Method and system for evaluating operation state of power equipment by adopting artificial intelligence

Abstract

The invention relates to the technical field of power equipment evaluation, and discloses a power equipment operation state evaluation method and system adopting artificial intelligence. The method comprises the following steps: firstly, collecting multi-dimensional operation parameters of power equipment, and constructing a dynamic state model; setting an artificial intelligence assessment framework, verifying the performance of the framework, and obtaining assessment capability indexes; integrating the framework into a dynamic state model; calculating a theoretical evaluation demand and an evaluation data flow by analyzing total energy consumption in a specified time period and combining an evaluation capability index; generating an adjustment instruction of the evaluation control unit based on the evaluation data traffic and performing parameter adjustment; and finally carrying out quantitative analysis on the evaluation effect. According to the method, through multi-dimensional parameter integration, framework performance verification, dynamic adjustment and effect quantification, the comprehensiveness, reliability and adaptability of power equipment operation state evaluation are improved, and the method is suitable for state monitoring and evaluation of various types of power equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment evaluation, in particular to a power equipment operation state evaluation method and system using artificial intelligence. BACKGROUND

[0002] In the stable operation of the power system, the state evaluation of the power equipment is an important link to ensure the reliability of power supply. With the continuous expansion of the power network and the improvement of the complexity of the equipment, the traditional power equipment operation state evaluation method gradually exposes many limitations.

[0003] Traditional methods mostly rely on manual inspection or single parameter monitoring, which is difficult to fully capture the running state of the equipment. For example, by monitoring only a few parameters such as temperature or voltage of the equipment, it is impossible to reflect the complex physical changes and potential failure risks inside the equipment. At the same time, these methods often use fixed evaluation cycles, and for equipment with large fluctuations in running state, the problem of evaluation lag or over-evaluation is likely to occur.

[0004] With the increase of the running time of the power equipment, the aging and performance degradation of the equipment show nonlinear characteristics, and the traditional evaluation method based on empirical model is difficult to accurately describe these dynamic changes. In large-scale power systems, the massive operation data also put forward higher requirements for the real-time and accuracy of the evaluation method, and the traditional data processing and analysis means have been difficult to meet the actual demand.

[0005] In recent years, the application of artificial intelligence technology in various industries has provided a new idea for power equipment state evaluation, but the existing evaluation methods based on artificial intelligence still have shortcomings. Some methods fail to fully integrate the multi-dimensional running parameters of the equipment, resulting in insufficient comprehensiveness of the evaluation model; some methods lack performance verification and dynamic adjustment in the evaluation framework, making the reliability and adaptability of the evaluation results need to be improved. Therefore, an artificial intelligence evaluation method that can comprehensively utilize multi-dimensional parameters, has dynamic adjustment capability and quantifiable evaluation effect is urgently needed to improve the accuracy and effectiveness of the power equipment operation state evaluation. SUMMARY

[0006] The purpose of the present application is to provide a power equipment operation state evaluation method and system using artificial intelligence to solve the problems raised in the background.

[0007] To achieve the above purpose, the present application provides a power equipment operation state evaluation method using artificial intelligence, which comprises: Collecting multi-dimensional running parameters of the power equipment during operation, and constructing a dynamic state model of the power equipment based on the multi-dimensional running parameters; setting an artificial intelligence evaluation framework, and performing performance verification on the artificial intelligence evaluation framework to obtain an evaluation capability index of the artificial intelligence evaluation framework; integrating the artificial intelligence evaluation framework into the dynamic state model; analyzing total energy consumption of the power equipment within a specified time period, calculating theoretical evaluation requirements of the artificial intelligence evaluation framework in the specified time period, combining the evaluation capability index of the artificial intelligence evaluation framework, and calculating evaluation data traffic of the artificial intelligence evaluation framework in the specified time period; generating adjustment instructions of an evaluation control unit in the artificial intelligence evaluation framework based on the evaluation data traffic of the artificial intelligence evaluation framework in the specified time period; performing parameter adjustment operations on the evaluation control unit based on the adjustment instructions of the evaluation control unit; quantitative analysis of the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment.

[0008] Preferably, the collection of multi-dimensional running parameters of the power equipment during operation and the construction of the dynamic state model of the power equipment based on the multi-dimensional running parameters comprise: obtaining physical structure information of the power equipment, and dividing the power equipment into multiple functional sub-regions; extracting parameter monitoring points of each functional sub-region to form a multi-dimensional running parameter set of the power equipment; calculating the energy accumulation capacity of each parameter monitoring point, and identifying external interference factors acting on the parameter monitoring points; based on the energy accumulation capacity of each parameter monitoring point and the external interference factors, establishing an energy balance equation of the power equipment.

[0009] Preferably, the establishment of the energy balance equation of the power equipment based on the energy accumulation capacity of each parameter monitoring point and the external interference factors comprises: extracting the interaction effect between each parameter monitoring point and the influence degree of the external interference factors on each parameter monitoring point, and constructing the energy balance equation of the power equipment based on the energy accumulation capacity of each parameter monitoring point, the interaction effect between each parameter monitoring point and the influence degree of the external interference factors on each parameter monitoring point.

[0010] Preferably, the setting of the artificial intelligence evaluation framework and the performance verification of the artificial intelligence evaluation framework to obtain the evaluation capability index of the artificial intelligence evaluation framework comprise: verifying the evaluation performance of the artificial intelligence evaluation framework in the data processing stage, the model inference stage and the result output stage respectively, and obtaining the evaluation capability index of the artificial intelligence evaluation framework; In the verification of the data processing stage, the pre-processing ability of the test framework for the multi-dimensional operating parameters of power equipment needs to be tested; the verification of the model inference stage focuses on evaluating the real-time performance and accuracy of the artificial intelligence algorithm; the verification of the result output stage focuses on evaluating the stability and explainability of the evaluation results.

