of preset load and temperature and humidity is matched, and the nodes and propagation sections of abnormal spread are traced to obtain the risk propagation range.

[0098] It is worth noting that the preset multi-level risk threshold is set to three levels. A deviation score greater than or equal to 80 and less than or equal to 100 is considered low risk; greater than or equal to 60 and less than 80 is considered medium risk; and greater than or equal to 0 and less than 60 is considered high risk. Based on the threshold range of the deviation score, the operating status segment is divided into three levels: low risk, medium risk, and high risk. Risk classification characteristics include risk level identifier, deviation score, risk threshold range, and judgment timestamp information. Related query conditions include the same risk level, the same abnormal parameter type, and the same power cabinet model; records of abnormal operating conditions that meet the conditions are retrieved from the historical database.

[0099] A historical similarity state chain is constructed in chronological order. Chronological association refers to sorting retrieved historical abnormal operating condition records according to the order in which the abnormalities occurred, forming a time series. The historical similarity state chain includes the time of occurrence, location, parameters, handling measures, results, and subsequent development information of the abnormality, used to analyze the development patterns of similar abnormalities.

[0100] The preset anomaly propagation paths are a knowledge base of anomaly propagation patterns derived from statistical analysis of a large number of historical anomaly cases. It includes seven basic types: load anomaly propagation paths, temperature anomaly propagation paths, humidity anomaly propagation paths, load-temperature coupling propagation paths, load-humidity coupling propagation paths, temperature-humidity coupling propagation paths, and three-parameter coupling propagation paths. Each propagation path describes the direction, speed, and degree of impact of the anomaly propagating from the starting node to adjacent nodes.

[0101] The tracing process begins at the node where the operational anomaly is located. Based on the matched propagation path, it chronologically reconstructs the process of the anomaly spreading to adjacent nodes, determining the expected time of anomaly occurrence at each node. The risk propagation scope includes a list of affected node numbers, the expected time of anomaly at each node, a propagation path diagram, and information on the total duration of the impact.

[0102] For example, a certain operating state segment has a deviation score of 49.12. Compared with the preset multi-level risk thresholds, 0 to 60 points is considered a high-risk range, and this state segment is determined to be of a high-risk level. By referencing the historical database, 15 historical records of the same high-risk level, abnormal temperature-load combination, and the same model of power cabinet were retrieved. A historical similar state chain was constructed in chronological order, and analysis revealed that this type of anomaly propagates to adjacent nodes on average within 120 seconds of its occurrence. Matching the preset load and temperature / humidity anomaly propagation path, it was determined to be a load-temperature coupling propagation path. According to the statistical regularity of the historical similar state chain, the average time interval for this type of anomaly to propagate from the starting node to adjacent nodes is 60 seconds, and the average time interval to the next adjacent node is 120 seconds. Tracing the anomaly propagation, the adjacent nodes of node 2 include nodes 1 and 3. It is predicted that from the moment the anomaly occurs at node 2, node 1 will experience an anomaly 60 seconds later, and node 3 will experience an anomaly 120 seconds later. The risk propagation scope is recorded as affecting nodes 1, 2, and 3. Node 2 is already abnormal. The estimated time of abnormality for node 1 is 14:32:30, and for node 3 it is 14:33:30. The propagation path is from node 2 to node 1, and from node 2 to node 3, with a total impact duration of 300 seconds.

[0103] It is worth noting that the preset fluctuation threshold, preset safety threshold range, preset anomaly propagation path, and preset adjustment rule base are all established based on statistical analysis of at least 1000 hours of historical operating data from integrated power cabinets of the same model. Taking a power cabinet with a rated power of 100kW and internally arranged with 6 temperature sensors and 3 humidity sensors as an example, multimodal data under normal operation and simulated fault conditions were collected, with the sampling frequency consistent with step S1. After extracting the complete operating deviation, abnormal location, and deviation degree score from the collected data according to steps S2 to S4, the change patterns of each parameter within the time window from 60 seconds before the anomaly occurred to 120 seconds after the anomaly ended were statistically analyzed.

