Multi-source data-based automatic electric charge returning and supplementing method and system

Abstract

The invention discloses an automatic electric charge returning and supplementing method and system based on multi-source data, and belongs to the technical field of computers. Attribution deviation caused by data splitting in the traditional technology is eliminated through systematic fusion of multi-source data, the timeliness and adaptability of operation state evaluation are ensured through a dynamic updating mechanism of an equipment state analysis model, and misjudgment caused by the fact that the equipment aging trend is not included in analysis is avoided; objective quantification of responsibility weight is realized through a joint analysis architecture of abnormal mode recognition and a return and supplement responsibility analysis model, so that influence degrees of different inducements such as equipment defects and environmental disturbance are distinguished, and the problem of high subjectivity of responsibility distribution is solved; the integration of the return and supplement strategy generator and the automatic execution link thoroughly replaces the manual coordination process, the return and supplement processing period is remarkably shortened, and meanwhile, the consistency and traceability of the processing result are guaranteed through the decision closed-loop design.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, in particular to an automatic electricity fee refund method and system based on multi-source data. BACKGROUND

[0002] In the operation of the power system, electricity fee refund processing is an important link in response to metering abnormalities, equipment failures and other events. At present, this process mainly relies on the experience of operation and maintenance personnel to make judgments on limited data, and is usually initiated and processed only according to single-dimensional information such as abnormal meter readings. This method has obvious technical limitations: on the one hand, due to the failure to systematically integrate multi-source data such as device operating status and environmental conditions, it is difficult to trace the root cause of the abnormality from a technical point of view; on the other hand, related methods lack quantitative attribution analysis capabilities for abnormal events, and cannot effectively distinguish the differentiated effects of device inherent defects, external environmental disturbances and other different causes on the metering results.

[0003] In addition, related technical means are fragmented at the data processing level, and the links of abnormality identification, responsibility analysis and refund execution are often handled by independent systems, forming information silos. This architecture results in a low degree of automation of the entire refund process, which is heavily dependent on manual intervention and cross-department coordination, not only resulting in long processing cycles and low efficiency, but also making it difficult to ensure the consistency of the processing results due to the lack of a unified technical decision-making model.

[0004] Therefore, how to build a technical system that can automatically analyze abnormal reasons and generate execution plans has become a key bottleneck in improving the efficiency of electricity fee refund processing. SUMMARY

[0005] The embodiments of the present application provide an automatic electricity fee refund method and system based on multi-source data, which can improve the efficiency of electricity fee refund processing, and the technical solutions are as follows: On the one hand, an automatic electricity fee refund method based on multi-source data is provided, the method comprising: in response to an electricity fee refund event trigger instruction, obtaining electricity metering data, device operating data and environmental monitoring data of a target user; based on the device operating data and the environmental monitoring data, performing operating state evaluation through a device state analysis model to obtain device health state indicators and environmental correlation analysis results, the device state analysis model being dynamically updated based on historical operating data of the device; based on the electricity metering data and the device health state indicators, performing joint analysis through an abnormal pattern recognition and refund responsibility analysis model to generate abnormal electricity consumption pattern features and a refund responsibility allocation scheme, the abnormal pattern recognition model being parameter-optimized based on electricity consumption abnormality cases, and the refund responsibility analysis model being model-adjusted based on historical responsibility definition results; Based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme, a comprehensive decision is made through the refund / refund strategy generator to generate a refund / refund execution scheme and complete the automatic refund / refund operation.

[0006] On the one hand, an automatic electricity bill refund / refund system based on multi-source data is provided, the system comprising: The acquisition module is used to acquire the target user's electricity metering data, equipment operation data, and environmental monitoring data in response to the electricity fee refund event trigger command. The status assessment module is used to assess the operational status based on the equipment operation data and environmental monitoring data through the equipment status analysis model, and obtain equipment health status indicators and environmental correlation analysis results. The equipment status analysis model is dynamically updated based on the equipment's historical operation data. The joint analysis module is used to perform joint analysis based on the electricity metering data and equipment health status indicators through anomaly pattern recognition and refund / refund responsibility analysis models to generate abnormal electricity consumption pattern characteristics and refund / refund responsibility allocation schemes. The anomaly pattern recognition model optimizes parameters based on abnormal electricity consumption cases, and the refund / refund responsibility analysis model is adjusted based on historical responsibility definition results. The comprehensive decision-making module is used to make comprehensive decisions based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme through the refund / refund strategy generator, generate a refund / refund execution scheme, and complete the automatic refund / refund operation.

[0007] On one hand, a computer device is provided, the computer device including one or more processors and one or more memories, the one or more memories storing at least one computer program, the computer program being loaded and executed by the one or more processors to implement the automatic electricity fee refund method based on multi-source data.

[0008] On the one hand, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer-readable storage medium, the computer program being loaded and executed by a processor to implement the automatic electricity fee refund and reimbursement method based on multi-source data.

[0009] On the one hand, a computer program product or computer program is provided, which includes program code stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium and executes the program code, causing the computer device to perform the above-mentioned automatic electricity fee refund and reimbursement method based on multi-source data. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a schematic diagram of the implementation environment of an automatic electricity fee refund method based on multi-source data provided in an embodiment of this application; Figure 2 This is a flowchart of an automatic electricity fee refund / refund method based on multi-source data provided in an embodiment of this application; Figure 3 This is a partial flowchart of an automatic electricity fee refund method based on multi-source data provided in an embodiment of this application; Figure 4 This is a partial flowchart of another automatic electricity fee refund method based on multi-source data provided in an embodiment of this application; Figure 5 This is a partial flowchart of another automatic electricity fee refund method based on multi-source data provided in the embodiments of this application; Figure 6 This is a schematic diagram of the structure of an automatic electricity fee refund system based on multi-source data provided in an embodiment of this application; Figure 7 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0013] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first", "second", and "n", nor are there any restrictions on the quantity or execution order.

[0014] Equipment operation data: Time-series parameters collected from the power equipment itself that characterize its electrical and mechanical operating status, such as voltage, current, power, power factor, frequency, equipment temperature, vibration data, etc.

[0015] Environmental monitoring data: Time-series parameters that reflect external conditions, collected from the physical environment where power equipment is located, such as ambient temperature, humidity, air pressure, wind speed, precipitation, salt spray concentration, and pollutant index.

[0016] Equipment health status index: A numerical or graded indicator that quantifies the degree of overall performance degradation and failure risk of equipment through comprehensive algorithmic evaluation.

[0017] Environmental correlation analysis results: A structured analysis conclusion describing whether there is a causal relationship, the intensity of the impact, the direction of the effect, and the time lag characteristics between a specific environmental factor and an equipment malfunction event.

[0018] Abnormal electricity consumption pattern characteristics: A set of features extracted from electricity metering data to formally describe abnormal electricity consumption behavior, typically including dimensions such as anomaly type, occurrence time, duration, and severity of deviation from normal levels.

[0019] Refund / Supplement Responsibility Allocation Scheme: A quantitative scheme that clearly stipulates the proportion of economic responsibility to be borne by each responsible party (such as equipment factors, environmental factors, and user factors) in electricity bill refund / supplement events, the specific refund / supplement amount, and the order of execution.

[0020] Multiscale time series analysis: a time series data processing method that aims to capture the characteristics of short-term fluctuations, medium-term cycles and long-term trends of data by analyzing data at different time resolutions (such as seconds, hours and days).

[0021] Steady-state characteristics of equipment operation: describes the long-term trend and periodic change patterns of the operating parameters of the equipment under normal operating conditions.

[0022] Transient anomaly characteristics: These describe the characteristics of sudden, transient abnormal events that briefly deviate from steady state in the operating parameters of equipment, such as spikes, sudden drops, and transient oscillations.

[0023] Multiscale time-frequency analysis: a signal processing technique that decomposes a time-series signal into frequency band components of different scales to reveal its local features at different time points and frequencies.

[0024] Trend components: Components in a time series that represent its long-term direction of change or slow drift characteristics, reflecting the overall evolution trend of the data after removing periodic and random fluctuations.

[0025] Periodic components: Regular fluctuations in a time series that repeat at fixed or quasi-fixed periods, such as daily, weekly, or seasonal cycles.

[0026] Residual components: The part of a time series that remains after removing trend and periodic components, which does not have obvious regularity and usually contains noise and unexplained anomalies.

[0027] Key environmental factor sequence: A time series of one or more environmental parameters that are strongly correlated with equipment malfunction events, obtained after screening, alignment and cleaning.

[0028] Dynamic association analysis: a method for analyzing the relationships, mutual influences, and causal directions among two or more time series variables as they change over time.

[0029] Impact Model: A formalized model that describes how environmental factors specifically affect equipment operation. It typically defines the intensity of the impact, the time lag of the effect, and the effective time window for the duration of the impact.

[0030] Relevant time period: The specific time interval during which changes in environmental factors are identified as having a significant impact on equipment malfunctions.

[0031] Key environmental parameters: A small number of core parameters selected from numerous environmental monitoring parameters that are statistically significantly correlated with the occurrence of equipment malfunctions.

[0032] Correlation coefficient: A statistic that takes values ​​between -1 and 1, used to quantify the strength and direction of the linear relationship between two variables.

[0033] Lag time parameter: A quantity representing the average time delay between the occurrence of a causal event (such as a change in environmental factors) and a consequential event (such as an equipment malfunction).

[0034] Granger causality test: a hypothesis testing method based on statistical prediction, used to determine whether past values ​​of one time series can help predict future values ​​of another time series, thus providing evidence for the existence of a causal relationship.

[0035] Cross-correlation function sequence: A sequence of correlation coefficients between two time series at different lag times, used to measure their similarity at different time offsets.

[0036] Direct impact weight: A normalized quantified value used to characterize the direct impact of a certain environmental factor on equipment malfunction events.

[0037] Electricity billing system: The core business system used by power companies to process user electricity bill calculations, bill generation, payment settlement and fund transfer.

[0038] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0039] In power system operations, the technical bottleneck caused by fragmented multi-source data in the electricity bill refund and subsidy processing is becoming increasingly prominent. Related technical solutions rely solely on single-dimensional information such as electricity meter readings for anomaly detection, failing to systematically integrate multi-source data such as equipment operating status and environmental conditions. This results in a lack of technical sophistication in tracing the root causes of anomalies. Furthermore, the differential impact of inherent equipment defects and external environmental disturbances on metering results cannot be effectively quantified and differentiated, leading to a lack of objective basis for allocating refund and subsidy responsibilities. Moreover, the anomaly identification, responsibility analysis, and refund / subsidy execution processes are handled by independent systems, creating information silos. This results in high levels of human intervention, complex cross-departmental coordination, and significantly restricts the automation level and decision-making consistency of the refund and subsidy process.

[0040] For example, after a 110kV substation was struck by lightning, the electricity meter readings showed abnormal fluctuations. The relevant technology initiated the refund / refund process solely based on the metering data deviation. However, environmental monitoring data indicated that the instantaneous voltage disturbance caused by the lightning strike was due to external environmental factors, while equipment operation data showed that the current transformer insulation performance was within the normal threshold. Due to the lack of a device condition analysis model to cross-validate the operational and environmental data, the system misjudged the environmental disturbance as an equipment failure, leading to an incorrect allocation of refund / refund responsibility. In this scenario, maintenance personnel had to manually retrieve environmental records and equipment logs scattered across different systems, tracing the root cause of the anomaly through empirical comparisons. This resulted in a prolonged refund / refund process and discrepancies in the determination of responsibility for the same event among different personnel.

[0041] If the aforementioned technical issues are not resolved, the electricity bill refund and subsidy process will continue to rely on manual experience-based judgment, and the lack of quantitative attribution capabilities for abnormal events will result in a lack of technical support for responsibility allocation schemes. Specifically, the absence of correlation analysis results between equipment health status indicators and the environment makes it impossible to distinguish the impact weight of equipment defects and external disturbances in refund and subsidy decisions, making it difficult to guarantee the technical rationality of the refund and subsidy implementation plan. Furthermore, the fragmented state of multi-source data processing will continue to hinder the realization of full-process automation, the increase in manual coordination will make the refund and subsidy processing cycle uncontrollable, and the consistency of the overall system decision-making model cannot be maintained, ultimately affecting the technical reliability of power operation services.

[0042] Figure 1 This is a schematic diagram illustrating the implementation environment of an automatic electricity bill refund / refund method based on multi-source data provided in this application embodiment. See also... Figure 1 This implementation environment may include node 110 and server 140.

[0043] Node 110 is connected to server 140 via a wireless or wired network. Node 110 has an application installed and running that supports automatic electricity bill refunds based on multi-source data.

[0044] Server 140 is a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms. Server 140 can provide background services for applications running on node 110.

[0045] Since the embodiments of this application involve a relatively complex data processing process, the computing power of node 110 is insufficient to meet the requirements under the current computing power background. Therefore, the relevant data will be sent to server 140, and server 140 will execute the technical solution provided in the embodiments of this application.

[0046] Based on the technical problems described above, this application proposes an automatic electricity bill refund / refund method based on multi-source data, see [link to relevant documentation]. Figure 2 Taking the server as the executing entity as an example, the following steps are included.