[0011] Preferably, the integration of the artificial intelligence evaluation framework into the dynamic state model comprises: analyzing the parameter monitoring points in the dynamic state model, identifying the parameter monitoring points with the greatest influence, and setting the artificial intelligence evaluation framework at the corresponding positions in the dynamic state model mapped from the parameter monitoring points with the greatest influence.

[0012] Preferably, the analysis of the total energy consumption of the power equipment in a specified time period and the calculation of the theoretical evaluation demand of the artificial intelligence evaluation framework in the specified time period, combined with the evaluation capability index of the artificial intelligence evaluation framework, to calculate the evaluation data flow of the artificial intelligence evaluation framework in the specified time period comprises: calculating and summarizing the energy input of each parameter monitoring point in a specified time period to obtain the total energy consumption of the power equipment in the specified time period; obtaining the energy consumed by the artificial intelligence evaluation framework in the specified time period, and combining the total energy consumption of the power equipment in the specified time period to obtain the theoretical evaluation demand of the artificial intelligence evaluation framework in the specified time period; combining the theoretical evaluation demand of the artificial intelligence evaluation framework in the specified time period with the evaluation capability index of the artificial intelligence evaluation framework to obtain the evaluation data flow of the artificial intelligence evaluation framework in the specified time period.

[0013] Preferably, the generation of the adjustment instruction of the evaluation control unit in the artificial intelligence evaluation framework based on the evaluation data flow of the artificial intelligence evaluation framework in the specified time period comprises: testing the unit data flow processing capability of the evaluation control unit in the activated state, and combining the evaluation data flow of the artificial intelligence evaluation framework in the specified time period to determine the activation time length of the evaluation control unit.

[0014] Preferably, the parameter adjustment operation of the evaluation control unit based on the adjustment instruction of the evaluation control unit comprises: dividing the specified time period into a plurality of discrete time period units, and uniformly distributing the activation time length of the evaluation control unit into the plurality of discrete time period units; after the end of the previous discrete time period unit, determining whether the data flow processing of the evaluation control unit in the discrete time period unit reaches the expected target; The activation time length of the evaluation control unit in the latter discrete time period unit is adaptively corrected until the parameter adjustment operation of the evaluation control unit in the specified time period is completed.

[0015] Preferably, the quantitative analysis of the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment comprises: obtaining the parameter variation of the power equipment before and after the state evaluation; obtaining the actual evaluation demand of the power equipment based on the parameter variation; comparing and analyzing the actual evaluation demand and the theoretical evaluation demand to obtain the control weight of the evaluation control unit in the artificial intelligence evaluation framework; quantitatively analyzing the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment based on the control weight.

[0016] Preferably, the present application further comprises a power equipment running state evaluation system using artificial intelligence, which is used to realize the power equipment running state evaluation method using artificial intelligence as described above, and the system comprises: a dynamic state model construction module, which is used to collect multi-dimensional running parameters of the power equipment in the running process, and construct a dynamic state model of the power equipment based on the multi-dimensional running parameters; an artificial intelligence evaluation framework construction module, which is used to set an artificial intelligence evaluation framework, and verify the performance of the artificial intelligence evaluation framework to obtain an evaluation capability index of the artificial intelligence evaluation framework; a framework integration module, which is used to integrate the artificial intelligence evaluation framework into the dynamic state model; an evaluation data flow calculation module, which is used to analyze the total energy consumption of the power equipment in a specified time period, calculate the theoretical evaluation demand of the artificial intelligence evaluation framework in the specified time period, and calculate the evaluation data flow of the artificial intelligence evaluation framework in the specified time period in combination with the evaluation capability index of the artificial intelligence evaluation framework; an adjustment instruction generation module, which is used to generate adjustment instructions of the evaluation control unit in the artificial intelligence evaluation framework based on the evaluation data flow of the artificial intelligence evaluation framework in the specified time period; a parameter adjustment control module, which is used to perform parameter adjustment operation on the evaluation control unit based on the adjustment instructions of the evaluation control unit; an evaluation effect verification module, which is used to quantitatively analyze the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment.

[0017] Compared with the prior art, the present application has the following advantages: By collecting multi-dimensional operating parameters to construct a dynamic state model, the state characteristics of the power equipment under different operating conditions can be comprehensively reflected. The integration of multi-dimensional parameters covers voltage, current, temperature, vibration and other aspects during equipment operation, so that the model can capture potential correlations and abnormal signals that cannot be reflected by a single parameter, thereby being closer to the actual operation of the equipment.

[0018] An artificial intelligence evaluation framework is set up and performance verification is carried out to obtain evaluation capability indicators, providing a reliable benchmark for the evaluation process. The performance verification link can discover differences in the performance of the framework under different scenarios in advance, ensuring that it has stable evaluation capability in actual application and reducing evaluation bias caused by defects in the framework itself.

[0019] The artificial intelligence evaluation framework is integrated into the dynamic state model, realizing the organic combination of the model and the evaluation algorithm. This integration is not a simple superposition, but an evaluation framework that can operate based on the real-time updated parameters of the dynamic model, so that the evaluation process is synchronized with the changes in the state of the equipment, enhancing the timeliness of the evaluation.

[0020] The total energy consumption is analyzed and the theoretical evaluation demand and evaluation data flow are calculated, providing data basis for the adjustment of the evaluation control unit. By combining the evaluation capability indicators, the data processing amount in the evaluation process can be reasonably planned to avoid excessive system burden caused by excessive data flow or insufficient data affecting evaluation accuracy, making the evaluation process more efficient.

[0021] Based on the evaluation data flow, adjustment instructions are generated and the parameters of the evaluation control unit are adjusted, giving the evaluation system the ability to adapt. When the operating state of the equipment changes significantly, the control unit can adjust the parameters in time to change the evaluation frequency, data sampling density, etc., so that the evaluation method can adapt to different operating scenarios. Whether it is a smooth stage of normal equipment operation or a complex stage of abnormal fluctuations, the evaluation method can maintain good evaluation performance.

