An edge-computing-based power consumption safety self-calibration method and system

Abstract

The application relates to an electricity safety self-calibration method and system based on edge computing, and belongs to the technical field of electricity safety monitoring. The method comprises the following steps: collecting operation parameters of an electricity equipment and performing health degree evaluation to generate a health degree score; according to the health degree score, a target calibration strategy is dynamically selected from multiple calibration strategies to execute self-calibration, and a calibration residual parameter is generated; the calibration residual parameter is fed back to a data acquisition module in real time to dynamically adjust a sampling frequency and a filtering coefficient; the calibration residual parameter is synchronously used to update a dynamic reference parameter to correct an error drift curve and the health degree score; and a calibration effect evaluation index is uploaded to the cloud for global strategy optimization and local strategy library updating. Through the double-loop mechanism of edge end self-closed loop calibration and cloud collaborative optimization, the application realizes adaptive calibration of the operation parameters of the electricity equipment and strategy continuous evolution, and significantly improves the calibration real-time performance, precision and resource self-adaptive capability of the system.

Description

Technical Field

[0001] This invention relates to the field of electricity safety monitoring technology, specifically to an electricity safety self-calibration method and system based on edge computing. Background Technology

[0002] With the rapid development of the power Internet of Things and the continuous improvement of the intelligence level of electrical equipment, the long-term operational accuracy and reliability of various electrical equipment (such as motors, distribution transformers, and electricity meters) are crucial to the safe and stable operation of the power system. During long-term operation, electrical equipment is affected by factors such as changes in ambient temperature, humidity fluctuations, electromagnetic interference, and component aging, which can gradually cause deviations in its measurement parameters (such as voltage, current, and power). If not calibrated in time, this can lead to safety hazards such as inaccurate electricity metering, protection setting deviations, and even equipment malfunctions. In existing technologies, the calibration of electrical equipment mainly relies on two methods: First, periodic offline calibration, where maintenance personnel bring standard instruments to the site to verify each device individually. This method is not only labor-intensive and inefficient, but also has a fixed calibration cycle, failing to adapt to the varying degradation rates of equipment under different operating conditions, resulting in both "over-calibration" and "under-calibration." Second, centralized online calibration based on cloud computing, which uploads equipment operating parameters to the cloud for unified analysis and calibration decisions. However, the real-time uploading of massive amounts of equipment data puts enormous pressure on network bandwidth and cloud computing resources, and data transmission delays lead to poor calibration response timeliness, making it difficult to meet the needs of scenarios with high real-time calibration requirements. Furthermore, both methods lack the ability to adaptively perceive the health status of the equipment and cannot dynamically adjust the calibration strategy according to the actual performance degradation of the equipment, making it difficult to achieve an optimal balance between calibration accuracy and resource consumption.

[0003] On the other hand, edge computing, as an emerging computing paradigm, enables localized data processing and real-time response by deploying computing, storage, and network resources at the network edge close to the data source. It has already demonstrated significant advantages in fields such as industrial equipment monitoring and power system operation and maintenance. In recent years, some research has attempted to introduce edge computing technology into the field of online calibration of electrical equipment, for example, using edge nodes to perform real-time analysis and error correction of the sampling signals from electricity meters, or constructing a distributed remote calibration system with a cloud-edge-device architecture. However, existing edge computing-based calibration schemes still have the following shortcomings: First, edge computing nodes typically only serve as relay nodes for data preprocessing and forwarding, lacking the ability to autonomously assess device health and make autonomous decisions on calibration strategies. Calibration still relies on cloud-based commands for triggering, failing to fully leverage the real-time and autonomous advantages of edge computing. Second, data acquisition strategies in existing schemes are usually fixed, unable to dynamically adjust sampling frequency and filtering parameters based on the actual operating status of the device and calibration residuals. This results in wasted acquisition resources when the device is in good condition and insufficient acquisition accuracy when the device is in poor condition. Third, after calibration, there is a lack of quantitative evaluation and closed-loop feedback mechanisms for the calibration effect. Calibration parameters cannot be continuously optimized based on actual results, making it difficult to maintain long-term stability of calibration accuracy. Fourth, the reference zero point of the error drift curve in existing schemes remains fixed for a long time, unable to dynamically correct itself according to the aging process of the device and the reference offset. This causes the health assessment results to gradually deviate from the true state of the device, thus affecting the accuracy of subsequent calibration decisions. Therefore, there is an urgent need for a power safety self-calibration method and system that can fully utilize the real-time processing capabilities of edge computing, possess autonomous health assessment and adaptive decision-making, and enable dynamic adjustment of data acquisition strategies and continuous correction of benchmark parameters. Summary of the Invention

[0004] To address the problems of the prior art, this invention provides a power safety self-calibration method and system based on edge computing.