[0047] 201. In response to the electricity bill refund / refund event trigger command, acquire the target user's electricity metering data, equipment operation data, and environmental monitoring data; 202. Based on equipment operation data and environmental monitoring data, the operation status is assessed through the equipment status analysis model to obtain equipment health status indicators and environmental correlation analysis results. The equipment status analysis model is dynamically updated based on historical equipment operation data. 203. Based on electricity metering data and equipment health status indicators, a joint analysis is conducted using anomaly pattern recognition and refund / refund responsibility analysis models to generate abnormal electricity consumption pattern characteristics and refund / refund responsibility allocation schemes. The anomaly pattern recognition model optimizes parameters based on abnormal electricity consumption cases, and the refund / refund responsibility analysis model is adjusted based on historical responsibility definition results. 204. Based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme, a comprehensive decision is made through the refund / refund strategy generator to generate a refund / refund execution scheme and complete the automatic refund / refund operation.

[0048] In electricity bill refund and subsidy processing, acquiring multi-source data is a fundamental step in achieving a closed-loop technology. Among these, the equipment status analysis model refers to the computational framework used to assess the operating status of equipment. In practical applications, it can be implemented using Kalman filters based on historical data or adaptive neural fuzzy inference systems. For example, state estimation can be performed using a Kalman filter, or dynamic prediction can be executed using an ANFIS model. Its main purpose is to achieve a quantitative assessment of equipment health status. Furthermore, the anomaly pattern recognition model refers to the algorithmic mechanism used to detect abnormal electricity consumption. Specifically, it can be implemented using reinforcement learning methods based on simulated data or rule engines based on expert knowledge. For example, anomaly detection parameters can be optimized using Q-learning, or anomaly patterns can be identified using decision tree rules. Its main purpose is to achieve automatic classification of abnormal electricity consumption behavior. In addition, the refund and subsidy responsibility analysis model refers to the decision-making system used to allocate refund and subsidy responsibilities. In practical applications, it can be implemented using case-based inference systems or support vector machine classifiers. For example, historical cases can be matched using a CBR system, or responsibility weights can be calculated using an SVM classifier. Its main purpose is to achieve objective quantification of responsibility allocation. The withdrawal / replacement strategy generator refers to the decision-making module used to generate withdrawal / replacement execution plans. It can be implemented using constraint-based optimization algorithms or Markov decision processes, such as solving for the optimal withdrawal / replacement plan through linear programming or formulating execution strategies using an MDP model. Its main purpose is to automate the execution of withdrawal / replacement operations. Therefore, this application systematically solves the problems of anomaly root cause tracing bias and subjectivity in responsibility allocation caused by data fragmentation in traditional methods by constructing a multi-source data-driven automated decision-making system, achieving a closed-loop technology across the entire chain from data acquisition to withdrawal / replacement execution.

[0049] After the electricity bill refund / refund event trigger command is responded to, the target user's electricity metering data, equipment operation data, and environmental monitoring data are acquired. This data acquisition mechanism breaks through the limitations of traditional single metering data, providing a complete data foundation for tracing the root cause of anomalies by integrating equipment status and external environmental disturbance information. Based on the acquired equipment operation data and environmental monitoring data, the equipment status analysis model performs an operational status assessment. This model is dynamically updated based on historical equipment operation data, thereby generating equipment health status indicators and environmental correlation analysis results. Among them, the equipment health status indicators objectively reflect the degree of defects in the equipment itself, while the environmental correlation analysis results quantify the real-time impact intensity and lag period of environmental factors on equipment operation, avoiding misjudging environmental disturbances as equipment failures. Furthermore, electricity metering data and equipment health status indicators are input into an anomaly pattern recognition model and a refund / refund responsibility analysis model for joint analysis. The anomaly pattern recognition model optimizes parameters based on historical electricity anomaly cases to improve the accuracy of anomaly type identification, while the refund / refund responsibility analysis model dynamically adjusts weights based on historical responsibility determination results. This generates anomaly electricity consumption pattern characteristics that include the time period, type, and severity of abnormal electricity consumption, and outputs a refund / refund responsibility allocation scheme that distinguishes the responsibility weights of equipment factors, environmental factors, and user factors. Based on the impact patterns and refund / refund responsibility allocation scheme in the environmental correlation analysis results, the refund / refund strategy generator performs a comprehensive decision-making operation, fusing multi-source data into decision input features. Through a strategy matching mechanism, it generates a refund / refund execution scheme that includes refund / refund amount allocation, execution sequence, and verification mechanism. Finally, the refund / refund operation is automatically completed through the electricity billing system, and the accuracy of the results is verified, achieving a closed-loop process from technical analysis to business execution.

[0050] Specifically, the equipment status analysis model can be implemented as a time-series analysis model based on a Long Short-Term Memory (LSTM) network. It uses a sliding window mechanism to dynamically extract features from equipment operation data and updates the model parameters every 24 hours based on historical operation data to adapt to equipment aging trends. The withdrawal and compensation strategy generator can adopt a strategy matching engine based on a decision tree algorithm. This engine has a pre-stored set of withdrawal and compensation rules for different combinations of responsibility weights. When the responsibility weight coefficient and the severity of the anomaly are correlated in the input environment analysis results, the optimal execution strategy is automatically matched. For example, in the event of an abnormal meter reading for a residential user, the system obtains the user's meter reading sequence, distribution transformer operating current data, and weather station temperature records. The equipment status analysis model identifies a strong correlation between the abnormal rise in transformer winding temperature and high ambient temperature, generating an equipment health status index of moderate degradation and an environmental correlation analysis result of a temperature influence weight of 0.7. The anomaly pattern recognition model, combined with the characteristics of electricity metering deviation, determines the anomaly type as environmentally induced metering drift. Based on this, the refund and compensation responsibility analysis model allocates a responsibility weight of 60% for environmental factors, 30% for equipment factors, and 10% for user factors. The refund and compensation strategy generator automatically generates an execution instruction that assigns the main refund and compensation responsibility to environmental factors based on this allocation scheme, and completes the transfer of refund and compensation amount within 48 hours through the electricity billing system API interface.

[0051] Therefore, this method eliminates the attribution bias caused by data fragmentation in traditional technologies through the systematic fusion of multi-source data. The dynamic update mechanism of the equipment status analysis model ensures the timeliness and adaptability of the operational status assessment and avoids misjudgments caused by the failure to include equipment aging trends in the analysis. The joint analysis architecture of anomaly pattern recognition and return-to-work responsibility analysis model realizes the objective quantification of responsibility weights, distinguishes the degree of influence of different factors such as equipment defects and environmental disturbances, and solves the problem of strong subjectivity in responsibility allocation. The integration of the return-to-work strategy generator and the automatic execution link completely replaces the manual coordination process, significantly shortens the return-to-work processing cycle, and ensures the consistency and traceability of the processing results through the decision-making closed-loop design, effectively overcoming the shortcomings of related technologies such as inability to trace the root cause of anomalies, inability to quantify responsibility allocation, and low efficiency of manual processing.

[0052] Specifically, in some of the above-mentioned implementation methods, an operational status assessment based on equipment operation data and environmental monitoring data is proposed to obtain equipment health status indicators and environmental correlation analysis results through equipment status analysis models. However, in the implementation process, due to the lack of in-depth decomposition of equipment operation data and systematic correlation mechanism with environmental factors, the equipment health status assessment remains only at the surface level. It is unable to effectively separate the long-term steady-state characteristics of equipment operation from instantaneous abnormal disturbances. At the same time, the impact of environmental factors on equipment abnormalities is simplified, making it difficult to quantify the dynamic correlation between changes in environmental factors and equipment abnormal events. This results in fuzzy equipment health status indicators and inconsistent environmental correlation analysis results, ultimately affecting the accuracy of subsequent compensation and refund responsibility allocation.

[0053] Of course, the responsibility allocation made by the technical solutions provided in this application is for reference by relevant personnel, and the specific responsibility shall be determined and judged by relevant personnel.

[0054] In response, this application further proposes the following technical solution, see [link to technical solution]. Figure 3 Taking the server as the executing entity as an example, the following steps are included.

[0055] 301. Using the equipment status analysis model, perform multi-scale time series analysis on the equipment operation data to obtain the steady-state characteristics and transient anomaly characteristics of the equipment operation; 302. Using the equipment status analysis model, conduct environmental correlation analysis on environmental monitoring data and transient anomaly characteristics to obtain the impact pattern and correlation time period of environmental factors on equipment operation; 303. Based on the steady-state characteristics, transient anomaly characteristics and environmental impact patterns of the equipment, the equipment health status index and environmental correlation analysis results are generated through the equipment status analysis model.

[0056] Multi-scale time series analysis refers to the process of extracting features from equipment operation data across multiple time scales. This can be achieved using generalization methods such as wavelet transform, empirical mode decomposition, or Fourier series expansion. Its purpose is to separate operational features from the raw data at different time scales, avoiding feature omissions caused by single-scale analysis. Steady-state characteristics of equipment operation can be understood as quantitative indicators reflecting the long-term stable operating state of the equipment. These can be constructed based on the statistical characteristics of trend and periodic components, aiming to characterize the equipment's continuous operating capability under normal conditions. Transient anomaly characteristics refer to the set of features of sudden disturbances during equipment operation. These can be identified through abrupt change detection of residual components or threshold discrimination methods. The key is to locate instantaneous failure events; environmental correlation analysis can be understood as establishing a dynamic correlation mechanism between environmental factors and equipment anomalies, which can be achieved using time series matching algorithms or dynamic time warping techniques, aiming to solve the asynchronous problem between environmental data and equipment anomalies on a time scale; environmental impact model refers to the functional representation of the effect of environmental factors on equipment operation, which can be constructed based on a combination model of correlation coefficient and lag time parameters, aiming to quantify the intensity and temporal characteristics of the impact of environmental factors on equipment anomalies; the correlation period refers to the effective time window of environmental factors affecting equipment operation, which can be determined by time boundary expansion algorithms, aiming to clarify the time range of environmental factors participating in responsibility analysis.

[0057] Specifically, this scheme first performs multi-scale time-series analysis on equipment operation data through an equipment status analysis model, decomposing the raw data into feature components at different time scales, thereby separating steady-state characteristics that characterize long-term stability and transient anomaly characteristics that reflect instantaneous disturbances. On this basis, the transient anomaly characteristics are dynamically matched with environmental monitoring data within a time window, and the correlation index is calculated through cross-correlation analysis and causality tests are performed to construct the impact pattern of environmental factors on equipment operation and the associated time periods. Finally, the steady-state characteristics, transient anomaly characteristics, and environmental impact patterns of equipment operation are fused from multiple sources. The steady-state characteristics ensure the long-term reliability of health assessment, the transient anomaly characteristics identify the immediate impact of sudden failures, and the environmental impact patterns quantify the contribution of external factors. The equipment health status index generated by the collaboration of these three factors can simultaneously cover the equipment's own state and external environmental disturbances, so that the environmental correlation analysis results have quantifiable time-series boundaries and impact intensity.

[0058] As a specific implementation method, the scheme of this application is implemented as follows: The equipment status analysis model uses discrete wavelet transform to decompose the equipment operation data into three levels, obtaining trend components, periodic components, and residual components; the rolling mean standard deviation is calculated based on the trend components as a long-term stability indicator, the harmonic component amplitude is extracted based on the periodic components as a periodic fluctuation mode, and the peak detection of the residual components identifies instantaneous abnormal events; the environmental correlation analysis module dynamically matches the timestamps of transient abnormal features with environmental monitoring data using a time window, and the window width is adaptively adjusted according to the equipment sampling frequency; by calculating the cross-correlation function between the environmental factor sequence and the abnormal event sequence, the maximum correlation coefficient and the corresponding lag time are extracted as correlation indicators; the Granger causality test is used to verify the driving effect of environmental factor changes on equipment abnormal events, and an environmental impact function is constructed with the correlation coefficient as the weight and the lag time as the offset; finally, the steady-state characteristics of equipment operation, transient abnormal characteristics, and environmental impact function are input into the decision fusion unit to generate an equipment health status index that includes health scores and environmental impact boundaries.

[0059] Through the above scheme, this application achieves a refined hierarchical assessment of equipment operating status. The effective separation of steady-state and transient anomaly characteristics avoids ambiguity in health status assessment. The dynamic matching of time windows and causal verification mechanisms in environmental correlation analysis ensure the quantitative accuracy of environmental impact models, thus providing clear equipment health status indicators and environmental correlation analysis results for the allocation of compensation responsibilities, improving the accuracy of anomaly root cause tracing and the objectivity of responsibility determination. Specifically, in some embodiments of this application, multi-scale time-series analysis of equipment operating data is proposed to obtain steady-state and transient anomaly characteristics. However, in its implementation, the specific implementation method of multi-scale time-series analysis is unclear, resulting in insufficient accuracy in extracting steady-state and transient anomaly characteristics of equipment operating data. This makes it impossible to effectively separate long-term trends, periodic fluctuations, and instantaneous anomalies, thereby affecting the accuracy of environmental correlation analysis and the reliability of compensation responsibility allocation.

[0060] In response, this application further proposes to perform multi-scale time-frequency analysis on equipment operation data to obtain the trend components, periodic components and residual components of equipment operation; Based on trend components and periodic components, the long-term stability indicators and periodic fluctuation patterns of equipment operation are determined through steady-state characteristic analysis. Based on residual components, transient anomaly detection identifies transient abnormal events and short-term disturbance characteristics in equipment operation; Based on long-term stability indicators, periodic fluctuation patterns, instantaneous abnormal events, and short-term disturbance characteristics, steady-state characteristics and transient abnormal characteristics of equipment operation are generated.