[0022] Quantitative analysis of the evaluation effect can clearly present the actual performance of the evaluation method. Through quantitative indicators, the degree of agreement between the evaluation result and the actual state of the equipment can be intuitively understood, which facilitates the discovery of problems in the evaluation process and targeted optimization, continuously improves the evaluation system, and improves the overall evaluation level. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The working principle diagram of the power equipment operating state evaluation method using artificial intelligence described in the present application; Figure 2 Flowchart for constructing a dynamic state model; Figure 3 Flowchart for integrating artificial intelligence evaluation framework; Figure 4 Flowchart for calculating evaluation data flow; Figure 5 Flowchart for adjusting control unit parameters for evaluation. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0025] Please refer to Figure 1 The present application provides a power equipment operation state evaluation method using artificial intelligence, which comprises the following steps: A dynamic state model is constructed by collecting multi-dimensional operation parameters of power equipment, and an artificial intelligence evaluation framework is integrated to achieve efficient evaluation. The specific process includes: collecting multi-dimensional parameters such as voltage, current, temperature and vibration during the operation of power equipment, and establishing a dynamic state model based on the correlation between the parameters; designing an artificial intelligence evaluation framework, verifying its performance in data processing, model reasoning and result output stage, and obtaining evaluation capability indicators; integrating the framework into the dynamic state model, calculating the evaluation data flow by analyzing the total energy consumption of the power equipment and the theoretical evaluation demand of the framework; generating adjustment instructions for the evaluation control unit according to the data flow, and dynamically adjusting its parameters; finally quantitatively analyzing the evaluation effect, and realizing accurate evaluation of the operation state of power equipment.

[0026] Embodiment 1: refer to Figure 2 The construction process of the dynamic state model of power equipment is based on multi-dimensional operation parameters. By analyzing the energy changes of each functional area inside the equipment and external interference factors, a mathematical model that can reflect the actual operation state of the equipment is established. The core of this model is to accurately capture the energy flow and loss characteristics during the operation of power equipment, thereby providing reliable data support for subsequent artificial intelligence evaluation.

[0027] Firstly, the power equipment is divided into multiple functional sub-regions according to its physical structural features. For example, for a transformer device, it can be mainly divided into a winding region, a core region, a cooling system, and an insulating oil circulation system, etc. Each functional sub-region plays a different role in the operation of the equipment, and the change of its operating state directly affects the performance of the overall equipment. The winding region is responsible for the transmission and conversion of electrical energy, and its operating state can be characterized by parameters such as current and temperature; the core region involves magnetic circuit closure and magnetic loss, and its state parameters include magnetic flux density and core temperature; the cooling system reflects the heat dissipation efficiency through parameters such as oil temperature and flow rate. By dividing the functional sub-regions, it is possible to more accurately locate abnormalities or potential problems in the operation of the equipment.

[0028] Within each functional sub-region, corresponding parameter monitoring points are set to collect multi-dimensional operating data. For example, the monitoring points of the winding region can include hot spot temperature sensors, current transformers, and partial discharge detection devices; the monitoring points of the core region can be configured with magnetic flux sensors and vibration sensors; the monitoring points of the cooling system include oil temperature sensors, oil flow meters, etc. These monitoring points constitute a set of multi-dimensional operating parameters of the power equipment, covering electrical, mechanical, and thermodynamic state information. The collected data not only includes real-time measurement values, but also records the historical change trend to analyze the dynamic evolution law of the parameters.

[0029] For each parameter monitoring point, its energy accumulation capability needs to be calculated. The energy accumulation capability reflects the characteristics of the monitoring point in absorbing, storing, or releasing energy during operation. For example, the energy accumulation capability of the winding monitoring point can be calculated through its resistance loss and heat capacity characteristics, which embodies the relationship between temperature rise and heat storage; the energy accumulation capability of the core monitoring point is related to the hysteresis loss and eddy current loss, which manifests as the efficiency of magnetic energy conversion into heat energy. By quantifying the energy accumulation capability of each monitoring point, the role of different regions in energy balance can be clearly defined.

[0030] At the same time, external interference factors acting on each monitoring point need to be identified. External interference factors include environmental temperature fluctuations, grid voltage fluctuations, load surges, etc. These factors directly affect the parameter changes of the monitoring points, for example, an increase in environmental temperature may cause a decrease in cooling system efficiency, which in turn causes an abnormal rise in winding temperature; grid voltage fluctuations may exacerbate core magnetic saturation, leading to increased vibration. By analyzing the type, intensity, and action path of the interference factors, the causes of parameter changes can be more comprehensively understood.

[0031] After completing the analysis of the energy accumulation ability of the monitoring points and external interference factors, further research on the interaction effects between the monitoring points is needed. Power equipment is a complex coupled system, and there is mutual transmission of energy and signals between various functional sub-regions. For example, the rise in winding temperature will affect the core temperature through heat conduction, and core vibration may in turn exacerbate the aging of winding insulation. This interaction effect makes the parameter change of a single monitoring point possibly trigger a chain reaction. By establishing the correlation matrix between the monitoring points, the influence weight between different regions can be quantified, so as to more accurately describe the dynamic behavior of the whole equipment.

[0032] Based on the above analysis, the energy balance equation of the power equipment is constructed. The equation takes the energy input, output and internal loss of each monitoring point as the core, and describes the energy flow in the equipment operation through mathematical relationship. For example, the energy input of the winding region is electric power, the output is heat loss and heat conduction to the core, and the internal loss includes resistance heating and insulation medium loss; the energy input of the core region is magnetic energy, the output is vibration and heat, and the internal loss includes hysteresis and eddy current loss. By integrating these relationships into a system of equations, the energy distribution state of the equipment under different operating conditions can be dynamically simulated.