[0104] The variance and gradient thresholds in the preset fluctuation thresholds are set based on the upper limit of the 99.7% confidence interval of the sliding window statistics of each parameter during normal operation. Specifically, load power data for 720 consecutive hours under normal operating conditions is taken, and the local variance of load power per second (window 60 seconds) is calculated to obtain a variance sequence. The mean of this sequence plus three times the standard deviation is taken as the variance threshold. Similarly, the gradient threshold is taken as the 99.7% quantile of the absolute value of the gradient under normal operating conditions. Example values ​​are: load power variance threshold 0.5, temperature variance threshold 0.3, humidity variance threshold 0.4, and gradient thresholds of load power 0.2 per second, temperature 0.15 per second, and humidity 0.1 per second.

[0105] The preset safety threshold range is determined based on the equipment's factory tolerance limits and industry safety standards. The upper limit of the load power safety limit is 120% of the rated power, the lower limit is 0kW, and the deviation safety threshold is set at 20% of the rated power; the upper limit of the temperature safety limit is 65℃, the lower limit is 0℃, and the deviation safety threshold is set at 15℃; the upper limit of the humidity safety limit is 85%, the lower limit is 10%, and the deviation safety threshold is set at 20%.

[0106] The pre-defined anomaly propagation path is constructed by analyzing the temporal relationships in historical anomaly cases. At least 200 cases containing clear anomaly propagation records are selected from the historical database. Each case records the anomaly initiation node number, anomaly parameter type, and timestamps of the anomalies occurring at each node. For load-temperature coupled propagation paths, the time difference between the occurrence of anomalies between each pair of adjacent nodes is calculated, and the median of all valid cases is taken as the propagation time interval. An exemplary construction process is as follows: 50 cases of dual load-temperature anomalies are selected, all starting at node 2. The average time interval between the occurrence of temperature anomalies at adjacent nodes 1 is 58 seconds, the median is 60 seconds, and the standard deviation is 12 seconds; therefore, the propagation time interval is set to 60 seconds. The average time interval between the occurrence of anomalies at node 3 is 115 seconds, and the median is 120 seconds; therefore, it is set to 120 seconds. For propagation paths with different parameter combinations, their propagation direction, time interval, and impact attenuation coefficient are statistically analyzed, forming a rule table for seven basic propagation paths.

[0107] The pre-defined adjustment rule base is built based on statistical analysis of historical successful control cases and expert experience. Each rule includes triggering conditions (abnormal operating condition type, deviation score range, risk level), control measure combination, priority of each measure, and safety constraints. The rule base is constructed as follows: First, at least 300 historical abnormal control records are collected. Each record includes the abnormal operating condition type, control measure, parameter drop after control, and whether side effects occurred. Then, the abnormal operating condition types are grouped, and the average drop and success rate of different control measure combinations within each group are statistically analyzed. Finally, control combinations with a drop greater than 70% and a success rate higher than 85% are selected as recommended rules. An example rule is: the triggering condition is that the load temperature is abnormal and the deviation score is less than 60 points. The control combination is to switch to the backup circuit and start forced air cooling, where the priority of switching to the backup circuit is 1, the priority of forced air cooling is 2, and the safety constraint is that it is forbidden to switch two backup circuits at the same time and the continuous operation time of air cooling does not exceed 600 seconds.

[0108] In step S6, a parameter configuration scheme is obtained by retrieving a preset adjustment rule base based on the risk propagation range. The parameter configuration scheme is then matched and verified dimension-by-dimensionally with the abnormal operating conditions of the risk propagation range to obtain a correction parameter sequence, including:

[0109] Based on the risk propagation range, identify the specific abnormal operating conditions such as overload and abnormal temperature and humidity, and obtain the risk operating condition feature set;

[0110] Based on the risk condition feature set, the control configuration strategies in the preset adjustment rule base are retrieved, and parameter control combinations that are suitable for abnormal conditions are selected to obtain parameter configuration schemes.

[0111] The parameter configuration scheme is matched and verified with the abnormal operating conditions of the risk propagation range dimension by dimension, and the degree of fit in each dimension is calculated to obtain the fit verification coefficient.

[0112] The parameter configuration scheme is optimized and corrected based on the adaptation verification coefficient to obtain the corrected parameter sequence.

[0113] Abnormal operating condition type identification is based on the combination and degree of deviation of abnormal parameters within the risk propagation range, including seven specific types: single load over-limit, single temperature anomaly, single humidity anomaly, load and temperature dual anomaly, load and humidity dual anomaly, temperature and humidity dual anomaly, and comprehensive anomaly of three parameters. The risk operating condition feature set includes an abnormal operating condition type identifier, a list of involved parameters, the current value of each parameter, the target value range of each parameter, the location of the abnormal node, and the size of the impact range.