[0061] In practical applications, multi-scale time-frequency analysis refers to the technique of jointly analyzing equipment operation data in the frequency and time domains from different time scales. It can be achieved using multi-resolution analysis such as wavelet transform, empirical mode decomposition, or Fourier transform. The purpose is to separate signal components at different time scales and avoid feature confusion caused by single-scale analysis. Among them, the trend component refers to the low-frequency component in the equipment operation data that reflects the long-term trend of change. It can be a smooth curve extracted by moving average filtering or low-pass filtering, and its purpose is to characterize the long-term evolution law of equipment performance. Periodic components refer to components in equipment operation data that exhibit regular fluctuations. They can be the dominant frequency components identified through spectrum analysis or the periodic patterns extracted through autocorrelation functions. The purpose is to capture the periodic behavioral characteristics of equipment operation. The residual component refers to the random fluctuation part remaining in the equipment operation data after removing trend and periodic components. It can be a high-frequency signal obtained through differential operation or high-pass filtering, with the aim of focusing on instantaneous anomalies and random disturbances. Steady-state characteristic analysis refers to a quantitative evaluation method for the characteristics of stable operating state of equipment. It can be achieved by statistical process control or long-term trend fitting techniques, with the aim of determining the performance benchmark of equipment under normal operating conditions. Long-term stability indicators are quantitative parameters that characterize the long-term operational stability of equipment. They can be the slope of a trend line or the variance coefficient, and their purpose is to assess the long-term consistency of equipment performance. Periodic fluctuation patterns refer to the characteristic description of periodic changes in equipment operation. These patterns can be periodic amplitude, frequency, or phase information, and their purpose is to identify the regular fluctuation characteristics of equipment operation. Transient anomaly detection refers to the identification technology for instantaneous abnormal events during equipment operation. It can be implemented using threshold detection, mutation point detection, or machine learning anomaly detection algorithms, with the aim of timely detecting sudden faults in equipment operation. Transient abnormal events refer to abnormal fluctuations in equipment operation that are short in duration but large in amplitude. These can be sudden drops in voltage or sudden increases in current. The purpose is to capture transient faults in equipment operation. Short-term disturbance characteristics refer to the short-duration fluctuation characteristics during equipment operation. These can be harmonic content or flicker indicators, and their purpose is to characterize the equipment's response characteristics to short-term disturbances. Generating steady-state characteristics of equipment operation refers to integrating long-term stability indicators and periodic fluctuation patterns to form a description of the steady-state operation characteristics of the equipment. It can be achieved by weighted fusion or multi-dimensional feature vector construction methods, with the aim of comprehensively characterizing the operating characteristics of the equipment under steady-state conditions. Generating transient anomaly features refers to integrating instantaneous abnormal events and short-term disturbance features to form a description of equipment transient anomaly features. This can be achieved using event clustering or feature association analysis methods, with the aim of accurately depicting the instantaneous abnormal state of equipment during operation.

[0062] Specifically, the proposed solution first performs multi-scale time-frequency analysis on the equipment operation data, decomposing the raw data into trend components, periodic components, and residual components to ensure effective separation of signal components at different time scales. Then, based on the separated trend and periodic components, long-term stability indicators and periodic fluctuation patterns of the equipment operation are determined through steady-state characteristic analysis, quantifying the equipment's operating characteristics under steady-state conditions. Simultaneously, based on the residual components, transient anomaly detection identifies instantaneous abnormal events and short-term disturbance characteristics of the equipment operation, focusing on capturing instantaneous changes during equipment operation. Finally, the long-term stability indicators, periodic fluctuation patterns, instantaneous abnormal events, and short-term disturbance characteristics are integrated to generate steady-state and transient anomaly characteristics of the equipment operation. This ensures that the steady-state characteristics fully reflect the long-term operating state of the equipment, and the transient anomaly characteristics accurately characterize instantaneous disturbances, providing high-precision input for subsequent environmental correlation analysis.

[0063] As a preferred embodiment, the specific implementation of this application is as follows: When analyzing the operating data of a transformer in a substation, discrete wavelet transform is used to decompose the collected voltage and current data into multi-scale components, obtaining trend components, periodic components, and residual components; the long-term slope of the transformer oil temperature is calculated using the trend components as a long-term stability index, while the load periodic fluctuation pattern is identified through the periodic components; a sliding window standard deviation detection algorithm is applied to the residual components to identify voltage drop events as instantaneous abnormal events, and harmonic distortion rate is extracted as a short-term disturbance feature; finally, the long-term stability index and the periodic fluctuation pattern are fused into a steady-state feature vector, and the instantaneous abnormal events and short-term disturbance features are correlated into a transient abnormal feature set for subsequent analysis.

[0064] Through the above-mentioned scheme, this application effectively improves the feature extraction accuracy of equipment operation data and achieves the separation of long-term trends, periodic fluctuations, and instantaneous anomalies, thereby improving the accuracy of environmental correlation analysis and the reliability of responsibility allocation. Specifically, in some embodiments of this application, the generation of steady-state features and transient anomaly features of equipment operation is proposed to support equipment status assessment. However, in its implementation, due to the lack of a multi-dimensional fusion mechanism for long-term stability indicators and periodic fluctuation patterns, as well as a spatiotemporal correlation analysis method for instantaneous abnormal events and short-term disturbance features, the steady-state features rely only on a simple combination of single indicators and cannot fully reflect the long-term stability and periodic patterns of equipment operation. At the same time, the transient anomaly features fail to effectively capture the spatiotemporal distribution patterns and intensity characteristics of abnormal events, resulting in coarse feature generation that is susceptible to noise interference, making it difficult to distinguish between the steady-state characteristics and instantaneous abnormal behavior of equipment operation, thereby affecting the accuracy of subsequent equipment health status assessment and the reliability of anomaly cause tracing.

[0065] In response, this application further proposes to generate steady-state and transient abnormal characteristics of equipment operation based on long-term stability indicators, periodic fluctuation patterns, transient abnormal events, and short-term disturbance characteristics, including: Multi-dimensional feature fusion is performed on long-term stability indicators and cyclical fluctuation patterns to generate steady-state characteristics of equipment operation. The multi-dimensional feature fusion includes a quantitative assessment of trend stability and cyclical consistency. Spatiotemporal correlation analysis is performed on transient abnormal events and short-term disturbance characteristics to generate transient abnormal features. The spatiotemporal correlation analysis includes abnormal event clustering analysis and disturbance intensity pattern recognition.

[0066] Multi-dimensional feature fusion refers to the technical means of comprehensively integrating multi-source operational indicators to form a unified feature representation. This can be achieved using methods such as weighted average, principal component analysis, or factor analysis. The aim is to comprehensively capture the long-term trend stability and periodic fluctuation patterns of equipment operation by quantitatively assessing trend stability and periodic consistency. Trend stability can be understood as an indicator of the degree to which the equipment's operational trend deviates from its normal state. It can be quantified by calculating the standard deviation or coefficient of variation of operational data, aiming to objectively reflect the stability level of equipment operation. Periodic consistency can be understood as a measure of the regularity of equipment's periodic fluctuations, which can be evaluated using autocorrelation function analysis or spectral analysis. The purpose is to identify periodic anomaly patterns; spatiotemporal correlation analysis refers to the technical method of analyzing the relationship between anomaly events in time and space dimensions. It can be implemented using density-based clustering algorithms or spatiotemporal correlation analysis, with the aim of revealing the distribution pattern of anomaly events; anomaly event clustering analysis can be understood as the process of grouping anomaly events with similar spatiotemporal characteristics into one category. It can be implemented using DBSCAN or hierarchical clustering algorithms, with the aim of mining potential environmental or equipment-induced patterns; disturbance intensity pattern recognition can be understood as a technique for quantifying the severity of short-term disturbance events. It can be implemented by analyzing the characteristics of disturbance amplitude and duration, with the aim of characterizing the intensity characteristics of disturbance events.

[0067] Specifically, the proposed solution first accepts long-term stability indicators, periodic fluctuation patterns, transient abnormal events, and short-term disturbance characteristics as input data. Then, it performs multi-dimensional feature fusion on the long-term stability indicators and periodic fluctuation patterns, quantifying trend stability and periodic consistency to integrate multi-source indicators into a comprehensive steady-state characteristic that fully characterizes the long-term operating status of the equipment. Simultaneously, it performs spatiotemporal correlation analysis on transient abnormal events and short-term disturbance characteristics, identifying spatiotemporal clustering patterns of abnormal events through cluster analysis and quantifying the severity of disturbance events through disturbance intensity patterns, thereby generating transient abnormal characteristics that reflect the dynamic characteristics of abnormal behavior. Finally, the generated steady-state and transient abnormal characteristics together constitute a complete feature basis for equipment status assessment, providing a basis for subsequent health status indicator calculations and anomaly cause tracing.

[0068] As a preferred embodiment, the specific implementation of the solution in this application is as follows: When generating steady-state characteristics of equipment operation, principal component analysis is used to reduce the dimensionality of long-term stability indicators and periodic fluctuation patterns, and the main components are extracted as steady-state feature vectors. The trend stability is quantified by calculating the moving standard deviation of the operating data, and the periodic consistency is evaluated by Fourier transform analysis of the stability of the periodic components. When generating transient anomaly characteristics, the density-based DBSCAN clustering algorithm is applied to perform spatiotemporal clustering of instantaneous anomaly events, grouping events with similar time intervals and spatially related locations into the same cluster. At the same time, the disturbance intensity pattern is identified by calculating the amplitude peak and duration of the disturbance event, thus forming transient anomaly characteristics.

[0069] Through the above scheme, this application realizes the refined generation of equipment operation characteristics, effectively solves the problems of insufficient multi-dimensional information fusion and lack of spatiotemporal correlation in the feature extraction process, so that the steady-state characteristics of equipment operation can fully reflect the long-term stability and periodic patterns, and the transient anomaly characteristics can capture the spatiotemporal distribution and intensity characteristics of abnormal events, thereby improving the accuracy of equipment health status assessment and providing reliable technical support for tracing the cause of anomalies.

[0070] Specifically, in some of the embodiments described above in this application, an operational status assessment based on equipment operation data and environmental monitoring data is proposed. However, in the process of its implementation, the correlation between environmental factors and abnormal equipment events is often difficult to quantify accurately and capture dynamically, and there is a lack of systematic analysis methods. This makes it impossible to construct the impact pattern and related time period of environmental factors on equipment operation, thereby affecting the reasonable consideration of environmental factors and the fairness of responsibility allocation in the compensation and refund responsibility analysis.

[0071] In this regard, this application further proposes the following steps for environmental correlation analysis: Key environmental parameters were extracted from environmental monitoring data to obtain a sequence of key environmental factors affecting equipment operation; Based on transient anomaly characteristics and key environmental factor sequences, the correlation between environmental factor changes and equipment anomaly events is determined through dynamic correlation analysis. The dynamic correlation analysis is modeled based on the spatiotemporal correlation between historical environmental monitoring data and equipment anomaly records. Based on the correlation, an impact model of environmental factors on equipment operation is constructed, and the correlation time periods of environmental factors on equipment operation are generated.

[0072] Among these, key environmental parameter extraction refers to screening a subset of parameters that are significantly related to the equipment's operating status from multi-source environmental monitoring data. This can be achieved using threshold screening methods based on statistical significance tests or feature grouping algorithms based on unsupervised clustering. The aim is to eliminate interference from irrelevant environmental variables and focus on core parameters that have a substantial impact on equipment anomalies. Dynamic correlation analysis can be understood as quantifying the temporal correlation and causal relationship between changes in environmental factors and equipment anomalies. This can be achieved using sliding window cross-correlation analysis or causal inference models based on time series. The aim is to dynamically capture the temporal characteristics and intensity changes of the impact of environmental factors on equipment operation. Constructing an impact model of environmental factors on equipment operation refers to establishing a quantitative functional relationship between environmental factors and equipment anomalies. This can be achieved based on parametric regression models or nonlinear mapping functions. The aim is to transform the correlation into an operable input for responsibility analysis, providing an objective basis for decision-making regarding compensation or refunds.

[0073] Specifically, the proposed solution first extracts key environmental parameters from environmental monitoring data to identify core environmental factors highly correlated with equipment anomalies, thus simplifying the input data dimensions for subsequent analysis. Then, based on transient anomaly characteristics and key environmental factor sequences, dynamic correlation analysis is performed. This involves modeling the spatiotemporal correlation between historical environmental monitoring data and equipment anomaly records to quantify the correlation index between environmental factor changes and equipment anomaly events. Finally, based on the correlation, an impact pattern of environmental factors on equipment operation is constructed, generating correlated time periods, and transforming the quantified results into quantifiable responsibility analysis parameters. This phased analysis process ensures the systematic and dynamic adaptability of environmental correlation analysis, making the consideration of environmental factors in the allocation of compensation and refund responsibilities more objective and comprehensive.

[0074] As a specific implementation method, the solution of this application is implemented as follows: In the key environmental parameter extraction stage, the system acquires temperature, humidity, and wind speed data from the environmental monitoring sensor network deployed around the power facilities. By calculating the correlation strength between each parameter and historical equipment anomaly events, temperature is selected as a key environmental factor and reconstructed into a time series synchronized with the equipment operation data. In the dynamic correlation analysis stage, the temperature change sequence is matched with the transient anomaly characteristics of the equipment through a time window, and the cross-correlation function is calculated to determine the maximum correlation coefficient and the corresponding lag time parameter. In the impact model construction stage, based on the correlation coefficient and lag time parameter, an impact strength model of temperature change on equipment anomalies is established, and the correlation period of environmental factors on equipment operation is generated, which covers the complete process of environmental impact from start to finish.