[0033] The solution of the energy balance equation depends on the real-time updated monitoring point data. During the operation of the equipment, the parameters of each monitoring point change continuously, and the equation needs to be dynamically adjusted according to the latest data. For example, when the load increases and the current rises, the electric power input of the winding region increases, and the equation needs to adjust the calculation weight of its heat loss and conduction heat accordingly; when the ambient temperature drops, the heat dissipation efficiency of the cooling system improves, and the equation needs to reevaluate its influence on the winding temperature. This dynamic adjustment mechanism enables the model to track the changes in the running state of the equipment in real time.

[0034] The output of the model not only includes the energy distribution of each functional sub-region, but also can further derive the evaluation indicators of the equipment health status. For example, by comparing the deviation of the actual energy loss from the theoretical value, the degree of insulation aging can be judged; by analyzing the frequency spectrum characteristics of the core vibration energy, mechanical looseness or magnetic saturation phenomenon can be identified. These indicators provide rich feature input for the subsequent artificial intelligence evaluation framework, enabling it to more comprehensively analyze the running state of the equipment.

[0035] The construction process of the dynamic state model fully considers the physical characteristics and operating rules of the power equipment, and realizes the fine description of the equipment state through multi-dimensional parameter acquisition, energy accumulation analysis, interference factor identification and interaction modeling. The dynamic adjustment capability of the model enables it to adapt to the changes of different operating conditions, providing a reliable data basis for artificial intelligence evaluation.

[0036] Example 2: see Figure 3The performance verification and integration process of the artificial intelligence evaluation framework is a key link in the method of power equipment operation state evaluation. The design of the framework needs to meet the requirements of real-time, accuracy and stability, and the performance verification needs to cover the whole process of data processing, model reasoning and result output, and form a collaborative optimization evaluation system through reasonable integration strategy and dynamic state model.

[0037] In the verification of the data processing stage, the framework's preprocessing ability for multi-dimensional operation parameters of power equipment needs to be tested. Power equipment monitoring data usually contains noise, missing values and outliers, and these data quality problems directly affect the reliability of subsequent analysis. The framework needs to have adaptive filtering function, which can dynamically adjust the filtering parameters according to the signal characteristics, such as using sliding window mean filtering for high-frequency noise in current signal and using differential compensation for slow drift of temperature signal. For missing data, the framework needs to support multiple completion strategies, including linear interpolation, pattern filling based on historical data and correlation completion of adjacent monitoring points. The abnormal value detection algorithm needs to dynamically adjust the threshold according to the equipment operation condition, to avoid false positives caused by normal working condition changes such as load mutation. By verifying the performance of different preprocessing methods under different data quality scenarios, the optimal combination of data cleaning and completion strategies can be determined.

[0038] The verification of the model reasoning stage focuses on the real-time and accuracy of artificial intelligence algorithms. Power equipment state evaluation usually uses deep neural network models, and the inference speed needs to meet the timeliness requirements of online monitoring. The framework needs to test the inference delay of different network structures on the target hardware platform, such as comparing the computational efficiency of convolutional neural network and lightweight Transformer architecture on edge computing devices. Model accuracy verification needs to build a test data set covering typical operating states, including normal working condition, early fault, intermediate fault and severe fault, etc. By analyzing the classification accuracy and confusion matrix of the model on different fault types, the sensitivity and specificity of the model to specific fault patterns can be identified. To address the overfitting problem of the model, regularization techniques and incremental learning strategies need to be used to ensure the model's generalization ability on new operating data.

[0039] The verification of the result output stage focuses on the stability and explainability of the evaluation results. Power equipment state evaluation needs to output clear health state levels or fault probabilities, and the volatility of these results needs to be controlled within a reasonable range. The framework needs to calculate the variance and trend consistency of consecutive evaluation results to avoid evaluation jumps caused by small changes in input data. To improve the credibility of the results, the framework should output auxiliary explanation information, such as the contribution ranking of key influencing parameters, comparison analysis with historical evaluation results, etc. For important fault warnings, a multi-level confirmation mechanism needs to be set up, which triggers an alarm only after detecting the same abnormal pattern multiple times, to reduce the risk of false positives.

[0040] After completing the phased performance verification, the artificial intelligence evaluation framework needs to be integrated into the dynamic state model. The core of the integration process is to establish a mapping relationship between the framework and the model parameter monitoring points. By analyzing the influence weight of each monitoring point in the dynamic state model, the key parameters that most determine the evaluation of the equipment state are identified. The principal component analysis method is used to calculate the variance contribution rate of each monitoring point parameter to determine the core dimension of the dominant equipment state change. For example, in the transformer state evaluation, the parameters of the top oil temperature, winding hot spot temperature and dissolved gas content in oil usually have a high weight. Align the input features of the artificial intelligence evaluation framework with these key monitoring points to ensure that the framework receives the most informative input data.

[0041] The output of the framework needs to form a closed-loop feedback with the prediction module of the dynamic state model. The artificial intelligence evaluation result is not only used as the basis for the final state judgment, but also needs to be fed back to the parameter adjustment mechanism of the model. For example, when the framework detects the trend of insulation aging, the coefficient of the insulation material thermal aging equation in the model can be dynamically adjusted; when the evaluation result shows that the cooling efficiency is decreasing, the related parameters of oil flow speed and heat dissipation capacity in the model can be corrected. This two-way interaction enables the dynamic state model to continuously optimize its parameters based on the evaluation results, forming a self-adaptive learning cycle.

[0042] The integrated system needs to be verified for its coordination, testing the data interaction efficiency and logical consistency of the artificial intelligence framework and the dynamic state model. Check if there is a conflict between the framework output and the model prediction, for example, whether the energy distribution state calculated by the model and the health level evaluated by the framework are logically consistent. For inconsistent cases, an arbitration mechanism needs to be established, such as using the Bayesian fusion method to integrate the two evaluation results. The system also needs to verify the rationality of resource allocation to ensure that the computational load of the artificial intelligence framework does not affect the real-time updating ability of the dynamic state model.