[0114] The preset adjustment rule base contains control configuration strategies for different abnormal operating condition types. Each strategy includes triggering conditions, control parameters, control amplitude, control priority, and safety constraints. During retrieval, the abnormal operating condition type and triggering conditions in the risk operating condition feature set are matched, and all control configuration strategies that meet the conditions are filtered out.

[0115] The parameter control combinations include three categories: load adjustment strategies, temperature control strategies, and humidity regulation strategies. Load adjustment strategies include switching to backup circuits, limiting output power, and balancing load distribution. Temperature control strategies include activating forced air cooling, adjusting air conditioning power, and opening ventilation holes. Humidity regulation strategies include activating dehumidifiers, adjusting ventilation volume, and activating humidification functions. The corresponding control combination is selected based on the type of abnormal operating condition to form a parameter configuration scheme. The scheme includes the specific parameter settings, execution order, and execution duration for each control measure.

[0116] When matching and verifying the parameter configuration scheme with the abnormal operating conditions within the risk propagation range dimension by dimension, each dimension includes three verification dimensions: load, temperature, and humidity. The matching and verification calculates the expected improvement effect of each control measure on the corresponding parameter, and the expected improvement effect is determined based on the statistical patterns of historical control cases.

[0117] Calculate the fit degree for each dimension to obtain the fit verification coefficient. The formula for calculating the fit verification coefficient is: the fit degree for each dimension equals the expected improvement value after adjustment of that dimension divided by the current deviation value of that dimension, multiplied by the weight coefficient of that dimension. The current deviation value is defined as the absolute value of the difference between the current actual parameter value and the target parameter value. The weight of the load dimension is 0.5, the weight of the temperature dimension is 0.3, and the weight of the humidity dimension is 0.2. The comprehensive fit verification coefficient is the weighted average of the fit degrees of the three dimensions.

[0118] It should be noted that the optimization and correction rules are as follows: if the adaptation verification coefficient is greater than 0.8, the original parameter configuration scheme is maintained; if the adaptation verification coefficient is between 0.5 and 0.8, the control amplitude is adjusted, and the control parameters are increased by 20%; if the adaptation verification coefficient is less than 0.5, the control strategy is changed, and an alternative control combination is selected. The corrected parameter sequence is an ordered set of optimized control parameters, including the final parameter values, execution sequence, execution conditions, and interlock protection settings for each control measure.

[0119] For example, a power cabinet has a rated power of 100kW. The risk propagation range indicates that the abnormal operating condition type is dual load and temperature anomalies. The current load power is 80kW (exceeding 80% of the rated power), and the target range is set to not exceed 70% of the rated power, i.e., 70kW, so the current deviation is 10kW. The current temperature is 38℃, the operating baseline is 25℃, and the target range is set to not exceed the baseline by 10℃, i.e., 35℃, so the current deviation is 3℃.

[0120] Search the preset adjustment rule library, match the control configuration strategy that triggers the dual abnormal conditions of load and temperature, filter out the parameter control combination, the load adjustment strategy is to switch to backup loop 1, the temperature control strategy is to start forced air cooling at level 3, and the execution time is 300 seconds.

[0121] Based on historical control case statistics, switching to the backup circuit is expected to reduce load power by 8kW, and starting forced air cooling is expected to reduce temperature by 2℃. Dimensional matching verification shows that the load dimension compatibility is (8kW / 10kW)×0.5=0.4; the temperature dimension compatibility is (2℃ / 3℃)×0.3=0.2. The overall compatibility verification coefficient is 0.4+0.2=0.6.

[0122] Since the adaptation verification coefficient of 0.6 falls between 0.5 and 0.8, according to the optimization and correction rules, the control parameters are increased by 20%, that is, the air cooling level is adjusted to level 4, while the execution time remains unchanged. The corrected parameter sequence is as follows: switch to backup circuit 1, start forced air cooling at level 4, execution time 300 seconds, execution condition is that the load power exceeds 70kW for 10 seconds, and the interlock protection is set to prohibit the simultaneous switching of two backup circuits.