[0075] Through the above scheme, this application can accurately quantify the correlation between environmental factors and equipment abnormal events, dynamically capture the temporal characteristics of environmental impact, thereby constructing the impact pattern and related time period of environmental factors on equipment operation, providing an objective basis for the analysis of electricity fee refund and subsidy responsibility, and improving the fairness and accuracy of refund and subsidy responsibility allocation.

[0076] In the process of electricity fee refunds and subsidies in the power system, the relevant technologies have significant limitations in analyzing the correlation between environmental factors and equipment operation anomalies. In some embodiments described above in this application, environmental correlation analysis is proposed to determine the impact patterns of environmental factors on equipment operation. However, in its implementation, the extraction of key environmental parameters lacks a screening mechanism, leading to irrelevant environmental data interfering with the correlation analysis results and affecting the accuracy of subsequent responsibility allocation.

[0077] In response, this application further proposes extracting key environmental parameters from environmental monitoring data to obtain a sequence of key environmental factors affecting equipment operation, including: Based on preset environmental parameter association rules, key environmental parameters that have a preset threshold of correlation with equipment anomalies are selected from environmental monitoring data. The environmental parameter association rules are determined based on the correlation analysis results of historical environmental monitoring data and equipment anomaly records. The selected key environmental parameters are reconstructed into time series to obtain an environmental parameter series with a unified time reference, which is consistent with the time reference for the acquisition of equipment operation data. The environmental parameter sequence is standardized and outlier correction is performed to obtain the key environmental factor sequence.

[0078] Among them, the preset environmental parameter association rules refer to the quantitative screening criteria established based on historical data. These can be implemented using threshold judgment conditions obtained through statistical correlation analysis or machine learning model training. The purpose is to eliminate environmental noise with low correlation and ensure that the screening process is based on objective historical data rather than subjective experience. The unified time benchmark refers to the time coordinate system shared by the environmental parameter sequence and equipment operation data. This can be implemented using timestamp alignment or resampling interpolation techniques. The purpose is to solve the problem of spatiotemporal misalignment caused by differences in the original sampling frequency. Standardization processing and outlier correction refer to the data normalization and quality optimization of the environmental parameter sequence. This can be implemented using Z-score standardization combined with a sliding window anomaly detection algorithm. The purpose is to eliminate dimensional differences and correct data outliers, ensuring that the environmental factor sequence truly reflects the trend of environmental change.

[0079] Specifically, the proposed solution effectively filters environmental monitoring data unrelated to equipment anomalies through a screening mechanism based on environmental parameter association rules, avoiding interference from low-correlation parameters in subsequent analysis. Building upon this, a time series reconstruction step adjusts the screened key environmental parameters to a time base consistent with the equipment operation data, providing an accurate temporal basis for the spatiotemporal matching of environmental factor changes and equipment anomaly events. Finally, standardization and outlier correction steps optimize the data quality of the reconstructed sequence, eliminating dimensional differences and correcting outliers, enabling the environmental factor sequence to objectively represent environmental change characteristics. These three steps form a progressive data processing flow: first, ensuring the targeting of the screening from the parameter dimension; second, guaranteeing data synchronization from the time dimension; and finally, improving the reliability of the sequence from the quality dimension, collectively constructing a high-precision key environmental factor sequence.

[0080] As a specific implementation method, the solution of this application is implemented as follows: In the substation electricity fee refund scenario, the environmental parameter association rule is determined based on historical data to be that the correlation coefficients between the three parameters of temperature, humidity and wind speed and equipment anomalies must be greater than 0.7; after screening, the data collected by the temperature sensor every 10 minutes is reconstructed by linear interpolation to keep it consistent with the time benchmark of the 5-minute collection cycle of the electricity metering data; then the Z-score method is used to standardize the reconstructed sequence, and the 3σ principle is used to identify and correct outliers that exceed the normal fluctuation range, and finally generate a key environmental factor sequence that reflects the real environmental impact.

[0081] Through the above technical solution, this application effectively solves the problem of correlation analysis bias caused by the failure to extract key environmental parameters, and enables the construction of the impact model of environmental factors on equipment operation to be based on high-quality data, thereby ensuring the objectivity and accuracy of the allocation of compensation responsibility and avoiding the deviation in responsibility definition caused by environmental data interference.

[0082] In practical applications, some of the embodiments described above in this application propose dynamic correlation analysis to determine the relationship between changes in environmental factors and abnormal equipment events. However, in its implementation, time window matching lacks a dynamic adaptation mechanism for the timestamps of abnormal equipment events and the sampling frequency of environmental factor sequences, the correlation index only focuses on static correlation and ignores the time lag characteristics, and the causal analysis does not introduce statistical verification methods. This results in time misalignment and attribution bias in the assessment of the impact of environmental factors on equipment operation, making it difficult to support the quantification of the allocation of compensation responsibility.

[0083] In response, this application further proposes a method based on transient anomaly characteristics and key environmental factor sequences to determine the correlation between changes in environmental factors and equipment anomaly events through dynamic correlation analysis, including: Time window matching is performed on transient anomaly features and key environmental factor sequences to obtain the temporal correspondence between environmental factor changes and equipment anomaly events. The time window matching is dynamically adjusted based on the timestamp of the equipment anomaly event and the sampling frequency of the environmental factor sequence. Based on the time correspondence, the correlation index between changes in environmental factors and abnormal equipment events is determined. The correlation index includes the correlation coefficient and the lag time parameter. Based on the correlation index, a causal analysis of environmental factors and equipment malfunction events was conducted to obtain the correlation between changes in environmental factors and equipment malfunction events. The Granger causality test method was used for the causal analysis.

[0084] Among them, time window matching refers to a mechanism for dynamically adjusting the size of the time window, which can be implemented using algorithms based on timestamp alignment and sampling frequency compensation, such as sliding window technology or adaptive time scaling methods. Its purpose is to ensure that the environmental factor sequence corresponds to the equipment anomaly event in the time dimension, avoiding time series deviations caused by differences in data collection frequency or the suddenness of events. The correlation index includes the correlation coefficient and the lag time parameter. The correlation coefficient is a statistical measure that quantifies the degree of linear correlation between changes in environmental factors and equipment anomalies. It can be implemented using Pearson correlation coefficient or Spearman rank correlation coefficient. The lag time parameter is an indicator that characterizes the time required for the impact of changes in environmental factors to be transmitted to equipment anomalies. It can be determined through cross-correlation analysis or phase difference calculation. Its purpose is to simultaneously capture the strength of the correlation and the dynamic characteristics of the time series, avoiding the limitation that a single indicator cannot reflect the influence process. Causal analysis adopts the Granger causality test method, which is a statistical test method based on predictive ability. It can be implemented based on the vector autoregression model. It verifies the causal relationship by testing whether the environmental factor sequence helps to predict the equipment anomaly event sequence. Its purpose is to distinguish between true causal association and accidental association, and to prevent non-causal associations from being misjudged as the basis for responsibility.

[0085] Specifically, the proposed solution first establishes a temporal correspondence by matching transient anomaly characteristics and key environmental factor sequences within a time window, providing a reliable time benchmark for correlation analysis. Based on this temporal correspondence, a correlation index is calculated, using the correlation coefficient to measure the linear correlation strength between environmental factor changes and equipment anomaly events. Furthermore, a lag time parameter is used to capture the temporal dynamics of influence transmission, thus comprehensively characterizing the time-delayed pattern of environmental factors affecting equipment anomalies. Finally, the correlation index is used to perform a Granger causality test, statistically validating the predictive ability of the environmental factor change sequence for the equipment anomaly event sequence, effectively distinguishing between causal relationships and accidental correlations. This step-by-step processing mechanism forms a complete logical chain from time alignment and correlation quantification to causal verification, ensuring the temporality and attribution reliability of the environmental factor impact assessment.

[0086] As a specific implementation method, the solution of this application is implemented as follows: When an overload abnormal event occurs in the distribution transformer, the system acquires the timestamp of the abnormal event and the ambient temperature monitoring sequence; the time window matching module dynamically adjusts the window size according to the sampling frequency of the temperature sequence, for example, expanding the fixed time window into an adaptive window to align the temperature change with the abnormal event; then, the correlation coefficient and lag time parameter between the temperature change and the abnormal event are calculated, and the lag time corresponding to the maximum correlation value is determined through cross-correlation analysis; finally, the Granger causality test is performed to verify whether the temperature rise sequence has a statistically significant predictive ability for the abnormal event sequence, thereby confirming the causal influence of environmental factors.

[0087] Through the above technical solutions, the accuracy of the assessment of the impact of environmental factors on equipment operation is improved, the problem of timing misalignment is solved, and the attribution bias is significantly reduced, thus providing a scientific and objective quantitative basis for the allocation of responsibility for electricity fee refunds and subsidies.

[0088] In some of the embodiments described above in this application, a correlation index based on time correspondence is proposed to quantify the association between environmental factors and equipment anomaly events. However, in its implementation, there is a lack of a method for calculating the correlation index, especially how to automatically extract the correlation coefficient and lag time parameter from time series data. This results in the correlation analysis results being affected by subjective experience, making it impossible to accurately capture the dynamic temporal relationship between changes in environmental factors and equipment anomaly events. Consequently, it affects the reliability of the subsequent Granger causality test, and ultimately makes it difficult to guarantee the scientific validity of the compensation and refund responsibility allocation scheme.

[0089] In response, this application further proposes steps for determining the correlation index between changes in environmental factors and abnormal equipment events based on time correspondence, including: Based on the time correspondence, synchronous time series data of environmental factor change sequences and equipment abnormal event sequences are constructed; Cross-correlation analysis is performed on synchronous time series data to calculate the cross-correlation function values ​​at different lag times, thus obtaining a cross-correlation function sequence; the maximum cross-correlation value is extracted from the cross-correlation function sequence as the correlation coefficient, and the lag time corresponding to the maximum cross-correlation value is extracted as the lag time parameter; Based on the correlation coefficient and lag time parameter, the correlation index between changes in environmental factors and abnormal equipment events is determined. The correlation index is jointly defined by the correlation strength measured by the correlation coefficient and the correlation time series characteristics measured by the lag time parameter.

[0090] Synchronous time series data refers to a data structure that strictly aligns environmental factor change sequences with equipment anomaly event sequences under a unified time reference. This can be achieved using dynamic timestamp matching algorithms or linear interpolation methods, aiming to eliminate data mismatch caused by sampling frequency differences or timestamp offsets, providing a reliable time synchronization basis for correlation analysis. Cross-correlation analysis calculates the statistical correlation between two time series at different time offsets. This can be achieved using fast Fourier transform frequency domain calculations or direct time domain convolution methods, aiming to systematically explore the dynamic process of the impact of environmental factor changes on equipment anomaly events and reveal potential time-series dependencies. Cross-correlation function sequences refer to the set of cross-correlation values ​​calculated at different lag times, which can be represented as discrete time series or continuous function forms. The purpose of this method is to quantify the changing trend of correlation strength over time. The maximum cross-correlation value refers to the global peak value in the cross-correlation function sequence, which can be automatically extracted through traversal search or optimization algorithms. Its purpose is to objectively quantify the strongest correlation between environmental factors and equipment anomalies, avoiding the subjectivity of manually setting thresholds. The lag time parameter refers to the time offset corresponding to the maximum cross-correlation value, which can be expressed as a time unit value. Its purpose is to capture the time delay characteristics of environmental influences transmitting to equipment anomalies, providing key time-series evidence for causal inference. The correlation index is a comprehensive quantitative representation that integrates correlation strength and time-series characteristics. It can be defined using a weighted function or vector combination method. Its purpose is to ensure that the correlation analysis results reflect both the degree of correlation and the dynamic process, thereby supporting the accuracy of subsequent causal tests.

[0091] Specifically, the proposed solution first constructs synchronous time series data based on time correspondence to ensure strict alignment between environmental factor change sequences and equipment anomaly event sequences under a unified time benchmark, eliminating data mismatch issues. On this basis, cross-correlation analysis is performed on the synchronous time series data to calculate cross-correlation function values ​​at different lag times, systematically exploring the dynamic process of the impact of environmental factor changes on equipment anomaly events. Subsequently, the maximum cross-correlation value and its corresponding lag time are automatically extracted from the cross-correlation function sequence as core parameters to objectively quantify the correlation strength and time delay characteristics. Finally, the correlation coefficient and lag time parameter are integrated into a comprehensive correlation index, ensuring that the correlation analysis results simultaneously contain dual information of correlation strength and time series characteristics, thereby providing input for Granger causality testing and ensuring the scientific validity of the compensation and refund responsibility allocation scheme.

[0092] As a specific implementation method, the scheme of this application is implemented as follows: In environmental monitoring data, temperature change sequence and equipment abnormal event sequence are aligned with timestamps to construct a synchronized time series; cross-correlation analysis is performed using fast Fourier transform to calculate cross-correlation function values ​​with lag times ranging from -24 hours to +24 hours; the maximum value of 0.85 is found by traversing the cross-correlation function sequence, corresponding to a lag time parameter of 3 hours; the correlation coefficient of 0.85 and the lag time parameter of 3 hours are used as correlation indicators for subsequent Granger causality tests.

[0093] Through the above scheme, this application realizes the quantification of the correlation between changes in environmental factors and abnormal equipment events, avoids the subjectivity of manually setting thresholds, accurately captures dynamic time-series relationships, improves the reliability of Granger causality tests, and thus ensures that the compensation and refund responsibility allocation scheme can scientifically distinguish the causal relationship between environmental factors and equipment abnormalities.