[0043] The performance optimization of the artificial intelligence evaluation framework is a continuous iterative process. With the accumulation of power equipment operation data, the model parameters and logical rules of the framework need to be updated regularly. The framework should support online learning function, which can automatically filter valuable new data samples and trigger model retraining. At the same time, a version control mechanism needs to be established to preserve the evaluation framework of historical versions as a benchmark for comparison, facilitating the analysis of the actual effect of algorithm improvement.

[0044] Example 3: refer to Figure 4 The calculation process of the evaluation data flow is based on the analysis of the energy consumption of the power equipment and the running demand of the artificial intelligence evaluation framework. This process determines the data processing scale required for the evaluation task by quantifying the relationship between the equipment operation energy consumption and the evaluation system's own consumption, providing accurate basis for the adjustment of the subsequent evaluation control unit. The calculation process needs to consider the dynamic change characteristics of the equipment operation characteristics, evaluation framework performance indicators and time dimension.

[0045] The total energy consumption of the power equipment in a specified time period is calculated from the energy input statistics of the parameter monitoring points. The monitoring points of each functional sub-region continuously record the energy conversion data during operation, such as the Joule heat loss represented by the product of current square and resistance in the winding area, the sum of hysteresis loss and eddy current loss in the core area, and the power consumption of the oil pump motor in the cooling system. After these data are collected at a fixed sampling frequency, their cumulative amount in the time dimension is calculated by numerical integration method. For discrete sampling systems, the trapezoidal rule is used for integration calculation to balance the precision and computational complexity: wherein, represents the total energy consumption, is the sampling time interval, represents the instantaneous power value of the i-th sampling point, and n is the total number of sampling points. This calculation needs to cover all key parameter monitoring points and couple the energy consumption of each sub-region according to the physical correlation. For example, the temperature rise caused by winding heating increases the cooling system load, and this interaction needs to be reflected in the total energy calculation.

[0046] The artificial intelligence evaluation framework also consumes energy during operation, which comes from the power supply system of the power equipment. The framework energy consumption mainly includes the computing power consumption of the data processing module, the operation energy consumption of the model inference unit, and the communication power consumption of the result output module. The energy consumption of the data processing module is linearly related to the input data volume, and the model inference energy consumption depends on the complexity of the neural network structure and the inference frequency. A real-time monitoring mechanism for framework energy consumption needs to be established to estimate the actual energy consumption by collecting indicators such as CPU / GPU instruction cycle number and cache hit rate through hardware performance counters, combined with chip-level power consumption models. The total energy consumption of the framework itself is obtained by time integrating the above energy components in a specified time period.

[0047] The theoretical evaluation demand reflects the actual net energy support required for the evaluation task, and its calculation needs to consider the allocation relationship between the total energy consumption of the equipment and the energy consumption of the framework itself. The remaining energy after deducting the basic energy consumption of the framework from the total energy consumption of the power equipment is the resource available for state evaluation tasks. This part of energy determines the maximum evaluation intensity that the system can support, including data processing depth, model inference frequency, and result output accuracy. The calculation of the theoretical evaluation demand needs to introduce an energy allocation weight coefficient, which is determined by the equipment operation priority strategy. For example, in heavy load conditions, the energy quota of the evaluation task may be appropriately reduced to ensure the safety of the equipment itself.

[0048] The calculation of the evaluation data flow converts the theoretical evaluation requirement into a specific data processing scale. The performance indicators of the artificial intelligence evaluation framework include a unit energy processing capacity parameter, which is obtained through offline testing and represents the standard amount of data that the framework can process per unit of energy consumed. In the voltage sag detection scenario, the unit energy processing capacity may be represented as the number of waveform sampling points that can be analyzed per joule of energy; in the temperature field reconstruction task, it may be represented as the number of grid nodes that can be processed per joule of energy. Multiplying the theoretical evaluation requirement by the unit energy processing capacity gives the upper limit of the evaluation data flow that the system can support within a specified time period.

[0049] The dynamic adjustment mechanism in the time dimension needs to be considered in the calculation process. The operating state of power equipment has time-varying characteristics, and its energy consumption often exhibits non-stationary characteristics. For example, when the transformer load fluctuates periodically, the winding loss will fluctuate with the change in current; when the ambient temperature changes day and night, the cooling system will also adjust the heat dissipation energy consumption accordingly. The calculation of evaluation data flow cannot simply use the average value, but needs to establish a sliding time window mechanism to dynamically update the calculation results at multiple time scales. Typical implementation methods include calculating the instantaneous data flow at a minute level to guide real-time control and using a hourly window to calculate the cumulative flow for resource planning.

[0050] The spatial distribution characteristics of the evaluation data flow also need to be analyzed. The data generated by different functional sub-regions of power equipment have different information density and value weights. For example, partial discharge monitoring data, although not large in data volume, is crucial for insulation state evaluation; while temperature distribution data is larger in volume, the change of a single measurement point may have limited impact. Under the constraint of total data flow, a data importance weighted distribution mechanism needs to be established to reserve sufficient processing resources for core monitoring parameters. This distribution can be achieved through a data value coefficient matrix, where the matrix elements represent the relative importance of each monitoring point data in state evaluation.

[0051] The verification of the calculation results uses the energy-data conservation principle. The total input energy of the system should be equal to the sum of the device's own energy consumption, the framework's running energy consumption, and the evaluation data processing energy consumption. By establishing an energy flow balance checking mechanism, errors that may accumulate or unreasonable distribution during the calculation of data flow can be found. When there is a significant imbalance, it is necessary to backtrack and check the accuracy of the energy consumption monitoring data, the rationality of the theoretical evaluation requirement, and the applicability of the unit energy processing capacity, etc.