[0005] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: Firstly, a power safety self-calibration method based on edge computing, comprising the following steps: S1: Data acquisition step, the operating parameter data of the electrical equipment is collected in real time by a data acquisition module deployed on the side of the electrical equipment, the operating parameter data including electrical parameters and environmental parameters; S2: Health assessment step, through the health assessment unit deployed on the edge computing node, the health of the electrical equipment is assessed based on the operating parameter data and the equipment's historical operating records, and a health score is generated; S3: Calibration strategy decision step, through the calibration strategy decision unit deployed on the edge computing node, dynamically selects the target calibration strategy from a variety of preset calibration strategies based on the health score. The variety of calibration strategies includes at least a first calibration strategy, a second calibration strategy, and a third calibration strategy. Different calibration strategies have different calibration accuracies and computing resource consumption. S4: Calibration execution step: The calibration execution unit deployed on the edge computing node performs self-calibration on the operating parameters of the electrical equipment according to the target calibration strategy, generates calibrated parameter data, and generates calibration residual parameters for this calibration. The calibration residual parameters are used to characterize the deviation amount and deviation distribution characteristics between the operating parameters before and after calibration. S5: Calibration evaluation step, using a calibration evaluation unit deployed on the edge computing node to evaluate the self-calibration results and generate calibration effect evaluation indicators; S6: Adaptive adjustment step of acquisition strategy, the calibration residual parameter is fed back to the data acquisition module in real time, and the data acquisition module dynamically adjusts the data acquisition strategy according to the amplitude and distribution characteristics of the calibration residual parameter: when the calibration residual parameter exceeds the first residual threshold, the sampling frequency is increased and the filtering coefficient is decreased; when the calibration residual parameter is lower than the second residual threshold, the sampling frequency is decreased and the filtering coefficient is increased; S7: Dynamic benchmark update step, the calibration residual parameters are synchronously used to update the dynamic benchmark parameters of the electrical equipment. The dynamic benchmark parameters are the reference zero point for calculating the error drift curve in the health assessment step S2. The health assessment unit recalculates the error drift curve based on the updated dynamic benchmark parameters and dynamically corrects the health score. S8: Cloud-based collaborative optimization step, uploading the calibration effect evaluation index and / or calibrated parameter data to the cloud server, whereby the cloud server performs long-term trend analysis and global strategy optimization, generates optimized calibration strategy parameters, and distributes them to the edge computing node; S9: Strategy update step, the edge computing node updates the local calibration strategy library according to the optimized calibration strategy parameters issued by the cloud server.

[0006] In one specific implementation of the first aspect, step S2, the health assessment includes: Based on the historical calibration records of the electrical equipment, analyze the trend of the measurement error of the electrical equipment over time and generate an error drift curve; The aging factor of the electrical equipment is calculated based on the operating time, cumulative workload, and environmental stress data of the electrical equipment. The health score is calculated based on the error drift curve and the aging factor, according to a preset health assessment model.

[0007] In one specific implementation of the first aspect, in step S3, the dynamic target selection calibration strategy includes: The health score is compared with multiple preset health threshold ranges to determine the target threshold range to which the health score belongs; Based on the preset mapping relationship between the target threshold range and the calibration strategy, the corresponding calibration strategy is selected as the target calibration strategy; Specifically, when the health score falls within a first threshold range, the first calibration strategy is selected. This first calibration strategy is a rapid calibration strategy that only verifies and corrects key parameters. When the health score falls within a second threshold range, the second calibration strategy is selected. This second calibration strategy is a standard calibration strategy that verifies and corrects all operating parameters. When the health score falls within a third threshold range, the third calibration strategy is selected. This third calibration strategy is a deep calibration strategy that verifies and corrects all operating parameters and performs additional accuracy compensation calculations.

[0008] In one specific embodiment of the first aspect, prior to step S3, the method further includes: The calibration trigger judgment step determines whether the preset self-calibration trigger condition is met. When the self-calibration trigger condition is met, step S3 is triggered. The self-calibration triggering condition includes at least one of the following: The health score is lower than the preset health threshold; The deviation of the operating parameters of the electrical equipment exceeds a preset deviation threshold; The change in the environmental parameters of the electrical equipment exceeds a preset environmental change threshold; The time interval since the last self-calibration exceeds the preset time interval threshold; The load of the electrical equipment experienced a sudden change exceeding a preset range.

[0009] In one specific implementation of the first aspect, the self-calibration triggering condition further includes a prediction triggering condition; Based on the error drift curve, trend extrapolation is performed to predict the time point when the health score of the electrical equipment will drop below the health threshold within a preset time period in the future. When the difference between the predicted time point and the current time is less than the preset warning time threshold, self-calibration is triggered.

[0010] In one specific implementation of the first aspect, step S5, the calibration effect evaluation includes: The calibrated parameter data is compared with the operating parameter data before calibration, and the calibration correction amount is calculated. The accuracy improvement of the self-calibration is calculated based on the calibration correction amount and the preset reference standard value. During the preset monitoring period after the self-calibration is completed, the fluctuation of the operating parameters of the electrical equipment is continuously monitored, and stability evaluation indicators are generated. Based on the accuracy improvement and the stability evaluation index, the calibration effect evaluation index is generated according to the preset rating rules.

[0011] In one specific embodiment of the first aspect, after step S5, the method further includes: The closed-loop feedback step adjusts the evaluation parameters used in the health assessment step S2 according to the calibration effect evaluation index, thereby achieving continuous optimization of the health assessment model.

[0012] In one specific implementation of the first aspect, in step S6, after the data acquisition module dynamically adjusts the data acquisition strategy according to the amplitude and distribution characteristics of the calibration residual parameter, the method further includes: Real-time monitoring of changes in computing resource utilization and thermal power consumption caused by adjustments to the data acquisition strategy is used as dynamic constraints for the calibration strategy decision in step S3: when the computing resource utilization is higher than a preset resource threshold, a calibration strategy with lower computing resource consumption is selected first.