[0123] In step S7, execution verification features are obtained by performing instruction transmission and integrity verification according to the modified parameter sequence. Data update features are obtained by hierarchically storing the execution verification features. The stability of the power cabinet's operating state is determined based on the data update features to obtain an adaptive stable state, including:

[0124] The modified parameter sequence is encapsulated into a control execution instruction. Based on the execution instruction, transmission and data integrity verification are performed. The verification instruction is free of packet loss and tampering, thus obtaining the execution verification characteristics.

[0125] The abnormal time, control parameters, and execution result information of the aforementioned execution verification features are integrated and recorded in a standardized format to form an operation log;

[0126] The operating logs are archived according to the preset abnormal control data, and the updated data is stored in layers to obtain data update characteristics.

[0127] Based on the data update characteristics, the statistical parameter decline amplitude and fluctuation frequency, the power cabinet operation recovery effect is evaluated, and the adaptive stable state of the integrated power cabinet is obtained.

[0128] It should be noted that the modified parameter sequence is encapsulated into a control execution instruction. The control execution instruction adopts a structured data format, including an instruction header, instruction type, control object, control parameters, execution sequence, and checksum field. The instruction header identifies the source and version of the instruction, the instruction type identifies the category of control measures, the control object identifies the hardware device number performing the control, the control parameters contain specific set values, the execution sequence identifies the start time and duration, and the checksum is generated using a cyclic redundancy check algorithm for integrity verification.

[0129] The transmission process sends instructions to the execution unit via the control bus, recording the transmission timestamp and reception confirmation signal during transmission. Data integrity verification involves recalculating the checksum at the receiving end and comparing it with the checksum in the instruction. If they match, no tampering is determined. Simultaneously, the integrity of instruction fields is checked; if no fields are missing, no packet loss is determined.

[0130] The execution verification features include command sending time, reception confirmation time, transmission delay duration, verification result, and retransmission count information, used to assess the reliability of command transmission. The anomaly time records the start time of the anomaly, the control parameters record the actual parameter settings issued, and the execution result records the actual execution status of each control measure, including success, failure, and partial success.

[0131] Record operations in a standardized format to form an operational log. The standardized format includes log identifier, recording time, exception identifier, list of control measures, execution result, operator, and audit status fields, forming a structured operational log record.

[0132] Pre-defined rules for archiving abnormal control data establish categorized directories based on operating condition type and risk level, storing operational logs in the corresponding directories and updating index information. Data tiered storage updates include adding new records to the historical abnormal database, updating statistical reports, and optimizing knowledge base rules. Data update characteristics include the number of updated records, update time, data version number, and storage location information.

[0133] The parameter decline magnitude calculation measures the degree to which each parameter returns to its normal range after the control is implemented. The formula is: decline magnitude equals the deviation value before control minus the deviation value after control, then divided by the deviation value before control, expressed as a percentage. The fluctuation frequency statistics count the number of times the parameter fluctuation exceeds the threshold within one hour after control. In this embodiment, the unit time is set to 1 hour.

[0134] The recovery effect of the power cabinet is evaluated to obtain the adaptive stable state of the integrated power cabinet. The recovery effect is evaluated by comprehensively considering the drop amplitude and fluctuation frequency. The evaluation rules are as follows: a drop amplitude greater than 80% and a fluctuation frequency of less than 2 times per hour are considered complete recovery; a drop amplitude of 50% to 80% and a fluctuation frequency of 2 to 5 times per hour are considered partial recovery; and other conditions are considered poor recovery. The adaptive stable state includes the recovery effect level, current operating parameters, stable duration, and next inspection time information.

[0135] For example, the modified parameter sequence is encapsulated as a control execution instruction, with instruction header version V2.1, instruction type load switching and air-cooling start, control objects backup circuit 1 and air-cooling unit 3, control parameters circuit switching to 1, air-cooling level 4, duration 300 seconds, and checksum 0xA3B5. This instruction is sent to the execution unit at 14:35:05, received confirmation at 14:35:06, with a transmission delay of 1 second. Verification passes, there are no retransmissions, and the execution verification feature record is complete. The abnormal time 14:30:00, control parameters circuit 1 and level 4, and execution result success are integrated to form an operation log, with log identifier RL20240115001.