[0052] The implementation establishes a quantitative bridge between the physical operation state of power equipment and the artificial intelligence evaluation digital system through strict energy metering and data conversion calculation. The calculation process fully considers the time dynamic characteristics and spatial distribution characteristics, so that the evaluation data flow can accurately reflect the actual system resource constraints, providing an objective basis for the formulation of evaluation control strategies. The entire calculation system forms a closed-loop verification mechanism, ensuring the rationality and reliability of the calculation results through the principle of energy conservation, supporting the long-term stable operation of the power equipment state evaluation system.

[0053] Example 4: Referring to Figure 4 The adjustment instruction generation and execution process of the evaluation control unit is based on the evaluation data flow calculated in Example 3, and realizes precise control through discrete time allocation and dynamic correction mechanism. Taking the transformer partial discharge online monitoring system as an example, this process shows how to adjust the working mode of the control unit according to the data flow demand.

[0054] In the transformer partial discharge monitoring scenario, the evaluation control unit is responsible for managing the working rhythm of the high-frequency signal acquisition card, data processing module and classification inference unit. Assuming that Example 3 calculates the total evaluation data flow in a 10-second time window to be 850MB, the control unit needs to reasonably allocate the activation time under the premise of ensuring data integrity. The system first tests the performance of the control unit in different working modes, and records the key parameters shown in Table 1.

[0055] Table 1: The system first tests the performance of the control unit in different working modes, and records the key parameters shown in the following table.

[0056] Based on the test data, the balanced mode is selected as the reference working mode, which achieves a good balance between data processing amount and quality. The 10-second evaluation period is divided into 20 time units, each lasting 500ms. According to the total data flow of 850MB, it is theoretically required to activate the balanced mode 28.3 times (850MB / 30MB per time), and the total activation time needs to be 2264ms. Due to the time window limit, the uniform distribution strategy is adopted, and 113.2ms of activation time is allocated in each 500ms unit.

[0057] The system performs the first round of adjustment, and the control unit is activated for 113 ms in the first time unit (0-500 ms). Actual monitoring found that there was a slight disturbance in the power grid during this period, resulting in increased signal noise, and the actual processing data volume was only 27 MB. The system starts the dynamic correction mechanism according to the deviation: the remaining data volume to be processed is 823 MB (850-27), and the remaining time is 9500 ms. The single activation duration of the subsequent unit is adjusted to 116 ms (823 MB / (30 MB / 80 ms)), and this correction value retains 3% redundancy to deal with possible interference.

[0058] During the execution process of the subsequent time units, the system continuously monitors two key indicators: the actual processing data volume in the unit and the signal quality score. When the signal quality significantly improves (reaches 97%) in the fifth time unit (2000-2500 ms), the system appropriately reduces the activation duration to 110 ms while keeping the data processing volume stable. Conversely, when the 12th time unit (5500-6000 ms) encounters strong electromagnetic interference, the system temporarily switches to high-speed mode, increasing the sampling rate while maintaining the original activation duration, to ensure complete capture of key discharge pulses.

[0059] The parameter adjustment of the control unit not only involves time allocation, but also includes the optimization of the combination of working modes. During the night period when the transformer load is low, the system may adopt an alternating strategy of energy-saving mode and balanced mode: using energy-saving mode to process basic monitoring data for 3 consecutive time units, and switching to balanced mode for fine analysis in the fourth unit. This combination not only maintains the evaluation coverage, but also reduces the overall power consumption by about 15%.

[0060] The abnormal handling mechanism during adjustment is particularly important. When the actual processing data volume of two consecutive time units is less than 85% of the expected value, the system automatically triggers the diagnostic process: first, check the impedance matching of the signal acquisition channel, second, verify the adaptability of the data processing algorithm, and finally evaluate the power grid working condition changes. According to the diagnostic results, three countermeasures may be taken: adjusting the cutoff frequency of the pre-filter, switching to a backup data processing thread, or temporarily increasing the activation duration by 10% for compensation.

[0061] When performing the final verification, the system summarizes the cumulative processing data volume of each time unit and compares it with the target value of 850 MB. Assuming that 842 MB is actually completed, the deviation rate is 0.94%, which is within the allowed 2% error range. The system records the parameters of this adjustment as historical benchmarks, including average activation duration, mode switching frequency, dynamic correction amplitude, etc., which are used to optimize the initial allocation strategy of the subsequent evaluation period.

[0062] Through specific cases of transformer partial discharge monitoring, the complete process of evaluation control unit adjustment is demonstrated. From the selection of working mode, the division of time unit to the execution of dynamic correction, the system always carries out fine control around the data flow target. The test data in the table provides objective basis for initial decision, while the real-time monitoring and feedback mechanism ensures that the adjustment process adapts to changes in actual operating conditions. The entire implementation process embodies the core concept of adaptive control: under the constraint of limited resources, the best evaluation effect is achieved through continuous optimization.

[0063] Example 5: Quantitative analysis process of evaluation effect By comparing the parameter changes of power equipment before and after state evaluation with the theoretical expectations, the actual effectiveness of the artificial intelligence evaluation framework is verified. Taking the insulation state evaluation of power transformers as an example, this process elaborates how to achieve objective measurement of evaluation effect through parameter difference analysis, demand deviation calculation and weight correction, etc.

[0064] In the baseline state establishment stage before evaluation, the system records the initial values of the key parameters of the transformer, including winding direct current resistance, dielectric loss tangent value, and oil dissolved gas content, etc. These parameters constitute the original feature set of the equipment state, and each parameter is assigned an initial monitoring weight according to its physical meaning and importance in health evaluation. For example, winding direct current resistance reflects the integrity of the conductor, and its change may indicate joint loosening or conductor fracture; dielectric loss tangent value represents the aging degree of insulating materials; the composition and content of dissolved gases in oil are key indicators for diagnosing internal latent faults.