[0094] In some of the embodiments described above in this application, environmental correlation analysis based on correlation index is proposed. However, in its implementation, the correlation index can only reflect statistical correlation and cannot determine causal relationship, which may lead to misjudging correlation as causation, thereby affecting the accuracy of the allocation of compensation responsibility.

[0095] In response, this application further proposes a causal analysis based on correlation indices to determine the relationship between changes in environmental factors and equipment malfunctions, including: Based on the correlation index, environmental factor sequences and equipment abnormal event sequences with correlation exceeding a preset threshold are selected as input data for causal analysis. The preset threshold is dynamically adjusted based on historical correlation analysis results. Granger causality tests are performed on the input data for causal analysis to obtain associations, where the associations are characterized by the results of tests that environmental factor changes are Granger causes of equipment malfunction events.

[0096] In practical applications, correlation index screening refers to the process of screening environmental factor sequences and equipment anomaly event sequences based on preset thresholds. This can be achieved using correlation coefficients and lag time parameters as screening criteria. The aim is to filter out noisy data with low correlation, ensuring that only data with significant statistical correlation enters the subsequent causal analysis stage. Dynamic adjustment of the preset threshold refers to a mechanism that automatically adjusts the threshold based on historical correlation analysis results. This can be adaptively set based on the statistical characteristics of correlation distribution in historical data, aiming to adapt the screening criteria to the characteristics of historical data under different environmental conditions and equipment states, thereby improving the targeting and adaptability of the screening. Specifically, the Granger causality test is a method to determine causal relationships by evaluating the predictive ability of environmental factor sequences on equipment anomaly event sequences. This can be achieved using a vector autoregression model for time series predictive analysis, aiming to effectively distinguish between causal relationships and simple statistical correlations, providing a statistical basis for determining causal relationships. It is understandable that association characterization refers to the method of quantifying the impact of environmental factors using Granger causality test results. It can use the test p-value or F-statistic as a measure of the degree of causal impact. Its purpose is to directly quantify the degree of causal impact of environmental factors and provide a basis for decision-making in the subsequent allocation of compensation responsibilities.

[0097] Specifically, the proposed solution first selects data sequences with correlations exceeding a dynamic adjustment threshold as input for causal analysis, based on the correlation coefficients and lag time parameters determined in the preliminary environmental correlation analysis. This ensures that the input data has significant statistical correlations and avoids interference from low-correlation noise data. Subsequently, a Granger causality test is performed on the selected input data. This test, by constructing a time series prediction model, assesses whether environmental factor sequences can significantly improve the prediction accuracy of equipment anomaly event sequences, thereby determining whether changes in environmental factors constitute a Granger cause of equipment anomaly events. Finally, the test results are used to characterize the correlation, directly reflecting the degree of causal influence of environmental factors on equipment anomaly events. This process design, from selection to testing to characterization, forms a complete causal analysis chain. Through the organic combination of preliminary correlation analysis and causal testing, it effectively solves the attribution bias problem caused by relying solely on statistical correlation.

[0098] As a preferred embodiment, the specific implementation of the solution in this application is as follows: When analyzing abnormal electricity metering events in a certain area, the system first acquires environmental monitoring data sequences such as temperature and humidity and abnormal electricity metering event sequences, and calculates their correlation index; when the correlation index exceeds the threshold dynamically adjusted based on historical data, the system automatically inputs these sequences into the Granger causality test module; this module uses a vector autoregression model to evaluate predictive ability. If the test results show that the temperature change sequence can significantly predict abnormal metering events, then the temperature change is confirmed as the Granger cause of the metering abnormality, and this result is used as the quantitative basis for the impact of environmental factors for the allocation of compensation responsibility.

[0099] Through the above scheme, this application effectively distinguishes between the causal relationship and statistical correlation between environmental factors and equipment abnormal events, avoiding the problem of inaccurate allocation of refund and subsidy responsibilities due to misjudging correlation as causation, thereby improving the scientificity and fairness of electricity fee refund and subsidy processing.

[0100] Specifically, in some of the embodiments described above in this application, environmental correlation analysis is proposed to determine the impact pattern and correlation period of environmental factors on equipment operation. However, in its implementation, there is a lack of a quantitative mechanism for the impact intensity and lag period, which results in the impact pattern remaining only at the level of qualitative description. It cannot accurately reflect the quantitative correlation between environmental factors and abnormal equipment events, thus making it difficult to distinguish the specific contribution of environmental factors in abnormal events due to the lack of reliable data support for the allocation of compensation responsibility.

[0101] In response, this application further proposes to construct an impact model of environmental factors on equipment operation based on correlation, and to generate the correlation time periods of environmental factors on equipment operation, including: Based on the correlation coefficient and lag time parameter in the correlation relationship, the impact intensity and lag time of environmental factors on equipment abnormal events are quantified to obtain the key quantitative parameters of the impact pattern. Based on the lag time parameter and the duration of the equipment malfunction event, the effective correlation time window between environmental factors and equipment malfunction is determined, and the correlation period is obtained. Based on key quantitative parameters and effective correlation time windows, an impact model of environmental factors on equipment operation is constructed. The impact model is characterized as a functional relationship in which environmental factors act on the equipment operating state with a specific impact intensity within a specific correlation period.

[0102] Among these, quantifying the impact intensity and lag period of environmental factors on equipment anomalies refers to transforming the statistical indicators output from dynamic correlation analysis into measurable quantitative parameters. This can be achieved using an impact intensity mapping function based on historical data and calculations based on equipment operating characteristic parameters. The aim is to avoid the shortcomings of traditional methods that rely on subjective experience and judgment, thus upgrading the impact of environmental factors from a vague description to a calculable quantitative basis. Determining the effective correlation time window between environmental factors and equipment anomalies refers to defining the period of action by combining the transmission characteristics of environmental impacts with the actual duration characteristics of equipment anomalies. This can be achieved using a dynamic time window expansion algorithm, aiming to solve the problem of time period definition deviation caused by ignoring equipment response characteristics in time correlation analysis. Constructing the impact model of environmental factors on equipment operation refers to integrating quantitative parameters and time windows into a structured functional relationship. This can be achieved using decision logic combining responsibility weight coefficients, time delay rules, and effective time boundaries, aiming to provide an embeddable objective input model for compensation and refund responsibility analysis.

[0103] Specifically, the proposed solution first transforms the correlation coefficient and lag time parameters in the correlation relationship into quantitative parameters of impact intensity and lag period, thereby establishing a benchmark for the intensity correlation between environmental factors and equipment anomaly events. Based on this, it dynamically determines the effective correlation time window by combining the duration of the equipment anomaly event, locating the start and end boundaries of the environmental factor's effect. Finally, it integrates the quantitative parameters and time window to construct a functional relationship model, characterizing the impact of environmental factors as a mechanism of action with a specific intensity within a specific time period. This step-by-step quantification and integration mechanism ensures that the impact model reflects both the intensity dimension of the environmental impact and the dynamic characteristics of the time dimension, thus providing a traceable and objective basis for the allocation of compensation responsibility.

[0104] As a specific implementation method, the solution of this application is implemented as follows: When the system detects a significant correlation between changes in ambient temperature and abnormal electricity meter readings, the influence intensity weight of temperature on the abnormal event is calculated based on the correlation coefficient and lag time parameter. The effective period of effect of temperature influence is determined by combining the duration of the abnormality. The influence intensity weight and the period of effect are integrated into a temperature-metering abnormality influence function model. This model is directly input into the refund and compensation responsibility analysis module to quantify the specific contribution ratio of environmental factors in the responsibility allocation, thereby guiding the allocation of refund and compensation amounts.

[0105] The above technical solutions enable the quantification and mathematical expression of the impact of environmental factors on equipment operation, allowing the allocation of refund and subsidy responsibilities to distinguish the specific contribution of environmental factors based on objective data. This effectively solves the problem of ambiguous responsibility definition caused by the impact model only remaining at the level of qualitative description, and improves the accuracy and consistency of electricity fee refund and subsidy processing.

[0106] In some of the embodiments described above in this application, the intensity and lag time of the impact of environmental factors on equipment abnormal events are proposed to construct the impact model of environmental factors on equipment operation. However, in this process, the correlation coefficient and lag time parameter in the correlation relationship are directly used for quantification. The correspondence between the correlation coefficient and the actual impact degree in historical data and the physical differences in equipment operating characteristics are not fully combined. This results in the lack of quantification results of impact intensity and lag time, which in turn makes the construction of the impact model of environmental factors in the allocation of compensation responsibility inaccurate and affects the reliability of subsequent responsibility weight calculation.

[0107] In response, this application further proposes to quantify the intensity and lag period of the impact of environmental factors on equipment anomaly events based on the correlation coefficient and lag time parameter in the correlation relationship, thereby obtaining key quantitative parameters of the impact model, including: Based on the correlation coefficient, the direct impact weight of environmental factors on equipment abnormal events is calculated through a preset impact intensity mapping function. The impact intensity mapping function is calibrated according to the correspondence between the correlation coefficient and the actual impact in historical data. Based on the lag time parameters and equipment operating characteristics, the lag time range of the impact of environmental factors is determined. The equipment operating characteristics include the thermal inertia time constant or electrical response delay of the equipment. The direct impact weights and lag time ranges are normalized and fused to obtain standardized impact intensity and lag time, which serve as key quantitative parameters.

[0108] Among them, the correlation coefficient is a statistical indicator that measures the degree of linear correlation between changes in environmental factors and abnormal equipment events. In practical applications, it can be the Pearson correlation coefficient or the Spearman rank correlation coefficient. Its purpose is to quantify the correlation strength between environmental factors and abnormal equipment events. The lag time parameter refers to the time delay required for changes in environmental factors to be transmitted to abnormal events in equipment. It can be understood as the lag time corresponding to the maximum cross-correlation value extracted based on cross-correlation analysis. Its purpose is to characterize the temporal characteristics of environmental impacts. The influence strength mapping function refers to the function that maps the correlation coefficient to the weight that directly affects the influence. Specifically, it can be a multinomial regression model trained based on historical data or a piecewise linear lookup table function. Its purpose is to transform the abstract statistical correlation into physical influence weights that conform to the actual scenario. Equipment operating characteristics refer to the inherent physical response attributes of equipment, such as thermal inertia time constant or electrical response delay. The purpose is to determine the range of lag time in combination with the differences in equipment type, so as to avoid quantification distortion caused by general parameters. Normalization and fusion processing refers to the standardization and integration of the weights and lag time ranges that directly affect the data. In practical applications, min-max normalization or z-score normalization methods can be used. Fusion processing can use principal component analysis or weighted average algorithms. Its purpose is to eliminate dimensional differences and generate key quantitative parameters with a uniform scale.

[0109] Specifically, the proposed solution first utilizes correlation coefficients as input, combined with a pre-defined influence intensity mapping function. This function is calibrated based on the correspondence between correlation coefficients and actual influence levels in historical data, thereby transforming statistical correlation into direct influence weights that reflect actual influence levels, effectively avoiding quantitative bias caused by relying solely on correlation coefficients. Second, based on lag time parameters and incorporating equipment operating characteristics such as thermal inertia time constants or electrical response delays, the lag time range of environmental factors is dynamically determined, ensuring that the lag parameters align with the dynamic response patterns of equipment in real operating environments. Finally, the direct influence weights and lag time ranges are normalized to eliminate dimensional differences, and standardized influence intensity and lag time are generated through fusion processing. This ensures that key quantitative parameters comprehensively reflect the intensity and temporal characteristics of environmental factors on a unified scale, thus providing a reliable data foundation for constructing environmental factor influence models.

[0110] As a specific implementation method, the solution of this application is implemented as follows: When quantifying the impact of environmental factors, the system can use a neural network model trained based on historical data as the impact intensity mapping function to map the correlation coefficient into the direct impact weight; for the lag time parameter, combined with the characteristics of the equipment type, for example, for distribution transformer equipment, its thermal inertia time constant is used to determine the lag time range; subsequently, the direct impact weight and the lag time range are subjected to min-max normalization processing, and standardized key quantitative parameters are generated through a weighted fusion algorithm, wherein the weight coefficient is dynamically adjusted according to the equipment type.

[0111] Through the above scheme, this application can more accurately quantify the intensity and lag period of the impact of environmental factors on equipment abnormal events, avoiding quantitative deviations caused by ignoring the correspondence of historical data and the physical differences of equipment, thereby improving the accuracy of the construction of environmental factor impact models and ensuring the reliability of the compensation and refund responsibility allocation scheme.

[0112] In practical applications, some embodiments of this application propose determining an effective associated time window to generate associated time periods for environmental factors. However, in its implementation, the time window is defined solely based on the lag time parameter and the duration of equipment malfunction events, failing to fully consider the dynamic decay characteristics of the environmental factor's impact. This results in the associated time period not being able to fully cover the entire process of environmental factors from initial impact to complete disappearance, potentially omitting the impact decay period or including irrelevant time periods. Consequently, this leads to deviations in defining the scope of environmental factors in the analysis of liability for compensation and refund, affecting the accuracy of liability allocation.

[0113] To address this, this application further proposes a method for determining the effective correlation time window between environmental factors and equipment malfunctions based on lag time parameters and the duration of equipment malfunction events, resulting in the correlation period, including: Based on the lag time parameter, the starting point of the transmission of environmental factor changes to equipment abnormal events is determined; Determine the end time of the equipment malfunction event based on its duration. The time interval from the start time to the end time is extended by a preset buffer period to obtain an effective associated time window. The buffer period is dynamically adjusted based on the decay characteristics of environmental factors. Based on the effective correlation time window, a correlation period is generated, which includes the complete time range from the onset to the complete disappearance of the influence of environmental factors.