[0013] Secondly, an edge computing-based power safety self-calibration system includes: A data acquisition module, deployed on the side of the electrical equipment, is used to collect the operating parameter data of the electrical equipment in real time, including electrical parameters and environmental parameters; An edge computing node, communicatively connected to the data acquisition module, is deployed at the network edge near the power-consuming equipment. The edge computing node includes: The health assessment unit is used to assess the health of the electrical equipment based on the operating parameter data and the equipment's historical operating records, and generate a health score. The calibration strategy decision unit is used to dynamically select a target calibration strategy from a variety of preset calibration strategies based on the health score. The variety of calibration strategies includes at least a first calibration strategy, a second calibration strategy, and a third calibration strategy. Different calibration strategies have different calibration accuracies and computational resource consumption. The calibration execution unit is used to perform self-calibration on the operating parameters of the electrical equipment according to the target calibration strategy, generate calibrated parameter data, and generate calibration residual parameters for this calibration. The calibration residual parameters are used to characterize the deviation amount and deviation distribution characteristics between the operating parameters before and after calibration. The calibration evaluation unit is used to evaluate the self-calibration results of the calibration execution unit and generate calibration effect evaluation indicators. A cloud server, which is communicatively connected to the edge computing node, is used to receive calibration effect evaluation indicators and / or calibrated parameter data uploaded by the edge computing node, perform long-term trend analysis and global strategy optimization, and send the optimized calibration strategy parameters to the edge computing node. The edge computing node updates its local calibration strategy library based on the optimized calibration strategy parameters issued by the cloud server. The data acquisition module includes a dynamic filtering adjustment submodule. The filtering coefficients and sampling frequency of the dynamic filtering adjustment submodule are controlled by the calibration residual parameters output by the calibration execution unit, thereby realizing the adaptive adjustment of the data acquisition strategy. The edge computing node also includes a dynamic benchmark update submodule, which is connected between the calibration execution unit and the health assessment unit. It is used to dynamically update the benchmark parameters of the electrical equipment according to the calibration residual parameters, and feed back the updated benchmark parameters to the health assessment unit to correct the error drift curve and health score.

[0014] Thirdly, a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of an edge computing-based power safety self-calibration method.

[0015] The beneficial effects of this invention are as follows: 1. By constructing a three-layer distributed self-calibration architecture of "device-side data acquisition—edge-side processing—cloud collaboration," health assessment, calibration strategy decision-making, calibration execution, and effect evaluation are decentralized to edge computing nodes closer to the data source. This effectively reduces network transmission latency and bandwidth consumption caused by uploading massive amounts of operational parameter data to the cloud, significantly improving the real-time performance and system reliability of self-calibration response. Based on this, by deploying a health assessment unit, a health score characterizing the degree of equipment degradation is generated in real time based on the historical operating records of the electrical equipment. This provides a quantitative basis for the dynamic selection of calibration strategies on demand, thus avoiding resource waste or insufficient accuracy caused by a uniform calibration mode. Simultaneously, by setting multiple calibration strategies corresponding to different calibration accuracies and computational resource consumption, and dynamically switching the target calibration strategy according to the threshold range of the health score, edge computing nodes can adaptively balance calibration accuracy and resource overhead under limited computational resource constraints. This ensures that high-healthy devices can complete calibration quickly with low overhead, while ensuring that low-healthy devices receive in-depth calibration to correct significant deviations. Furthermore, after the calibration is completed, the residual parameters of this calibration are fed back to the data acquisition module in real time, enabling the sampling frequency and filter coefficients to adaptively adjust according to the residual amplitude and distribution characteristics. This achieves dynamic matching between data acquisition density and signal quality as the device status changes, further improving the accuracy and reliability of the data source. At the same time, it provides more refined raw data support for subsequent calibrations during periods of high residuals.

[0016] 2. By synchronously using calibration residual parameters for dynamic benchmark updates, the reference zero point of the error drift curve relied upon in health assessment is corrected. This allows the error drift curve to respond promptly to the actual degradation trajectory of the equipment, preventing health assessments from deviating from the true state due to a fixed benchmark. This forms a self-closing iterative mechanism at the edge, consisting of "calibration execution—benchmark correction—assessment optimization," continuously improving the accuracy of health assessments and the targeting of self-calibration. Furthermore, calibration effect evaluation indicators and post-calibration parameters are uploaded to the cloud. The cloud performs long-term trend analysis and global strategy optimization for all network devices, and then distributes the optimized strategy parameters in batches to each edge computing node. This achieves a complementary advantage between the real-time response capability of the edge and the global optimization capability of the cloud, enabling the calibration strategy library to continuously evolve with changes in the statistical characteristics of the device group, overcoming the policy rigidity defects of isolated operation of a single edge node or device. Simultaneously, the calibration effect evaluation results are fed back to the health assessment model, continuously adjusting the evaluation parameters. This forms a dual correction link with the dynamic benchmark update; the former optimizes the model itself, while the latter updates the data benchmark, jointly promoting the health assessment model's approximation of the true state of the equipment. Building upon this foundation, by real-time monitoring of changes in computing resource utilization and thermal power consumption resulting from adjustments to the data acquisition strategy, and using resource utilization as a dynamic constraint for calibration strategy selection, a reverse adjustment link of "resource awareness - strategy adaptation" is constructed. This ensures that edge computing nodes automatically downgrade to lightweight calibration strategies when resources are scarce, enhancing the system's operational stability and adaptability under fluctuating resource conditions. Through this dual-loop mechanism of edge-end self-closed-loop and cloud-based collaborative optimization, a comprehensive improvement in accuracy maintenance, resource adaptation, adaptive adjustment, and continuous evolution of the power safety self-calibration system is achieved. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall self-calibration method of the present invention.

[0018] Figure 2 This is a schematic diagram of the calibration trigger judgment and strategy decision process of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figures 1 to 2 This paper presents a power safety self-calibration method and system based on edge computing.

[0021] I. The power safety self-calibration system based on edge computing described in this invention includes a data acquisition module deployed on the side of the power-consuming equipment, an edge computing node deployed on the network edge, and a cloud server that communicates with the edge computing node.