[0065] After the evaluation process starts, the artificial intelligence evaluation framework conducts multi-dimensional analysis on the running state of the transformer, and outputs the insulation state grade and potential risk prediction. The system continuously monitors the dynamic changes of each parameter during the evaluation process and records the final value of the parameter at the end of the evaluation period. By comparing the differences between the initial and final values of each parameter, the relative change rate is calculated. For example, the dielectric loss tangent value rises from 0.005 to 0.006, with a change rate of 20%; the ethyne gas content in oil increases from 3 μL / L to 4 μL / L, with a change rate of 33%. These change amounts reflect the actual evolution of the equipment state during the evaluation process.

[0066] Based on the parameter change amount, the actual evaluation demand is deduced, which needs to establish the mapping relationship between parameter change and energy consumption. The change degree of each parameter corresponds to a specific evaluation resource demand, such as gas chromatography detection energy consumption for oil gas analysis and energy consumption for applying test voltage for dielectric loss measurement. The system converts the change amount of each parameter into equivalent evaluation energy value according to the preset parameter-energy conversion coefficient, and summarizes the actual evaluation total demand. This value reflects the actual resource scale consumed by the evaluation framework to capture the state change of the equipment, and forms a comparable benchmark with the theoretical evaluation demand calculated in Example 3.

[0067] The demand deviation analysis is the core step of comparing the actual evaluation demand with the theoretical evaluation demand. The system calculates the absolute difference and the relative deviation rate of the two, and analyzes the source distribution of the deviation. The deviation may come from multiple aspects: the theoretical model's insufficient estimation of the sensitivity of certain parameters, such as underestimating the energy consumption of trace gas detection in oil; unexpected changes in the evaluation process leading to additional resource consumption, such as temporary increase in partial discharge monitoring demand; or fluctuations in the efficiency of the evaluation framework's algorithm. By decomposing the contribution of each parameter to the total deviation, the main influencing factors are identified, providing directional guidance for control weight adjustment.

[0068] The control weight correction of the evaluation control unit is the key output of quantitative analysis. According to the results of deviation analysis, the system dynamically adjusts the decision weight of each parameter in the evaluation process. For parameter categories whose actual demand is consistently higher than the theoretical expectation, appropriately increase their control weight and increase resource allocation; for parameters whose actual demand is lower than the theoretical expectation, appropriately reduce the weight. The weight adjustment adopts a gradual strategy, with the correction amplitude controlled within 5%-10% of the original value, avoiding drastic fluctuations affecting the stability of the evaluation. The adjusted weight value will be used for the calculation of the theoretical demand in the next evaluation period, forming a closed-loop optimization mechanism.

[0069] Long-term tracking of evaluation effectiveness requires the establishment of a historical data comparison mechanism. The system saves the parameter change pattern, demand deviation characteristics and weight adjustment record of each evaluation, forming an evaluation efficiency evolution curve. By analyzing the convergence trend of the deviation rate in multiple evaluation cycles, it is determined whether the system is gradually approaching the optimal evaluation state. At the same time, monitor the frequency and amplitude changes of weight adjustment to verify the adaptive learning ability of the evaluation framework. Historical data is also used to identify seasonal patterns or load correlations in the evolution of device state, providing reference for periodic optimization of evaluation strategies.

[0070] Through the specific application of transformer insulation state evaluation, the complete process of evaluation effect quantitative analysis is demonstrated. From parameter change capture to demand deviation calculation, and then to control weight dynamic adjustment, the system realizes the objective evaluation and continuous improvement of the artificial intelligence evaluation framework. The whole process emphasizes data-driven analysis logic, avoiding the interference of subjective judgment, ensuring the scientificity and reliability of evaluation optimization. The quantitative analysis results not only verify the effectiveness of the current evaluation system, but also provide a clear improvement direction for subsequent algorithm upgrade and model iteration, forming a positive cycle of continuous improvement of power equipment state evaluation capability.

[0071] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and illustrative figures, it should be apparent that the scope of the present application is not limited to these specific embodiments.

[0072] While the embodiments of the application have been shown and described herein, it is to be understood that the scope of the application, jointly pointed out in the appended claims, is not to be limited to the above-described embodiments but can be otherwise variously changed, modified, replaced, and altered within the principles and spirit of the present application.

Claims

1. A power equipment operation state evaluation method using artificial intelligence, characterized by, The method comprises the following steps: Collecting multi-dimensional operation parameters of the power equipment during operation, and constructing a dynamic state model of the power equipment based on the multi-dimensional operation parameters; Setting up an artificial intelligence evaluation framework, and verifying the performance of the artificial intelligence evaluation framework to obtain evaluation capability indicators of the artificial intelligence evaluation framework; Integrating the artificial intelligence evaluation framework into the dynamic state model; Analyzing the total energy consumption of the power equipment within a specified time period, and calculating the theoretical evaluation demand of the artificial intelligence evaluation framework within the specified time period, combining the evaluation capability indicators of the artificial intelligence evaluation framework to calculate the evaluation data flow of the artificial intelligence evaluation framework within the specified time period; Generating adjustment instructions of an evaluation control unit in the artificial intelligence evaluation framework based on the evaluation data flow of the artificial intelligence evaluation framework within the specified time period; Performing parameter adjustment operations on the evaluation control unit based on the adjustment instructions of the evaluation control unit; Quantitatively analyzing the evaluation effect of the artificial intelligence evaluation framework on the operation state of the power equipment; The step of setting up an artificial intelligence evaluation framework, and verifying the performance of the artificial intelligence evaluation framework to obtain evaluation capability indicators of the artificial intelligence evaluation framework comprises the following steps: Respectively verifying the evaluation performance of the artificial intelligence evaluation framework in the data processing stage, the model inference stage and the result output stage to obtain the evaluation capability indicators of the artificial intelligence evaluation framework; In the verification in the data processing stage, the pre-processing capability of the framework on the multi-dimensional operation parameters of the power equipment needs to be tested; the real-time performance and accuracy of the artificial intelligence algorithm are the focus of the verification in the model inference stage; the stability and interpretability of the evaluation results are concerned in the verification in the result output stage.