[0114] Specifically, the lag time parameter refers to the time delay characteristic between changes in environmental factors and abnormal equipment events. It can be determined through Granger causality tests or cross-correlation analysis of historical environmental monitoring data and equipment abnormality records. The purpose is to accurately capture the critical moment when environmental changes trigger equipment abnormalities and avoid starting point deviations caused by fixed delay assumptions. The buffer period can be understood as an additional time period added outside the basic time window. It can be dynamically calculated based on the decay model of environmental factors (such as the exponential decay model or the hyperbolic decay model). The purpose is to cover the entire process of environmental impact from peak to complete decay, solving the problem of rigid window boundaries in traditional methods. In practical applications, the effective correlation time window is formed by extending the time interval from the start time point to the end time point into a buffer period. The purpose is to ensure that the time window can truly match the decay patterns of different environmental factors and avoid omissions in the decay period due to an excessively short window or the introduction of noise interference due to an excessively long window.

[0115] Specifically, the proposed solution first determines the starting point of the transmission of environmental factor changes to equipment malfunction events based on lag time parameters. This utilizes the quantified time delay characteristics in causal analysis to provide a scientific starting point benchmark for the time window. Then, it determines the ending point based on the duration of the equipment malfunction event, closely matching the actual duration of the malfunction event to ensure that the time window covers the complete occurrence process of the malfunction event. Next, it extends the time interval from the starting point to the ending point by a preset buffer period, and the buffer period is dynamically adjusted based on the decay characteristics of the environmental factor impact. By introducing a decay model, it adaptively extends the window boundary to cover the decay process of the environmental impact from its peak to its complete disappearance. Finally, it generates a related time period based on the effective associated time window. This time period clearly includes the complete time range from the beginning to the complete disappearance of the environmental factor impact, providing a time benchmark for the analysis of compensation responsibility.

[0116] As a specific implementation method, the solution of this application is implemented as follows: When the system detects that the change in ambient temperature is related to the abnormal transformer oil temperature event, the starting time point of the temperature change being transmitted to the abnormal oil temperature is first determined according to the lag time parameter; then the ending time point is determined according to the duration of the abnormal oil temperature event; the basic time window is extended to a buffer period dynamically adjusted based on the temperature decay characteristics to obtain an effective correlation time window; finally, a correlation period containing the complete time range from the start to the complete disappearance of the temperature influence is generated.

[0117] Through the above scheme, this application can define the time range of the impact of environmental factors on abnormal equipment operation, avoid the deviation in responsibility allocation caused by incomplete window definition, and thus improve the accuracy and traceability of electricity fee refund and subsidy decisions.

[0118] Specifically, in some of the embodiments described above in this application, a model for constructing the impact of environmental factors on equipment operation is proposed to quantify the intensity and temporal characteristics of environmental impact. However, in its implementation, the impact model only characterizes the mechanism of action of environmental factors as a functional relationship and fails to directly transform the quantitative parameters into operable responsibility allocation decision rules. This results in a lack of weight basis, time boundary definition, and logical decision framework when environmental factors participate in the analysis of compensation and refund responsibilities. Consequently, the attribution of responsibility still relies on manual experience judgment and cannot achieve automation and normalization of compensation and refund responsibility allocation.

[0119] In response, this application further proposes steps for constructing an impact model of environmental factors on equipment operation based on key quantitative parameters and effective correlation time windows, including: Based on the standardized impact intensity in key quantitative parameters, determine the responsibility weight coefficient of environmental factors in the allocation of compensation and refund responsibilities; Based on the standard lag period in key quantitative parameters, a time delay rule for the impact of environmental factors is established. The time delay rule defines the expected time range for equipment abnormalities to occur after changes in environmental factors. Based on the effective correlation time window, the effective time boundary of the impact of environmental factors is determined. The effective time boundary is used to limit the effective period of environmental factors participating in the responsibility analysis. By combining the responsibility weight coefficient, time delay rule, and effective time boundary, an environmental impact decision logic is formed for the analysis of compensation and refund responsibilities. This environmental impact decision logic serves as the impact model of environmental factors on equipment operation.

[0120] Among them, the responsibility weighting coefficient refers to the decision parameter that maps the objective quantitative value of environmental impact to the responsibility allocation ratio. It can be implemented using a linear mapping function or a nonlinear piecewise function based on historical data. Its purpose is to eliminate the defects of environmental factor weights relying on subjective experience and ensure that responsibility allocation is strictly based on the measured impact intensity. The time delay rule can be understood as the constraint condition that defines the time characteristics of environmental disturbances transmitted to equipment anomalies. It can be implemented using a dynamic threshold model based on the thermal inertia time constant of equipment or a fixed interval model based on electrical response delay. Its purpose is to match the actual timing of environmental changes and abnormal events and prevent erroneous attribution during unrelated periods. The effective time boundary specifically refers to the effective time period boundary for environmental factors to participate in responsibility analysis. It can be implemented using a dynamic expansion algorithm based on the duration of abnormalities or a buffer period adjustment mechanism based on the decay characteristics of environmental factors. Its purpose is to eliminate interference from irrelevant time periods and ensure that responsibility analysis focuses on the actual effective period of environmental impact. Environmental impact decision logic refers to a structured decision framework that integrates multi-dimensional parameters. It can be understood as an executable scheme that logically combines responsibility weight coefficients, time delay rules, and effective time boundaries through a rule engine or decision tree model. Its purpose is to transform abstract impact patterns into direct input basis for compensation and refund responsibility analysis, supporting automated decision-making.

[0121] Specifically, the proposed solution directly maps the standardized impact intensity in key quantitative parameters to responsibility weight coefficients, transforming the objective quantitative value of environmental impact into a quantifiable responsibility allocation ratio. Simultaneously, it establishes time delay rules based on standard lag periods and defines the expected time range from environmental factor changes to equipment anomalies by combining equipment operating characteristics, ensuring that responsibility analysis strictly follows the actual timing of environmental disturbances. Furthermore, it utilizes effective correlation time windows to determine effective time boundaries, dynamically defining the effective time periods for environmental factors to participate in the analysis and effectively isolating interference from uncorrelated time periods. Finally, it logically combines the responsibility weight coefficients, time delay rules, and effective time boundaries to form an environmental impact decision-making logic, upgrading the impact model of environmental factors from a functional relationship to an executable decision rule, thereby driving the automated determination of compensation and refund responsibility analysis.

[0122] As a specific implementation method, this application is implemented as follows: In the process of environmental factor impact analysis, the standardized impact intensity is converted into a responsibility weight coefficient through a mapping function calibrated by historical data. For example, when the standardized impact intensity is 0.8, the responsibility weight coefficient can be determined to be 80%. The time delay rule sets the expected time range from environmental change to anomaly occurrence based on the equipment thermal inertia time constant. For example, a lag period of 15-30 minutes is set for transformer equipment. The effective time boundary dynamically expands the effective correlation time window according to the duration of equipment anomaly. For example, a 10% buffer period is added on the basis of an anomaly lasting 2 hours. The environmental impact decision logic combines the above parameters through a rule engine. When the environmental factor change is within the effective time boundary and meets the time delay rule, the responsibility weight coefficient is automatically triggered to participate in the calculation of refund and compensation allocation.

[0123] Through the above technical solution, this application realizes the direct transformation of quantitative parameters of environmental factors into decision rules for responsibility allocation, enabling the analysis of refund and subsidy responsibilities to have weight basis, clear time boundary definition and logical decision framework, effectively eliminating the processing deviation caused by human experience judgment, ensuring that the allocation of refund and subsidy responsibilities is strictly based on the objective quantitative results of environmental impact and the actual timing of action, and improving the automation level and judgment of electricity fee refund and subsidy processing.

[0124] Specifically, in some of the embodiments described above in this application, a joint analysis based on power metering data and equipment health status indicators through anomaly pattern identification and refund / refund responsibility analysis model is proposed to generate abnormal power consumption pattern characteristics and refund / refund responsibility allocation scheme. However, in its implementation process, due to the lack of a specific data fusion mechanism and dynamic adjustment method for responsibility weight, the accuracy of anomaly identification is insufficient, the responsibility attribution is ambiguous, and the quantitative allocation of refund / refund responsibility cannot be achieved.

[0125] In response, this application further proposes the following technical solution, see [link to technical solution]. Figure 4 Taking the server as the executing entity as an example, the following steps are included.

[0126] 401. Perform time series alignment and data fusion on the electricity metering data and equipment health status indicators to obtain a comprehensive data series for anomaly analysis; 402. Based on the comprehensive data sequence, multi-dimensional feature extraction and anomaly detection are performed through an anomaly pattern recognition model to obtain abnormal power consumption pattern features, which include abnormal power consumption time period, anomaly type and anomaly severity. 403. Based on the characteristics of abnormal power consumption patterns, a responsibility attribution analysis model is used to obtain the responsibility weights of equipment factors, environmental factors, and user factors in abnormal events. 404. Generate a refund / reimbursement responsibility allocation scheme based on responsibility weight and severity of anomalies.

[0127] Time series alignment refers to the technical process of adjusting data from different sources to a unified time reference. This can be achieved using timestamp-based linear interpolation or dynamic time warping algorithms, aiming to eliminate time series deviations between equipment operation data and metering data caused by differences in sampling frequencies. Data fusion refers to the method of integrating multi-source heterogeneous data to form a unified representation. This can be achieved using feature-level splicing or decision-level weighted fusion, such as dimensionality reduction through principal component analysis or Kalman filtering for data correction, aiming to improve the completeness and spatiotemporal consistency of the data required for anomaly analysis. Anomaly pattern recognition models refer to computational frameworks used to identify abnormal behavior. These can be built based on statistical learning methods or deep neural networks, aiming to extract multidimensional features from comprehensive data and detect abnormal events. Multidimensional feature extraction refers to technical means of mining data features from time domain, frequency domain, and other perspectives, which can include mean, variance, spectral energy, or wavelet coefficients, etc. The calculation of feature parameters aims to comprehensively characterize the dynamic features of electricity consumption behavior. Abnormal electricity consumption pattern features refer to the set of key attributes describing abnormal events, including the abnormal electricity consumption period, abnormal type, and abnormal severity. Abnormal types can be categorized as equipment failure, environmental interference, or user violation, with the aim of providing structured input for responsibility analysis. The refund / reimbursement responsibility analysis model refers to a decision-making mechanism that quantifies responsibility attribution. It can be implemented based on a rule engine or classification algorithm, with the aim of dynamically evaluating the contribution of each factor to the abnormal event. Responsibility weights refer to the quantified responsibility proportions of equipment factors, environmental factors, and user factors in the abnormal event. These can be obtained through normalization, with the aim of providing an objective basis for refund / reimbursement allocation. Generating a refund / reimbursement responsibility allocation scheme refers to the calculation process that determines the specific details of refund / reimbursement execution. It can use a weighted fusion algorithm to map responsibility weights to refund / reimbursement proportions, with the aim of achieving automation and fairness in refund / reimbursement decisions.

[0128] Specifically, the proposed solution ensures that electricity metering data and equipment health status indicators form a comprehensive data sequence under a unified time benchmark through time series alignment and data fusion, laying a reliable data foundation for anomaly analysis. Based on this comprehensive data sequence, the anomaly pattern recognition model performs multi-dimensional feature extraction and anomaly detection, identifying abnormal electricity consumption periods, types, and severity, thereby distinguishing different anomaly patterns such as equipment failure, environmental interference, or user violations. Subsequently, the refund / reimbursement responsibility analysis model performs responsibility attribution analysis based on the identified abnormal electricity consumption pattern characteristics, dynamically adjusting the responsibility weights of equipment factors, environmental factors, and user factors, so that the weights reflect the real-time impact of the degree of equipment health status deterioration and the environmental correlation analysis results. Finally, the refund / reimbursement responsibility allocation scheme is generated by combining the responsibility weights and the severity of the anomaly. When mapping the responsibility weights to specific refund / reimbursement ratios, the impact of the severity of the anomaly on the base amount is fully considered, realizing the transformation of refund / reimbursement decisions from qualitative description to quantitative execution, forming a complete closed loop of anomaly identification and responsibility allocation.

[0129] As a preferred embodiment, the solution of this application is implemented as follows: In the time series alignment stage, a timestamp alignment method based on network time protocol synchronization can be used to adjust the power series of electricity metering data and the vibration amplitude series of equipment health status indicators to a unified sampling point per minute; in the data fusion stage, the root mean square value feature of the power series and the peak value feature of the vibration amplitude can be concatenated at the feature level to form a two-dimensional comprehensive data series; in the anomaly detection stage, a convolutional neural network can be used to extract time-frequency domain features from the comprehensive data series to identify abnormal power consumption periods and equipment failure anomalies; in the responsibility analysis stage, equipment failure anomalies can be matched based on a preset rule base, and the weight of equipment factors can be dynamically increased when the equipment health status indicator is below the threshold; finally, based on the normalized responsibility weight and the severity of the anomaly, an allocation scheme including the refund and compensation amount of the equipment operation and maintenance party and the execution order is generated.