[0136] Archived according to load and temperature dual anomalies and high risk levels, updated the historical anomaly database to add 1 record, the statistical report update control success rate is 100%, data update feature record update quantity 1, time 14 hours 35 minutes 30 seconds, version V20240115, storage location partition 3.

[0137] After adjustment, the load power decreased from 80kW to 72kW, and the deviation decreased from 10kW to 2kW, a reduction of (10-2)÷10=80%; the temperature decreased from 38℃ to 36℃, and the deviation decreased from 3℃ to 1℃, a reduction of (3-1)÷3≈66.7%. The fluctuation frequency was twice per hour. The overall reduction was taken as the average of the load and temperature reduction, 73.35%. According to the evaluation rules, a reduction greater than 80% and a fluctuation frequency less than twice per hour constitutes a complete recovery. In this case, the reduction of 73.35% is less than 80%, and the fluctuation frequency is equal to 2 times, so it is judged as a partial recovery. The adaptive stable state record is: recovery effect level: partial recovery; current load power: 72kW; temperature: 36℃; humidity: 60%; stable duration: 0 seconds (adjustment just completed); and the next inspection needs to be arranged.

[0138] Reference Figure 2 The second embodiment of the present invention provides an adaptive monitoring system for an integrated power supply cabinet, comprising:

[0139] The data acquisition module is used to collect multimodal sensor data including temperature, humidity, and load fluctuations, and to preprocess the multimodal sensor data to obtain environmental state information.

[0140] The status analysis module is used to analyze the load and temperature and humidity parameter characteristics based on the environmental status information, generate fluctuation trend characteristics, compare the fluctuation trend characteristics with the preset fluctuation threshold, and obtain the complete operating deviation.

[0141] An anomaly identification module is used to perform a node-by-node time-series scan of the complete operational deviation to obtain a multi-parameter anomaly superposition segment, analyze the fluctuation amplitude and deviation degree of the multi-parameter anomaly superposition segment, and obtain the location of the operational anomaly.

[0142] The deviation comparison module is used to detect the environmental state records of the abnormal operation location, extract abnormal time-series features, and calculate the deviation degree score based on the preset data hierarchical storage and the abnormal time-series features.

[0143] The risk identification module is used to classify the operational state segments of the complete operational deviation according to the deviation degree score to obtain risk classification features, and match the risk classification features with a preset abnormal transmission path to obtain the risk propagation range;

[0144] The parameter correction module is used to retrieve a preset adjustment rule base based on the risk propagation range to obtain a parameter configuration scheme, and to match and verify the parameter configuration scheme with the abnormal working conditions of the risk propagation range dimension by dimension to obtain a correction parameter sequence.

[0145] The stability determination module is used to obtain execution verification features by performing instruction transmission and integrity verification according to the modified parameter sequence, store the execution verification features in layers to obtain data update features, determine the stability of the power cabinet's operating state according to the data update features, and obtain an adaptive stable state.

[0146] It should be noted that the adaptive monitoring system for integrated power cabinets provided in this embodiment of the invention is used to execute all the process steps of the adaptive monitoring method for integrated power cabinets in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0147] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0148] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. An adaptive monitoring method for an integrated power supply cabinet, characterized in that, include: Collect multimodal sensor data including temperature, humidity, and load fluctuations, and preprocess the multimodal sensor data to obtain environmental state information; Based on the environmental status information, the load and temperature and humidity parameters are analyzed to generate fluctuation trend characteristics. The fluctuation trend characteristics are compared with the preset fluctuation threshold to obtain the complete operating deviation. The complete operational deviation is scanned node by node in time sequence to obtain the multi-parameter abnormal superposition section. The fluctuation amplitude and deviation degree of the multi-parameter abnormal superposition section are analyzed to obtain the location of the operational abnormality. The abnormal time-series features are extracted from the environmental state records of the abnormal operation location. Based on the preset data hierarchical storage and the abnormal time-series features, the deviation degree score is calculated. Based on the deviation degree score, the operational state segment of the complete operational deviation is classified to obtain risk classification features. The risk classification features are then matched with a preset abnormal transmission path to obtain the risk propagation range. Based on the risk propagation range, a preset adjustment rule base is retrieved to obtain a parameter configuration scheme. The parameter configuration scheme is then matched and verified dimension by dimension with the abnormal operating conditions of the risk propagation range to obtain a corrected parameter sequence. The execution verification features are obtained by performing instruction transmission and integrity verification based on the modified parameter sequence, and the execution verification features are stored hierarchically to obtain data update features. The stability of the power cabinet's operating state is determined based on the data update features to obtain an adaptive stable state.

2. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The acquisition includes multimodal sensor data on temperature, humidity, and load fluctuations. The multimodal sensor data is preprocessed to obtain environmental state information, including: Collect multimodal sensor data on temperature, humidity, and load fluctuations, and filter the multimodal sensor data for electromagnetic noise and remove physical extreme anomalies to obtain cleaning sensor data; The cleaning sensor data is time-aligned and calibrated according to a unified timestamp to obtain time-series calibration features; The time-series calibration features are processed to unify the dimensions to obtain environmental time-series data. The topological relationship of the internal space of the power cabinet is obtained, and the location labels of the environmental time series data are completed according to the topological relationship. The environmental time series data and the location labels are fused in a spatiotemporal dimension to obtain environmental status information.

3. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The process involves analyzing load and temperature / humidity parameters based on the environmental state information, generating fluctuation trend characteristics, comparing these characteristics with a preset fluctuation threshold, and obtaining the complete operational deviation, including: Based on the environmental state information, the load current, cabinet temperature, and ambient humidity parameters are analyzed to obtain a multidimensional state parameter set. Based on the multidimensional state parameter set, a multidimensional state vector at the same time section is constructed. Based on the multidimensional state vector, the local variance and gradient of the sliding time window are calculated to obtain the parameter change feature set. Based on the parameter change feature set, the fluctuation trend of load and temperature and humidity is generated to obtain the fluctuation trend feature. If the fluctuation trend characteristics exceed the preset fluctuation threshold, abnormal periods are selected; if the fluctuation trend characteristics do not exceed the preset fluctuation threshold, normal operation is determined. Abnormal data segments of environmental state information are extracted based on the abnormal time period. The deviation value between the abnormal data segments and the preset standard operating baseline is calculated based on the abnormal data segments to obtain the deviation distribution characteristics. The abnormal segments and parameter deviation information are integrated based on the deviation distribution characteristics to obtain the complete operating deviation.

4. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The process involves sequentially scanning the complete operational deviation node by node to obtain multi-parameter anomaly superposition segments, analyzing the fluctuation amplitude and deviation degree of these segments, and determining the operational anomaly location, including: The complete operational deviation is scanned node by node in time sequence. The abnormal frequency and parameter linkage features of nodes within the complete operational deviation are extracted to obtain a deviation feature set. Based on the deviation feature set, the abnormal fluctuations and intertwining of load and temperature and humidity are identified to obtain the fluctuation intertwining points. Based on the analysis of the interlacing points of the fluctuation, the interlacing distribution characteristics within the continuous sampling period are analyzed. Based on the interlacing distribution characteristics, the time intervals in which the load and temperature and humidity parameters simultaneously exceed the preset safety threshold range are screened, and the multi-parameter abnormal superposition segment is located. The fluctuation amplitude and deviation values ​​of the multi-parameter abnormal superposition sections are statistically analyzed item by item to obtain a quantitative statistical feature set. The abnormal superposition coefficient is calculated based on the quantitative statistical feature set, and the abnormal duration and comprehensive deviation level are determined according to the abnormal superposition coefficient. Based on the duration of the anomaly and the overall deviation level, the core anomaly node with the highest overall deviation level is identified in the complete operational deviation, thus obtaining the location of the operational anomaly.

5. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The environmental state records at the detected abnormal operation location are used to extract abnormal time-series features. Based on the preset data hierarchical storage and the abnormal time-series features, a deviation score is calculated, including: Detect the environmental status records at the location of the abnormal operation, and extract the time-series change sequences of load and temperature and humidity from the environmental status records; Based on the time-series change sequence, retrieve historical abnormal data of the same working condition and parameter type stored in the preset historical data hierarchy, compare the similarity between the historical abnormal data and the time-series change sequence to obtain similar abnormal segments. The environmental state records are matched point-by-point in time sequence with the similar abnormal segments, and the differences in numerical deviations and trends are statistically analyzed to obtain a sequence difference feature set. The sequence difference feature set is weighted and calculated according to a preset weighting coefficient to obtain a time-series difference quantification value. The overall deviation degree is determined based on the time-series difference quantification value to obtain a deviation degree score.

6. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The risk classification features are obtained by classifying the operational state segments of the complete operational deviation according to the deviation degree score, and the risk classification features are matched with the preset abnormal transmission path to obtain the risk propagation range, including: By comparing the deviation score with the preset multi-level risk threshold, the risk level of each operating state segment is divided to obtain risk classification features; Based on the risk classification features, the same level and type of abnormal working condition records in the historical database are associated with each other, and a time-series association of historical similar states is constructed in chronological order. Based on the historical similarity state chain, the abnormal transmission path of preset load and temperature and humidity is matched, and the nodes and propagation sections of abnormal spread are traced to obtain the risk propagation range.

7. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The step involves retrieving a parameter configuration scheme from a preset adjustment rule base based on the risk propagation range, and then matching and verifying the parameter configuration scheme against the abnormal operating conditions within the risk propagation range dimension by dimension to obtain a correction parameter sequence, including: Based on the risk propagation range, identify the specific abnormal operating conditions such as overload and abnormal temperature and humidity, and obtain the risk operating condition feature set; Based on the risk condition feature set, the control configuration strategies in the preset adjustment rule base are retrieved, and parameter control combinations that are suitable for abnormal conditions are selected to obtain parameter configuration schemes. The parameter configuration scheme is matched and verified with the abnormal operating conditions of the risk propagation range dimension by dimension, and the degree of fit in each dimension is calculated to obtain the fit verification coefficient. The parameter configuration scheme is optimized and corrected based on the adaptation verification coefficient to obtain the corrected parameter sequence.

8. The adaptive monitoring method for an integrated power supply cabinet according to claim 1, characterized in that, The process of obtaining execution verification features by performing instruction transmission and integrity verification according to the modified parameter sequence, storing the execution verification features hierarchically to obtain data update features, and determining the stability of the power cabinet's operating state based on the data update features to obtain an adaptive stable state includes: The modified parameter sequence is encapsulated into a control execution instruction. Based on the execution instruction, transmission and data integrity verification are performed. The verification instruction is free of packet loss and tampering, thus obtaining the execution verification characteristics. The abnormal time, control parameters, and execution result information of the aforementioned execution verification features are integrated and recorded in a standardized format to form an operation log; The operating logs are archived according to the preset abnormal control data, and the updated data is stored in layers to obtain data update characteristics. Based on the data update characteristics, the statistical parameter decline amplitude and fluctuation frequency, the power cabinet operation recovery effect is evaluated, and the adaptive stable state of the integrated power cabinet is obtained.

9. An adaptive monitoring system for an integrated power supply cabinet, characterized in that, include: The data acquisition module is used to collect multimodal sensor data including temperature, humidity, and load fluctuations, and to preprocess the multimodal sensor data to obtain environmental state information. The status analysis module is used to analyze the load and temperature and humidity parameter characteristics based on the environmental status information, generate fluctuation trend characteristics, compare the fluctuation trend characteristics with the preset fluctuation threshold, and obtain the complete operating deviation. An anomaly identification module is used to perform a node-by-node time-series scan of the complete operational deviation to obtain a multi-parameter anomaly superposition segment, analyze the fluctuation amplitude and deviation degree of the multi-parameter anomaly superposition segment, and obtain the location of the operational anomaly. The deviation comparison module is used to detect the environmental state records of the abnormal operation location, extract abnormal time-series features, and calculate the deviation degree score based on the preset data hierarchical storage and the abnormal time-series features. The risk identification module is used to classify the operational state segments of the complete operational deviation according to the deviation degree score to obtain risk classification features, and match the risk classification features with a preset abnormal transmission path to obtain the risk propagation range; The parameter correction module is used to retrieve a preset adjustment rule base based on the risk propagation range to obtain a parameter configuration scheme, and to match and verify the parameter configuration scheme with the abnormal working conditions of the risk propagation range dimension by dimension to obtain a correction parameter sequence. The stability determination module is used to obtain execution verification features by performing instruction transmission and integrity verification according to the modified parameter sequence, store the execution verification features in layers to obtain data update features, determine the stability of the power cabinet's operating state according to the data update features, and obtain an adaptive stable state.

Images