[0022] The data acquisition module includes a dynamic filtering adjustment submodule, whose filtering coefficients and sampling frequency are controlled by the calibration residual parameters output by the calibration execution unit. The edge computing node includes a health assessment unit, a calibration strategy decision unit, a calibration execution unit, a calibration evaluation unit, and a dynamic benchmark update submodule. The dynamic benchmark update submodule is connected between the calibration execution unit and the health assessment unit, and is used to dynamically update the benchmark parameters of the electrical equipment based on the calibration residual parameters, and feed the updated benchmark parameters back to the health assessment unit to correct the error drift curve and health score.

[0023] II. S1: Data Acquisition Steps The data acquisition module is deployed on the side of the electrical equipment and collects the operating parameter data of the equipment in real time through a sensor array. The operating parameter data includes electrical parameters and environmental parameters.

[0024] Electrical parameters include, but are not limited to: voltage (collection range: 0~1000V, accuracy ±0.5%), current (collection range: 0~5000A, accuracy ±0.5%), power (active power, reactive power, apparent power), power factor (0~1), frequency (45~65Hz, accuracy ±0.01Hz), harmonic content (total harmonic distortion (THD), 2nd to 50th harmonics), etc.

[0025] Environmental parameters include, but are not limited to: temperature (sampling range: -40℃~+125℃, accuracy ±0.5℃), humidity (sampling range: 0~100%RH, accuracy ±3%RH), vibration (acceleration, sampling range: 0~10g, accuracy ±0.1g), etc.

[0026] The default sampling frequency of the data acquisition module is 1kHz, and the default filter coefficient is 0.5 (first-order low-pass filter, with a filter coefficient range of 0~1). The dynamic filter adjustment submodule dynamically adjusts the acquisition strategy within the following range based on feedback from the calibration residual parameters: Sampling frequency adjustment range: 100Hz~10kHz; Filter coefficient adjustment range: 0.1~0.9.

[0027] The data acquisition module packages the collected operating parameter data and transmits it to the edge computing node via communication methods such as industrial Ethernet, Wi-Fi, or 5G.

[0028] Input: Analog / digital sensor signals from the operating site of the electrical equipment.

[0029] Processing logic: After the sensor signal is processed by the signal conditioning circuit (amplification, filtering, isolation), it is converted into a digital signal by the ADC analog-to-digital converter (resolution ≥ 16 bits), and then a structured data packet is formed after data preprocessing.

[0030] Output: A data package containing runtime parameters, including timestamps, device IDs, parameter values, and data quality identifiers.

[0031] S2: Health Assessment Steps The health assessment unit is deployed on edge computing nodes. Based on received operating parameter data and historical operating records, it assesses the health of electrical equipment and generates a health score. The health score ranges from 0 to 100, with higher scores indicating better equipment health.

[0032] The health assessment specifically includes the following sub-steps: (1) Generation of error drift curve Based on the historical calibration records of the electrical equipment (containing at least the timestamps and calibration deviation data of the most recent 10 calibrations), the trend of measurement error of the electrical equipment over time is analyzed. A least squares method is used to perform polynomial fitting on the historical error data to generate an error drift curve. The fitting function is in the form of:

[0033] in, For time The prediction error value at that location, These are the fitting coefficients. The cumulative operating time of the equipment (in hours). The fit coefficients are determined by minimizing the sum of squared residuals from the historical data.

[0034] (2) Calculation of aging factors Based on the operating time of electrical equipment (Unit: hours), cumulative workload (Unit: kWh) and environmental stress data Calculate the aging factor of electrical equipment :

[0035] in, This is a reference service life for the equipment (e.g., 87,600 hours, or 10 years). The reference cumulative operating load for the equipment (e.g., 1 million kWh). For reference environmental stress (such as a combined baseline value of temperature 25℃ and humidity 50%RH); , , Let be the weighting coefficient, satisfying Recommended value .

[0036] Environmental stress The calculation method is as follows:

[0037] in, and For the first Temperature and humidity at each sampling time, and The reference temperature and humidity (e.g., 25°C and 50%RH) are used as the reference. and The maximum allowable temperature and humidity for the equipment.

[0038] (3) Health score calculation Based on the error drift curve and aging factor, a health score is calculated according to a pre-defined health assessment model. :

[0039] in, This is the initial health baseline value (set to 100 points when the new equipment leaves the factory); This is the error impact coefficient (recommended value is 0.5~2.0 points / %, representing the deduction of the corresponding points for every 1% measurement error); The aging effect coefficient (recommended value: 10-30 points); This represents the absolute value of the prediction error at the current moment (expressed as a percentage). This is the aging factor (value range 0~1).

[0040] when When the calculation result is less than 0, take ;when When the calculation result is greater than 100, take .

[0041] Input: Operating parameter data, equipment historical operating records (including historical calibration records).

[0042] Processing logic: Read the historical calibration database, perform polynomial fitting to generate error drift curves; read equipment runtime, cumulative workload and environmental parameter records, calculate aging factor; substitute into the health assessment model to calculate score.

[0043] Output: Health score (0~100 points).

[0044] S3: Calibration Strategy Decision-Making Steps The calibration strategy decision unit is deployed on edge computing nodes and is based on health scores. It dynamically selects the target calibration strategy from a variety of preset calibration strategies.

[0045] (1) Health threshold range division Three preset health threshold ranges: First threshold interval: A score of 1 indicates that the equipment is in good health. Second threshold interval: A score of 1 indicates that the device is in a moderate health condition. Third threshold interval: A score of 0 indicates that the equipment is in poor health. (2) Calibration strategy mapping relationship Select the corresponding calibration strategy based on the preset mapping relationship between the target threshold range and the calibration strategy: First calibration strategy (fast calibration strategy): When Time-sharing selection. Verification and correction are performed only on critical parameters (RMS values ​​of voltage and current). Computational resource consumption is approximately 30% of the standard calibration strategy, and calibration time is approximately 40% of the standard calibration strategy. Calibration accuracy target: Critical parameter error ≤2%.