[0130] Through the above scheme, this application effectively avoids the risk of misjudgment of anomalies caused by time series deviations of multi-source data, and improves the objectivity and comprehensiveness of anomaly identification; at the same time, it realizes the dynamic quantification of responsibility weight and the allocation of refund and compensation amounts, so that the refund and compensation responsibility allocation process is freed from the ambiguity of experience judgment, and ensures the automation and fairness of electricity fee refund and compensation processing.

[0131] In practical applications, some of the embodiments described above in this application propose to generate responsibility weights by analyzing the characteristics of abnormal power consumption patterns. However, in the implementation process, the allocation of initial responsibility weights lacks a dynamic adjustment mechanism and fails to fully combine equipment health status indicators, environmental correlation analysis results, and user historical power consumption behavior data. This makes it difficult to distinguish the differentiated impacts of inherent equipment defects, external environmental disturbances, and user operation behaviors on abnormal events. It is also prone to deviations in responsibility allocation due to static rules and cannot adapt to complex and ever-changing abnormal event scenarios.

[0132] In response, this application further proposes a method for analyzing the responsibility attribution of equipment factors, environmental factors, and user factors in abnormal events based on the characteristics of abnormal power consumption patterns and using a compensation and refund responsibility analysis model. The steps include: By using the refund and compensation responsibility analysis model, based on the abnormality type in the abnormal power consumption pattern characteristics, and matching the preset responsibility allocation rule library, the initial responsibility weights of equipment factors, environmental factors and user factors are obtained; Based on the equipment health status indicators and the results of environmental correlation analysis, the initial responsibility weights of equipment factors and environmental factors are dynamically adjusted. Specifically, when the equipment health status indicators are lower than the preset threshold, the weight of equipment factors is increased, and when the environmental correlation analysis results show a strong correlation, the weight of environmental factors is increased. Based on users' historical electricity consumption data, the initial responsibility weight of user factors is adjusted, and the weight of user factors is increased when users have records of overloading electricity or violating regulations. The weights of dynamically adjusted equipment factors, environmental factors, and user factors are normalized to obtain the responsibility weights.

[0133] The pre-defined responsibility allocation rule base refers to a decision knowledge base that stores the mapping relationship between anomaly types and responsibility weights. It can be implemented using rule-based expert systems or decision tree algorithms, aiming to provide an initial weight framework for responsibility attribution analysis. Dynamic adjustment of equipment factor weights based on equipment health status indicators refers to real-time correction of responsibility weights according to equipment operating status. This can be implemented using a threshold comparison mechanism; when equipment health status indicators fall below a preset threshold, the weight of equipment factors is increased, aiming to reflect the direct impact of equipment performance degradation on abnormal events. Dynamic adjustment of environmental factor weights based on environmental correlation analysis results refers to adjusting weights using quantitative correlation evidence between environmental factors and equipment anomalies. This can be implemented using a threshold comparison mechanism. The system employs a threshold judgment based on correlation indicators. When the environmental correlation analysis results show a strong correlation, the weight of environmental factors is increased to objectively reflect the degree of impact of external environmental disturbances. The correction of user factor weights based on historical user electricity consumption data refers to adjusting responsibility weights in conjunction with historical user behavior risks. This can be achieved using a historical violation record matching mechanism. When a user has records of overloading electricity or violating regulations, the user factor weight is increased to ensure that the allocation of user factor responsibility matches the actual operational risks. Weight normalization refers to standardizing the adjusted weights to a sum of 1, which can be achieved using a linear normalization algorithm to ensure the logical consistency of the weights for equipment, environment, and user factors.

[0134] Specifically, this application's solution first generates initial responsibility weights based on an anomaly type matching rule base, establishing a basic framework for responsibility allocation. Then, it adjusts the weights of equipment factors in real time based on equipment health status indicators, increasing the responsibility proportion when equipment performance deteriorates, thereby accurately reflecting the impact of inherent equipment defects. Simultaneously, it dynamically increases the weights of environmental factors based on environmental correlation analysis results, incorporating quantitative evidence of the correlation between environmental factors and equipment anomalies into responsibility determination, objectively reflecting the effect of external disturbances. Furthermore, it corrects the weights of user factors using historical user electricity consumption data, using historical violations as a dynamic basis to ensure that user responsibility matches risk. Finally, it normalizes the adjusted weights to form a logically consistent weight system. This multi-dimensional dynamic adjustment mechanism enables responsibility attribution analysis to adapt to the specific scenarios of abnormal events, effectively distinguishing the differentiated impacts of equipment, environment, and user factors, and avoiding biases caused by static rules.

[0135] As a specific implementation method, the solution of this application is implemented as follows: In the case of voltage anomaly events, the compensation and refund responsibility analysis model identifies the anomaly type as voltage fluctuation and matches the initial weights from the responsibility allocation rule library; the equipment health status index indicates equipment performance degradation, and the weights of equipment factors are increased accordingly; the environmental correlation analysis results confirm that temperature changes are strongly correlated with anomalies, and the weights of environmental factors are increased; the user historical data contains overload records, and the weights of user factors are adjusted upwards; finally, the responsibility weight allocation is generated through normalization processing.

[0136] Through the above technical solution, this application can realize the dynamic quantification of responsibility attribution analysis, distinguish the impact of inherent equipment defects, external environmental disturbances and user operation behavior on abnormal events, effectively avoid responsibility attribution deviation caused by static rules, and significantly improve the accuracy and adaptability of electricity fee refund and subsidy decisions.

[0137] In practical applications, some of the embodiments described above in this application propose a scheme for generating refund and compensation responsibility allocation to achieve quantitative allocation of refund and compensation responsibility. However, in its implementation process, the conversion of responsibility weight to responsibility ratio lacks clear algorithmic support, the determination of the benchmark refund and compensation amount does not fully integrate the correlation between the severity of the anomaly and the actual measurement data, and the generation of the refund and compensation execution order does not introduce priority rules for dynamic optimization. This results in the scheme generation relying on manual experience judgment, insufficient automation, and subjective bias in the responsibility allocation results, making it difficult to guarantee the consistency and execution efficiency of refund and compensation processing.

[0138] In response, this application further proposes a scheme for allocating compensation and refund responsibilities based on responsibility weight and severity of anomalies, including: The refund / refund responsibility ratio is determined based on the responsibility weights of equipment factors, environmental factors, and user factors. The refund / refund responsibility ratio is mapped to specific responsibility allocation coefficients through a weighted fusion algorithm. The base refund / refund amount is determined based on the severity of the anomaly and the electricity metering data during the abnormal electricity consumption period. Based on the proportion of compensation and refund liability and the benchmark compensation and refund amount, the specific compensation and refund amount for each responsible party is determined, and the order of execution of compensation and refund is determined based on the compensation and refund priority rules; Generate a refund and compensation responsibility allocation plan that includes the refund and compensation amount for each responsible party, the order of refund and compensation execution, and the time nodes.

[0139] The refund / reimbursement responsibility ratio refers to the technical means of converting the responsibility weights of equipment factors, environmental factors, and user factors into quantifiable responsibility allocation coefficients. This can be achieved using linear weighting algorithms, exponential decay algorithms, or nonlinear mapping algorithms based on decision trees. The aim is to ensure that the generation of the responsibility ratio strictly relies on objective weight data and avoids interference from human experience. The benchmark refund / reimbursement amount refers to a quantitative benchmark value reflecting the degree of impact of abnormal events. It can be determined based on the product relationship between the severity level of the abnormality and the electricity metering data during the abnormal period, or by fitting historical refund / reimbursement case data through a regression model. The aim is to ensure that the amount calculation reflects both the degree of technical impact and is anchored to real metering data. The refund / reimbursement execution order refers to the timing arrangement of refund / reimbursement operations by each responsible party. Priority rules can be set according to the nature of the responsibility, such as prioritizing equipment factors over environmental factors, environmental factors over user factors, or dynamically adjusting the execution order according to the size of the responsibility weight. The aim is to optimize the timing logic of refund / reimbursement operations and improve execution efficiency. The refund / reimbursement responsibility allocation scheme refers to a structured and integrated complete execution document of refund / reimbursement elements. It can include elements such as amount details, execution time series, and verification requirements. The aim is to provide a complete instruction set that can directly drive automatic refund / reimbursement operations.

[0140] Specifically, the proposed solution first maps the responsibility weights of equipment, environmental, and user factors into specific responsibility allocation coefficients using a weighted fusion algorithm, forming a refund / refund responsibility ratio. Simultaneously, it dynamically calculates a benchmark refund / refund amount reflecting the degree of impact of the anomaly by combining an anomaly severity index with actual electricity metering data during abnormal power consumption periods. Then, it mathematically couples the refund / refund responsibility ratio with the benchmark refund / refund amount to calculate the specific refund / refund amount for each responsible party, and determines the optimal execution order based on preset refund / refund priority rules. Finally, it integrates the amount data, execution order, and time requirements into a structured refund / refund responsibility allocation scheme, achieving seamless integration from technical attribution to operational implementation. This process systematically solves the problems of insufficient responsibility allocation quantification and execution disconnect through algorithm-driven responsibility ratio transformation, data-fusion-based amount determination, and rule-guided order optimization.

[0141] As a specific implementation method, when the system identifies an abnormal electricity metering event for a user in a residential area, based on the previous analysis, the responsibility weights for equipment factors are 0.65, environmental factors are 0.25, and user factors are 0.10. A weighted fusion algorithm is used to calculate the responsibility ratios for the equipment party (65%), the environment party (25%), and the user party (10%). Combining the anomaly severity level 4 (out of 5) and the metering data of 800 kWh during the abnormal period, a linear model is used to determine the baseline refund amount as 400 yuan. Then, the refund amount is calculated as 260 yuan for the equipment party, 100 yuan for the environment party, and 40 yuan for the user party. According to the refund priority rule of prioritizing equipment factors, the execution order is determined as follows: first process the refund for the equipment party, then the environment party, and finally the user party. Finally, a refund responsibility allocation scheme is generated, which includes the specific refund amount, the execution time sequence (equipment party within 24 hours, environment party within 48 hours, user party within 72 hours), and verification requirements.

[0142] Through the above scheme, this application realizes the automated mapping of responsibility weight to responsibility ratio, avoiding the interference of human experience in weight allocation; dynamically correlates the severity of anomalies with electricity metering data to determine the benchmark amount, improving the accuracy of amount estimation; introduces priority rules for refunds and subsidies to optimize the execution order, solving the problem of inefficiency caused by the arbitrary execution order in traditional methods; and finally generates a structured and complete scheme, effectively bridging the gap between responsibility analysis and execution, and ensuring the consistency and efficiency of refund and subsidy processing.

[0143] Specifically, in some of the embodiments described above in this application, a comprehensive decision-making process based on environmental correlation analysis results and refund / refund responsibility allocation schemes is proposed through a refund / refund strategy generator. However, in its implementation, the environmental correlation analysis results and refund / refund responsibility allocation schemes fail to be effectively integrated to form a unified decision-making basis. The strategy matching relies on simple rules and lacks intelligent optimization capabilities. The refund / refund execution verification mechanism is imperfect, resulting in low efficiency and insufficient accuracy in generating refund / refund execution schemes, low automation, and difficulty in adapting to the complex and ever-changing electricity fee refund / refund scenarios.

[0144] This application further proposes, see [link to relevant documentation] Figure 5 Taking the server as the executing entity as an example, the following steps are included.

[0145] 501. Integrate the environmental impact decision-making logic from the environmental correlation analysis results with the compensation and refund responsibility allocation scheme through multi-source data fusion to generate a comprehensive decision input feature vector; 502. Based on the comprehensive decision input feature vector, the optimal withdrawal strategy is matched from the preset strategy library through the strategy matching engine in the withdrawal strategy generator. The strategy matching engine adopts a deep neural network model based on the attention mechanism. 503. Based on the matched optimal refund and compensation strategy, generate a specific refund and compensation execution plan that includes refund and compensation amount allocation, execution time sequence, and verification mechanism; 504. Based on the refund and compensation execution plan, the refund and compensation operation is automatically executed through the API interface of the electricity billing system, and the accuracy of the refund and compensation result is verified through a verification algorithm after the execution is completed.

[0146] Among them, environmental impact decision logic refers to a set of rules that quantify the impact of environmental factors on equipment operation. This can be implemented using a decision tree model trained on historical data or a fuzzy logic system, aiming to transform the intensity and time window of environmental factors into calculable decision parameters. Multi-source data fusion refers to the process of integrating heterogeneous data into a unified feature representation. This can be achieved using feature concatenation, weighted fusion, or deep learning fusion networks, aiming to eliminate the disconnect between environmental factor analysis and responsibility allocation information, forming a complete basis for decision input. The policy matching engine refers to a deep neural network model based on an attention mechanism, which can employ Transformer... The ORMER architecture or a custom attention network is used to dynamically focus on key dimensions in the feature vector that are highly relevant to the refund / refund scenario, thereby improving the degree of policy matching. The refund / refund execution scheme refers to a structured instruction set that includes the entire refund / refund operation process. It can be represented as a JSON-formatted refund / refund instruction file or an executable script. Its purpose is to clarify the rules for refund / refund amount allocation, the execution time sequence arrangement, and the design of the verification mechanism to ensure the standardization and traceability of the operation. The verification algorithm refers to the calculation method for verifying the accuracy of the refund / refund results. It can be implemented using data consistency checks or cross-validation algorithms. Its purpose is to ensure the reliability of the refund / refund operation through multi-dimensional comparison.