[0046] Second calibration strategy (standard calibration strategy): When Time-sharing selection. Verification and correction are performed on all operating parameters (voltage, current, power, power factor, frequency, etc.). Computational resource consumption is based on the standard baseline (set to 1.0), and calibration time is approximately the standard time (set to 1.0). Calibration accuracy target: Error of all parameters ≤1%.

[0047] Third calibration strategy (deep calibration strategy): When Time-sharing selection. All operating parameters are verified and corrected, and additional accuracy compensation calculations are performed (including temperature compensation, nonlinearity compensation, and cross-sensitivity compensation). Computational resource consumption is approximately 2.5 times that of the standard calibration strategy, and calibration time is approximately 2.0 times that of the standard calibration strategy. Calibration accuracy target: All parameter errors ≤ 0.3%.

[0048] Input: Health score .

[0049] Processing logic: Compare with the preset threshold range to determine the range to which it belongs, and select the corresponding calibration strategy according to the mapping relationship.

[0050] Output: Target calibration strategy identifier (first / second / third calibration strategy).

[0051] Calibration trigger judgment step (executed before S3) Before executing S3, it is first determined whether the preset self-calibration trigger conditions are met. S3 is triggered when any of the following conditions are met: Health Trigger: Health Score Below the preset health threshold (recommended threshold) point).

[0052] Parameter deviation trigger: The operating parameters of the electrical equipment deviate beyond the preset deviation threshold. Parameter deviation The calculation method is the percentage relative deviation between the current measurement value and the historical benchmark value. Triggered at time. Among them, , This is the current measured parameter value. This is a historical benchmark value.

[0053] Environmental change trigger: The environmental parameters of the electrical equipment change beyond the preset environmental change threshold. Temperature change or humidity change Triggered during RH.

[0054] Time interval trigger: The time interval since the last self-calibration exceeds the preset time interval threshold (the recommended threshold is 720 hours, or 30 days).

[0055] Load change trigger: The load on the electrical equipment changes abruptly beyond a preset range. (Load change rate) Triggered at time, where For the current load, The load from the previous moment.

[0056] Prediction trigger: based on the error drift curve By extrapolating trends, we can predict when the health score of electrical equipment will drop below the health threshold within a preset time period in the future. .when Self-calibration is triggered at time, where The warning time threshold is set (a recommended value is 168 hours, or 7 days). The prediction method is: Let... Extrapolate the fitted function to the future and calculate achieve The time.

[0057] Inputs: health score, operating parameters, environmental parameters, last calibration time, load data, and error drift curve.

[0058] Processing logic: Each trigger condition is checked one by one, and a trigger signal is generated if any condition is met.

[0059] Output: Trigger signal (True / False) and trigger reason code.

[0060] S4: Calibration Execution Steps The calibration execution unit is deployed on edge computing nodes and performs self-calibration on the operating parameters of electrical equipment according to the target calibration strategy.

[0061] (1) Calibration model The calibration execution unit has a built-in calibration model for any parameter being calibrated. calibrated parameter values for:

[0062] in, These are the original measured values. This is the system deviation compensation amount. This is the drift compensation amount.

[0063] System deviation compensation amount Derived from the calibration data at the time of device delivery, and stored in the local calibration strategy library. Drift compensation amount. From the error drift curve The extrapolated value is determined at the current moment.

[0064] (2) Differences in the implementation of different calibration strategies First calibration strategy (fast calibration): The above calibration calculation is performed only on the effective values ​​of voltage and current; other parameters are passed directly. The calibration calculation uses linear interpolation, and the computational workload is approximately 100 floating-point operations.

[0065] The second calibration strategy (standard calibration) involves performing the above calibration calculations on all electrical parameters (voltage, current, power, power factor, frequency). The calibration calculations employ a piecewise linear fitting method, requiring approximately 500 floating-point operations.

[0066] The third calibration strategy (deep calibration): Calibration calculations are performed on all electrical parameters, with additional temperature compensation, nonlinearity compensation, and cross-sensitivity compensation. Temperature compensation coefficient. (Unit: % / ℃) Obtained according to the equipment factory calibration, the compensation amount is... Nonlinear compensation employs cubic spline interpolation; cross-sensitivity compensation considers the mutual influence between voltage and current measurement channels and uses a decoupling matrix method. The total computational load is approximately 1500 floating-point operations.

[0067] (3) Generation of calibration residual parameters :

[0068] in, The number of parameters to be calibrated. For the first The original measured values ​​of each parameter, For the first After the parameters are calibrated, the calibration execution unit generates the calibration residual parameters for this calibration upon completion of the calibration. For the first The full-scale value of each parameter.

[0069] The calibration residual parameters also include the deviation distribution characteristics, i.e., the mean of the residuals of each parameter. Standard deviation and maximum residual .

[0070] Input: target calibration strategy, raw operating parameter data, local calibration strategy library (including calibration data).

[0071] Processing logic: Select a calibration algorithm based on the target calibration strategy, perform calibration calculations, and generate post-calibration parameters and residual parameters.

[0072] Output: calibrated parameter data, calibration residual parameters (including mean, standard deviation, and maximum residual).

[0073] S5: Calibration Evaluation Steps The calibration evaluation unit is deployed on edge computing nodes to evaluate the self-calibration results and generate calibration effect evaluation indicators.

[0074] (1) Calculation of calibration correction amount The calibrated parameter data is compared with the operating parameter data before calibration, and the calibration correction amount is calculated. :

[0075] (2) Calculation of accuracy improvement Calculate the accuracy improvement of self-calibration based on the calibration correction amount and the preset reference standard value. :

[0076] in, This represents the overall error before calibration (estimated based on historical data). This is the overall error after calibration (estimated based on the calibration residual parameters). Positive values ​​indicate improved accuracy, while negative values ​​indicate deteriorated accuracy.

[0077] (3) Stability assessment Within the preset monitoring period after self-calibration (recommended monitoring duration is 60 minutes), continuously monitor the fluctuations of the operating parameters of the electrical equipment and generate stability assessment indicators. :

[0078] in, The standard deviation of the parameters before calibration. This represents the standard deviation of the parameters during the monitoring period after calibration. The value ranges from 0 to 1, and the larger the value, the more stable the parameter is after calibration.

[0079] (4) Generation of calibration effect evaluation indicators Based on accuracy improvement and stability assessment indicators Generate calibration effect evaluation indicators according to preset rating rules.

[0080] excellent: and ; good: and ; generally: or ; Poor: 0 (Accuracy deterioration) Input: Pre-calibration parameter data, post-calibration parameter data, calibration residual parameters.

[0081] Processing logic: Calculate calibration correction amount and accuracy improvement; continuously collect parameters and calculate standard deviation during the monitoring period; generate a rating by combining accuracy improvement and stability indicators.

[0082] Output: Calibration correction amount Accuracy improvement Stability assessment indicators Calibration effect evaluation indicators .

[0083] Closed-loop feedback steps (after S5) Based on calibration effectiveness evaluation indicators Adjust the assessment parameters used in health assessment step S2 to achieve continuous optimization of the health assessment model: when If the rating is "excellent", keep the current evaluation parameters unchanged.

[0084] when When the result is "good", the weighting coefficient is slightly adjusted (adjustment range ±5%).

[0085] when When the value is "normal", the weighting coefficient should be adjusted appropriately (adjustment range ±10%).

[0086] when When the result is "poor", the evaluation model parameter recalibration process is triggered.

[0087] Input: Calibration effectiveness evaluation indicators .

[0088] Processing logic: Select the corresponding parameter adjustment strategy based on the rating results and update the health assessment model parameters.

[0089] Output: Updated health assessment model parameters.

[0090] S6: Adaptive Adjustment Steps for Acquisition Strategy Calibration residual parameters The data is fed back to the data acquisition module in real time, and the data acquisition module dynamically adjusts the data acquisition strategy based on the amplitude and distribution characteristics of the calibration residual parameters.

[0091] (1) Setting the residual threshold First residual threshold (High threshold): Recommended value is 5%; Second residual threshold (Low threshold): Recommended value is 1%. (2) Data collection strategy adjustment rules when When the residual exceeds the high threshold: the sampling frequency is increased from the current value to twice the current value (maximum not exceeding 10kHz), and the filter coefficient is decreased from the current value to 0.5 times the current value (minimum not less than 0.1).

[0092] when When the residual is below the low threshold: the sampling frequency is reduced from the current value to 0.5 times the current value (minimum not lower than 100Hz), and the filter coefficient is increased from the current value to 1.5 times the current value (maximum not exceeding 0.9).

[0093] when Time: Keep the current data collection strategy unchanged.

[0094] (3) Adjusting smoothing processing To avoid system oscillations caused by frequent switching of acquisition strategies, a hysteresis control strategy is adopted: when Crossing from low to high Immediately trigger an increase in sampling frequency; when Crossing from high to low Immediately trigger a reduction in sampling frequency; the intermediate region remains unchanged.

[0095] Input: Calibration residual parameters (including mean) Standard deviation .

[0096] Processing logic: comparison Based on the comparison results with the residual threshold, a new sampling frequency and filter coefficient are calculated, and control commands are output to the dynamic filter adjustment submodule.

[0097] Output: New sampling frequency value and filter coefficient value.

[0098] Optimal calibration strategy under resource constraints (after S6) Real-time monitoring of changes in computing resource utilization and thermal power consumption caused by adjustments in data acquisition strategies is used as dynamic constraints for calibration strategy decisions in S3. Calculate resource utilization (Unit: %): Obtained in real time through the CPU monitoring module of the edge computing node.

[0099] Thermal power consumption (Unit: W): Real-time data is obtained through the temperature sensor and power consumption monitoring module of the edge computing node.

[0100] when or When the maximum allowable thermal power consumption of the device is 85% of the rated power consumption, the calibration strategy decision unit should prioritize the calibration strategy with lower computational resource consumption, i.e., prioritize the first calibration strategy, then the second calibration strategy, and avoid the third calibration strategy.

[0101] Input: CPU utilization and thermal power consumption data.

[0102] Processing logic: Determine if resource usage exceeds the threshold; if so, send a constraint signal to the calibration strategy decision unit.

[0103] Output: Resource constraint signal (high / low).

[0104] S7: Dynamic Baseline Update Steps The calibration residual parameters are used synchronously to update the dynamic reference parameters of the electrical equipment. The dynamic reference parameters are the reference zero point for calculating the error drift curve in the health assessment step S2.

[0105] (1) Baseline parameter update Dynamic reference parameters The update formula is:

[0106] in, This is the current baseline parameter value. To update the step size coefficient (recommended value is 0.1~0.3), For calibration residual parameters (expressed as a percentage). This is the full-scale value of the parameter.

[0107] (2) Error drift curve correction The health assessment unit recalculates the error drift curve based on the updated dynamic baseline parameters. Specifically, the updated baseline parameters are used as the new reference zero point; historical error data is shifted and corrected; polynomial fitting is then performed again to generate the corrected error drift curve. .