[0147] Specifically, the proposed solution integrates multi-source data from environmental impact decision-making logic derived from environmental correlation analysis with the refund / refund responsibility allocation scheme to generate a comprehensive decision input feature vector. This vector integrates multi-dimensional information such as environmental factor responsibility weights and anomaly severity, providing a unified input basis for decision-making. Based on this feature vector, the strategy matching engine utilizes an attention-based deep neural network model to match the optimal refund / refund strategy from a pre-defined strategy library. This model dynamically calculates the attention weights of each dimension in the feature vector to identify the strategy template that best matches the current refund / refund scenario. Subsequently, a specific execution plan is generated based on the matched optimal strategy, including refund / refund amount allocation rules, time-segmented execution plans, and verification mechanisms, ensuring refined management of the refund / refund operation. Finally, the refund / refund instruction is automatically executed through the API interface of the electricity billing system, and a verification algorithm is used to compare and verify the electricity bill data and user account balances before and after the refund / refund in real time, forming a closed-loop process of decision-making, execution, and verification. This effectively solves the core problems of insufficient data integration, lack of intelligent decision-making, and incomplete execution verification.

[0148] As a specific implementation method, the policy matching engine can be implemented as a deep neural network based on the Transformer architecture. The attention mechanism dynamically adjusts the focus on different dimensions by calculating the weight distribution of key parameters such as the responsibility weight of environmental factors and the severity of anomalies in the feature vector. The refund / refund execution plan can be specifically represented as a structured JSON instruction set, where the refund / refund amount is calculated based on the responsibility weight ratio, the execution time series is set as a weekday time-segmented execution plan, and the verification mechanism includes automatic comparison of electricity bill data before and after the refund / refund and real-time verification of the user's account balance. The verification algorithm can specifically adopt an anomaly detection model based on historical refund / refund records, determining the accuracy of the operation by comparing the deviation between the actual refund / refund result and the expected value, ensuring that the refund / refund process conforms to preset rules.

[0149] Through the above scheme, this application achieves deep integration of environmental correlation analysis results and refund / refund responsibility allocation scheme, improving the generation efficiency and decision accuracy of refund / refund execution scheme; the intelligent optimization capability of the strategy matching engine effectively adapts to the needs of complex and ever-changing electricity fee refund / refund scenarios; the sound verification mechanism ensures the reliability of automatic refund / refund operations, significantly reduces the need for manual intervention, and improves the automation level of electricity fee refund / refund processing.

[0150] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0151] Figure 6 This is a schematic diagram of an automatic electricity fee refund / refund system based on multi-source data provided in an embodiment of this application. See also... Figure 6 The system includes: The acquisition module 601 is used to acquire the target user's electricity metering data, equipment operation data, and environmental monitoring data in response to the electricity fee refund event trigger command. The status assessment module 602 is used to assess the operational status based on equipment operation data and environmental monitoring data through the equipment status analysis model, and obtain equipment health status indicators and environmental correlation analysis results. The equipment status analysis model is dynamically updated based on historical equipment operation data. The joint analysis module 603 is used to perform joint analysis based on electricity metering data and equipment health status indicators through anomaly pattern recognition and refund / refund responsibility analysis models to generate abnormal electricity consumption pattern characteristics and refund / refund responsibility allocation schemes. The anomaly pattern recognition model optimizes parameters based on abnormal electricity consumption cases, and the refund / refund responsibility analysis model adjusts the model based on historical responsibility definition results. The comprehensive decision-making module 604 is used to make comprehensive decisions based on the results of environmental correlation analysis and the allocation plan for refund and compensation responsibilities, through the refund and compensation strategy generator, to generate a refund and compensation execution plan and complete the automatic refund and compensation operation.

[0152] It should be noted that the automatic electricity bill refund system based on multi-source data provided in the above embodiments is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer equipment can be divided into different functional modules to complete all or part of the functions described above. In addition, the automatic electricity bill refund system based on multi-source data provided in the above embodiments and the automatic electricity bill refund method embodiments based on multi-source data belong to the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0153] Figure 7 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 700 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 701 and one or more memories 702. The one or more memories 702 store at least one computer program, which is loaded and executed by the one or more processors 701 to implement the methods provided in the various method embodiments described above. Of course, the server 700 may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The server 700 may also include other components for implementing device functions, which will not be elaborated upon here.

[0154] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including a computer program that can be executed by a processor to perform the automatic electricity bill refund method based on multi-source data in the above embodiments. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage device, etc.

[0155] In an exemplary embodiment, a computer program product or computer program is also provided, which includes program code stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium and executes the program code, causing the computer device to perform the above-described automatic electricity fee refund method based on multi-source data.

[0156] In some embodiments, the computer program involved in the present application embodiments may be deployed and executed on a computer device, or executed on multiple computer devices located in one location, or executed on multiple computer devices distributed in multiple locations and interconnected through a communication network. Multiple computer devices distributed in multiple locations and interconnected through a communication network may constitute a blockchain system.

[0157] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0158] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for automatic electricity fee refund / reimbursement based on multi-source data, characterized in that, The method includes: In response to an electricity bill refund / refund event trigger command, acquire the target user's electricity metering data, equipment operation data, and environmental monitoring data; Based on the equipment operation data and environmental monitoring data, the operation status is evaluated through the equipment status analysis model to obtain equipment health status indicators and environmental correlation analysis results. The equipment status analysis model is dynamically updated based on the equipment's historical operation data. Based on the electricity metering data and equipment health status indicators, a joint analysis is performed using an anomaly pattern recognition and refund / refund responsibility analysis model to generate abnormal electricity consumption pattern characteristics and refund / refund responsibility allocation schemes. The anomaly pattern recognition model optimizes parameters based on abnormal electricity consumption cases, and the refund / refund responsibility analysis model is adjusted based on historical responsibility definition results. Based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme, a comprehensive decision is made through the refund / refund strategy generator to generate a refund / refund execution scheme and complete the automatic refund / refund operation.

2. The method according to claim 1, characterized in that, Based on the equipment operation data and environmental monitoring data, the operational status is assessed using an equipment status analysis model to obtain equipment health status indicators and environmental correlation analysis results, including: The equipment status analysis model is used to perform multi-scale time series analysis on the equipment operation data to obtain the steady-state characteristics and transient anomaly characteristics of the equipment operation. By using the equipment status analysis model, environmental correlation analysis is performed on the environmental monitoring data and transient anomaly characteristics to obtain the impact pattern and correlation time period of environmental factors on equipment operation; Based on the steady-state characteristics, transient anomaly characteristics, and environmental impact patterns of the equipment, the equipment health status index and the environmental correlation analysis results are generated using the equipment status analysis model.

3. The method according to claim 2, characterized in that, The multi-scale time-series analysis of the equipment operation data yields steady-state characteristics and transient anomaly characteristics of the equipment operation, including: Multi-scale time-frequency analysis was performed on the equipment operation data to obtain the trend components, periodic components, and residual components of the equipment operation. Based on the trend component and the periodic component, the long-term stability index and periodic fluctuation pattern of the equipment operation are determined through steady-state characteristic analysis. Based on the residual components, transient anomaly detection identifies instantaneous abnormal events and short-term disturbance characteristics in the operation of the equipment. Based on the long-term stability index, the periodic fluctuation pattern, the instantaneous abnormal event, and the short-term disturbance characteristics, steady-state characteristics and transient abnormal characteristics of equipment operation are generated.

4. The method according to claim 3, characterized in that, The process of generating steady-state and transient abnormal characteristics of equipment operation based on the long-term stability index, the periodic fluctuation pattern, the instantaneous abnormal events, and the short-term disturbance characteristics includes: The long-term stability index and the periodic fluctuation pattern are fused with multi-dimensional features to generate the steady-state characteristics of the equipment operation. The multi-dimensional feature fusion includes a quantitative assessment of trend stability and periodic consistency. Spatiotemporal correlation analysis is performed on the instantaneous abnormal events and the short-term disturbance features to generate the transient abnormal features. The spatiotemporal correlation analysis includes abnormal event clustering analysis and disturbance intensity pattern recognition.

5. The method according to claim 2, characterized in that, The environmental correlation analysis of the environmental monitoring data and transient anomaly characteristics yields the impact patterns and correlation time periods of environmental factors on equipment operation, including: Key environmental parameters are extracted from the environmental monitoring data to obtain a sequence of key environmental factors affecting equipment operation; Based on the transient anomaly characteristics and the key environmental factor sequence, the correlation between environmental factor changes and equipment anomaly events is determined through dynamic correlation analysis. The dynamic correlation analysis is modeled based on the spatiotemporal correlation between historical environmental monitoring data and equipment anomaly records. Based on the aforementioned correlation, an impact pattern of environmental factors on equipment operation is constructed, and the associated time periods of environmental factors on equipment operation are generated.

6. The method according to claim 1, characterized in that, Based on the electricity metering data and equipment health status indicators, the system performs joint analysis using anomaly pattern recognition and compensation responsibility analysis models to generate abnormal electricity consumption pattern characteristics and compensation responsibility allocation schemes, including: The power metering data and equipment health status indicators are time-series aligned and fused to obtain a comprehensive data sequence for anomaly analysis. Based on the comprehensive data sequence, multi-dimensional feature extraction and anomaly detection are performed through the anomaly pattern recognition model to obtain the abnormal electricity consumption pattern features, which include abnormal electricity consumption time period, anomaly type and anomaly severity. Based on the characteristics of the abnormal power consumption pattern, the responsibility attribution analysis is performed through the compensation and refund responsibility analysis model to obtain the responsibility weights of equipment factors, environmental factors and user factors in the abnormal event. Based on the responsibility weight and the severity of the anomaly, the refund and compensation responsibility allocation scheme is generated.

7. The method according to claim 6, characterized in that, Based on the characteristics of the abnormal power consumption pattern, the responsibility attribution analysis is performed using the compensation and refund responsibility analysis model to obtain the responsibility weights of equipment factors, environmental factors, and user factors in the abnormal event, including: By using the aforementioned compensation and refund responsibility analysis model, based on the abnormality type in the abnormal power consumption pattern characteristics, and matching the preset responsibility allocation rule library, the initial responsibility weights of equipment factors, environmental factors, and user factors are obtained; Based on the equipment health status indicators and environmental correlation analysis results, the initial responsibility weights of equipment factors and environmental factors are dynamically adjusted respectively. Specifically, when the equipment health status indicators are lower than a preset threshold, the weight of equipment factors is increased, and when the environmental correlation analysis results show a strong correlation, the weight of environmental factors is increased. Based on users' historical electricity consumption data, the initial responsibility weight of user factors is adjusted, and the weight of user factors is increased when users have records of overloading electricity or violating regulations. The weights of dynamically adjusted equipment factors, environmental factors, and user factors are normalized to obtain the responsibility weights.

8. The method according to claim 6, characterized in that, The step of generating the refund / compensation responsibility allocation scheme based on the responsibility weight and the severity of the anomaly includes: The refund / refund responsibility ratio for equipment factors, environmental factors, and user factors is determined based on the responsibility weights. The refund / refund responsibility ratio is then mapped to specific responsibility allocation coefficients using a weighted fusion algorithm. Based on the severity of the anomaly and the electricity metering data during the abnormal electricity consumption period, the benchmark refund / refund amount is determined. Based on the aforementioned refund and compensation liability ratio and the aforementioned benchmark refund and compensation amount, the specific refund and compensation amount for each responsible party is determined, and the refund and compensation execution order is determined based on the refund and compensation priority rules; Generate a refund and compensation responsibility allocation plan that includes the refund and compensation amount for each responsible party, the order of refund and compensation execution, and the time nodes.

9. The method according to claim 1, characterized in that, Based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme, a refund / refund strategy generator makes a comprehensive decision, generates a refund / refund execution plan, and completes the automatic refund / refund operation, including: The environmental impact decision-making logic from the environmental correlation analysis results is fused with the compensation and refund responsibility allocation scheme using multi-source data to generate a comprehensive decision input feature vector. Based on the comprehensive decision input feature vector, the optimal withdrawal strategy is matched from the preset strategy library through the strategy matching engine in the withdrawal strategy generator. The strategy matching engine adopts a deep neural network model based on the attention mechanism. Based on the optimal refund and compensation strategy, a specific refund and compensation execution plan is generated, which includes refund and compensation amount allocation, execution time sequence and verification mechanism; Based on the aforementioned refund and compensation execution scheme, the refund and compensation operation is automatically executed through the API interface of the electricity billing system, and the accuracy of the refund and compensation result is verified through a verification algorithm after the execution is completed.

10. An automatic electricity fee refund / reimbursement system based on multi-source data, characterized in that, The system includes: The acquisition module is used to acquire the target user's electricity metering data, equipment operation data, and environmental monitoring data in response to the electricity fee refund event trigger command. The status assessment module is used to assess the operational status based on the equipment operation data and environmental monitoring data through the equipment status analysis model, and obtain equipment health status indicators and environmental correlation analysis results. The equipment status analysis model is dynamically updated based on the equipment's historical operation data. The joint analysis module is used to perform joint analysis based on the electricity metering data and equipment health status indicators through anomaly pattern recognition and refund / refund responsibility analysis models to generate abnormal electricity consumption pattern characteristics and refund / refund responsibility allocation schemes. The anomaly pattern recognition model optimizes parameters based on abnormal electricity consumption cases, and the refund / refund responsibility analysis model is adjusted based on historical responsibility definition results. The comprehensive decision-making module is used to make comprehensive decisions based on the environmental correlation analysis results and the refund / refund responsibility allocation scheme through the refund / refund strategy generator, generate a refund / refund execution scheme, and complete the automatic refund / refund operation.

Images