[0108] (3) Dynamic correction of health score Based on the corrected error drift curve With updated aging factors, the health score is recalculated to achieve dynamic correction of the health score.

[0109] Input: Calibration residual parameters Current reference parameters .

[0110] Processing logic: Calculate new baseline parameters and update the local baseline parameter library; trigger the health assessment unit to recalculate the error drift curve and health score.

[0111] Output: Updated dynamic baseline parameters, corrected error drift curve, and corrected health score.

[0112] S8: Cloud Collaboration Optimization Steps Upload the calibration effectiveness evaluation indicators and / or calibrated parameter data to the cloud server via an encrypted communication channel. Uploads can be made immediately after each calibration or in batches according to a preset cycle (24 hours recommended).

[0113] After receiving the data, the cloud server performs the following processing: (1) Long-term trend analysis The cloud server aggregates calibration data uploaded by all connected edge computing nodes to construct long-term trend analyses across three dimensions: device-level, site-level, and global-level. Device-level trends: The changing trend of a single device's health score over time. Site-level trends: Statistical distribution and changing trends of health scores for all devices at the same site.

[0114] Global-level trend: Overall distribution and changing trend of health scores for all connected devices (2) Global strategy optimization Based on long-term trend analysis results, the cloud server uses machine learning methods (random forest or gradient boosting decision tree GBDT recommended) to globally optimize the calibration strategy parameters. The optimization objective function is:

[0115] in, For global average accuracy improvement, This represents the global average CPU utilization. This is the weighting coefficient (recommended value: 0.5).

[0116] The optimization variables include: boundary values ​​of the health threshold range, accuracy targets of each calibration strategy, and threshold values ​​of the calibration trigger conditions.

[0117] The optimized calibration strategy parameters are encapsulated in JSON format and pushed to each edge computing node through the cloud server's distribution channel.

[0118] Input: Calibration effect evaluation indicators, calibrated parameter data (including device ID, timestamp, geographical location, etc.).

[0119] Processing logic: Data aggregation and cleaning → Multi-dimensional trend analysis → Global strategy optimization → Generation of optimization parameter package.

[0120] Output: Optimized calibration strategy parameters (JSON format).

[0121] S9: Strategy Update Steps After receiving the optimized calibration strategy parameters from the cloud server, the edge computing node updates its local calibration strategy library. (1) Parameter verification Edge computing nodes perform integrity checks (checking whether all required fields are included) and rationality checks (checking whether parameter values ​​are within a preset reasonable range, such as thresholds being between 0 and 100) on the received optimization parameters.

[0122] (2) Policy library update Once verification is successful, the optimized parameters will be written to the local calibration policy library, replacing the corresponding old parameters. The update uses an atomic operation method to ensure that a power outage during the update process will not cause damage to the policy library.

[0123] (3) Update confirmation After the update is complete, the edge computing node sends an update confirmation message to the cloud server, which includes the policy version number and timestamp of the successful update.

[0124] Input: Optimized calibration strategy parameters (JSON format) sent from the cloud.

[0125] Processing logic: Parameter validation → Atomic update of policy library → Generation of confirmation message.

[0126] Output: Updated local calibration policy library, update confirmation message.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A power safety self-calibration method based on edge computing, characterized in that, Includes the following steps: S1: Data acquisition step, the operating parameter data of the electrical equipment is collected in real time by a data acquisition module deployed on the side of the electrical equipment, the operating parameter data including electrical parameters and environmental parameters; S2: Health assessment step, through the health assessment unit deployed on the edge computing node, the health of the electrical equipment is assessed based on the operating parameter data and the equipment's historical operating records, and a health score is generated; S3: Calibration strategy decision step, through the calibration strategy decision unit deployed on the edge computing node, dynamically selects the target calibration strategy from a variety of preset calibration strategies based on the health score. The variety of calibration strategies includes at least a first calibration strategy, a second calibration strategy, and a third calibration strategy. Different calibration strategies have different calibration accuracies and computing resource consumption. S4: Calibration execution step: The calibration execution unit deployed on the edge computing node performs self-calibration on the operating parameters of the electrical equipment according to the target calibration strategy, generates calibrated parameter data, and generates calibration residual parameters for this calibration. The calibration residual parameters are used to characterize the deviation amount and deviation distribution characteristics between the operating parameters before and after calibration. S5: Calibration evaluation step, using a calibration evaluation unit deployed on the edge computing node to evaluate the self-calibration results and generate calibration effect evaluation indicators; S6: Adaptive adjustment step of acquisition strategy, the calibration residual parameter is fed back to the data acquisition module in real time, and the data acquisition module dynamically adjusts the data acquisition strategy according to the amplitude and distribution characteristics of the calibration residual parameter: when the calibration residual parameter exceeds the first residual threshold, the sampling frequency is increased and the filtering coefficient is decreased; when the calibration residual parameter is lower than the second residual threshold, the sampling frequency is decreased and the filtering coefficient is increased; S7: Dynamic benchmark update step, the calibration residual parameters are synchronously used to update the dynamic benchmark parameters of the electrical equipment. The dynamic benchmark parameters are the reference zero point for calculating the error drift curve in the health assessment step S2. The health assessment unit recalculates the error drift curve based on the updated dynamic benchmark parameters and dynamically corrects the health score. S8: Cloud-based collaborative optimization step, uploading the calibration effect evaluation index and / or calibrated parameter data to the cloud server, whereby the cloud server performs long-term trend analysis and global strategy optimization, generates optimized calibration strategy parameters, and distributes them to the edge computing node; S9: Strategy update step, the edge computing node updates the local calibration strategy library according to the optimized calibration strategy parameters issued by the cloud server.

2. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, In step S2, the health assessment includes: Based on the historical calibration records of the electrical equipment, analyze the trend of the measurement error of the electrical equipment over time and generate an error drift curve; The aging factor of the electrical equipment is calculated based on the operating time, cumulative workload, and environmental stress data of the electrical equipment. The health score is calculated based on the error drift curve and the aging factor, according to a preset health assessment model.

3. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, In step S3, the dynamic target selection calibration strategy includes: The health score is compared with multiple preset health threshold ranges to determine the target threshold range to which the health score belongs; Based on the preset mapping relationship between the target threshold range and the calibration strategy, the corresponding calibration strategy is selected as the target calibration strategy; Specifically, when the health score falls within a first threshold range, the first calibration strategy is selected. This first calibration strategy is a rapid calibration strategy that only verifies and corrects key parameters. When the health score falls within a second threshold range, the second calibration strategy is selected. This second calibration strategy is a standard calibration strategy that verifies and corrects all operating parameters. When the health score falls within a third threshold range, the third calibration strategy is selected. This third calibration strategy is a deep calibration strategy that verifies and corrects all operating parameters and performs additional accuracy compensation calculations.

4. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, Before step S3, the procedure also includes: The calibration trigger judgment step determines whether the preset self-calibration trigger condition is met. When the self-calibration trigger condition is met, step S3 is triggered. The self-calibration triggering condition includes at least one of the following: The health score is lower than the preset health threshold; The deviation of the operating parameters of the electrical equipment exceeds a preset deviation threshold; The change in the environmental parameters of the electrical equipment exceeds a preset environmental change threshold; The time interval since the last self-calibration exceeds the preset time interval threshold; The load of the electrical equipment experienced a sudden change exceeding a preset range.

5. The power safety self-calibration method based on edge computing according to claim 4, characterized in that, The self-calibration triggering condition also includes a prediction triggering condition; Based on the error drift curve, trend extrapolation is performed to predict the time point when the health score of the electrical equipment will drop below the health threshold within a preset time period in the future. When the difference between the predicted time point and the current time is less than the preset warning time threshold, self-calibration is triggered.

6. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, In step S5, the calibration effect evaluation includes: The calibrated parameter data is compared with the operating parameter data before calibration, and the calibration correction amount is calculated. The accuracy improvement of the self-calibration is calculated based on the calibration correction amount and the preset reference standard value. During the preset monitoring period after the self-calibration is completed, the fluctuation of the operating parameters of the electrical equipment is continuously monitored, and stability evaluation indicators are generated. Based on the accuracy improvement and the stability evaluation index, the calibration effect evaluation index is generated according to the preset rating rules.

7. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, After step S5, the method further includes: The closed-loop feedback step adjusts the evaluation parameters used in the health assessment step S2 according to the calibration effect evaluation index, thereby achieving continuous optimization of the health assessment model.

8. The power safety self-calibration method based on edge computing according to claim 1, characterized in that, In step S6, after the data acquisition module dynamically adjusts the data acquisition strategy based on the amplitude and distribution characteristics of the calibration residual parameters, it further includes: Real-time monitoring of changes in computing resource utilization and thermal power consumption caused by adjustments to the data acquisition strategy is used as dynamic constraints for the calibration strategy decision in step S3: when the computing resource utilization is higher than a preset resource threshold, a calibration strategy with lower computing resource consumption is selected first.

9. A power safety self-calibration system based on edge computing, characterized in that, include: A data acquisition module, deployed on the side of the electrical equipment, is used to collect the operating parameter data of the electrical equipment in real time, including electrical parameters and environmental parameters; An edge computing node, communicatively connected to the data acquisition module, is deployed at the network edge near the power-consuming equipment. The edge computing node includes: The health assessment unit is used to assess the health of the electrical equipment based on the operating parameter data and the equipment's historical operating records, and generate a health score. The calibration strategy decision unit is used to dynamically select a target calibration strategy from a variety of preset calibration strategies based on the health score. The variety of calibration strategies includes at least a first calibration strategy, a second calibration strategy, and a third calibration strategy. Different calibration strategies have different calibration accuracies and computational resource consumption. The calibration execution unit is used to perform self-calibration on the operating parameters of the electrical equipment according to the target calibration strategy, generate calibrated parameter data, and generate calibration residual parameters for this calibration. The calibration residual parameters are used to characterize the deviation amount and deviation distribution characteristics between the operating parameters before and after calibration. The calibration evaluation unit is used to evaluate the self-calibration results of the calibration execution unit and generate calibration effect evaluation indicators. A cloud server, which is communicatively connected to the edge computing node, is used to receive calibration effect evaluation indicators and / or calibrated parameter data uploaded by the edge computing node, perform long-term trend analysis and global strategy optimization, and send the optimized calibration strategy parameters to the edge computing node. The edge computing node updates its local calibration strategy library based on the optimized calibration strategy parameters issued by the cloud server. The data acquisition module includes a dynamic filtering adjustment submodule. The filtering coefficients and sampling frequency of the dynamic filtering adjustment submodule are controlled by the calibration residual parameters output by the calibration execution unit, thereby realizing the adaptive adjustment of the data acquisition strategy. The edge computing node also includes a dynamic benchmark update submodule, which is connected between the calibration execution unit and the health assessment unit. It is used to dynamically update the benchmark parameters of the electrical equipment according to the calibration residual parameters, and feed back the updated benchmark parameters to the health assessment unit to correct the error drift curve and health score.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the edge computing-based power safety self-calibration method as described in any one of claims 1 to 8.

Images