2. The power equipment operation state evaluation method using artificial intelligence according to claim 1, characterized by, The step of collecting multi-dimensional operation parameters of the power equipment during operation, and constructing a dynamic state model of the power equipment based on the multi-dimensional operation parameters comprises the following steps: Obtaining physical structure information of the power equipment, and dividing the power equipment into a plurality of functional sub-regions; Extracting parameter monitoring points of each functional sub-region to form a multi-dimensional operation parameter set of the power equipment; Calculating the energy accumulation capability of each parameter monitoring point, and identifying external interference factors acting on the parameter monitoring points; Based on the energy accumulation capability of each parameter monitoring point and the external interference factors, an energy balance equation of the power equipment is established.

3. The power equipment operation state evaluation method using artificial intelligence according to claim 2, characterized by, The step of establishing an energy balance equation of the power equipment based on the energy accumulation capability of each parameter monitoring point and the external interference factors comprises the following steps: Respectively extracting the interaction effect between each parameter monitoring point and the influence degree of the external interference factors on each parameter monitoring point, and based on the energy accumulation capability of each parameter monitoring point, the interaction effect between each parameter monitoring point and the influence degree of the external interference factors on each parameter monitoring point, constructing an energy balance equation of the power equipment.

4. The power equipment operation state evaluation method using artificial intelligence according to claim 1, characterized by, The step of integrating the artificial intelligence evaluation framework into the dynamic state model comprises the following steps: Analyzing the parameter monitoring points in the dynamic state model, identifying the parameter monitoring point with the greatest influence, and setting the artificial intelligence evaluation framework at the corresponding position in the dynamic state model mapped from the parameter monitoring point with the greatest influence. 5.The power equipment operation state evaluation method using artificial intelligence according to claim 1, characterized in that, The analysis power equipment total energy consumption in a specified period, and calculates the theoretical evaluation demand of the artificial intelligence evaluation framework in a specified period, combined with the evaluation ability index of the artificial intelligence evaluation framework, calculates the evaluation data flow of the artificial intelligence evaluation framework in a specified period, including: Calculate and aggregate the energy input of each parameter monitoring point in a specified period, and obtain the total energy consumption of the power equipment in a specified period; Obtain the energy consumed by the artificial intelligence evaluation framework itself in a specified period, and combine the total energy consumption of the power equipment in a specified period to obtain the theoretical evaluation demand of the artificial intelligence evaluation framework in a specified period; Combine the theoretical evaluation demand of the artificial intelligence evaluation framework in a specified period with the evaluation ability index of the artificial intelligence evaluation framework to obtain the evaluation data flow of the artificial intelligence evaluation framework in a specified period. 6.The power equipment operation state evaluation method using artificial intelligence according to claim 1, characterized in that, Based on the evaluation data flow of the artificial intelligence evaluation framework in a specified period, the adjustment instruction of the evaluation control unit in the artificial intelligence evaluation framework is generated, including: Test the unit data flow processing capability of the evaluation control unit in the activated state, and combine the evaluation data flow of the artificial intelligence evaluation framework in a specified period to determine the activation time length of the evaluation control unit.

7. The power equipment operation state evaluation method using artificial intelligence according to claim 6, characterized by, Based on the adjustment instruction of the evaluation control unit, the parameter adjustment operation of the evaluation control unit includes: Divide the specified period into multiple discrete time period units, and evenly distribute the activation time length of the evaluation control unit to multiple discrete time period units; After the end of the previous discrete time period unit, judge whether the data flow processing of the evaluation control unit in the discrete time period unit reaches the expected target; Adaptively correct the activation time length of the evaluation control unit in the next discrete time period unit until the parameter adjustment operation of the evaluation control unit in a specified period is completed. 8.The power equipment operation state evaluation method using artificial intelligence according to claim 1, wherein, The quantitative analysis of the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment includes: Obtain the parameter change amount of the power equipment before and after state evaluation; Based on the parameter change amount, obtain the actual evaluation demand of the power equipment; Compare and analyze the actual evaluation demand and the theoretical evaluation demand to obtain the control weight of the evaluation control unit in the artificial intelligence evaluation framework; Based on the control weight, the evaluation effect of the artificial intelligence evaluation framework on the running state of the power equipment is quantitatively analyzed.

9. An electric power equipment operation state evaluation system using artificial intelligence, characterized by, The system for implementing the artificial intelligence-based power equipment running state evaluation method as claimed in any one of claims 1 to 8, the system comprising: A dynamic state model construction module, the dynamic state model construction module is used for collecting multi-dimensional running parameters of power equipment in the running process, and constructing a dynamic state model of the power equipment based on the multi-dimensional running parameters; An artificial intelligence evaluation framework construction module, the artificial intelligence evaluation framework construction module is used for setting an artificial intelligence evaluation framework, and verifying the performance of the artificial intelligence evaluation framework to obtain the evaluation ability index of the artificial intelligence evaluation framework; a framework integration module, configured to integrate the artificial intelligence evaluation framework into the dynamic state model; an evaluation data flow calculation module, configured to analyze total energy consumption of the power equipment in a specified time period, and to calculate theoretical evaluation requirements of the artificial intelligence evaluation framework in the specified time period, and to calculate evaluation data flow of the artificial intelligence evaluation framework in the specified time period in combination with evaluation capability indexes of the artificial intelligence evaluation framework; an adjustment instruction generation module, configured to generate adjustment instructions of an evaluation control unit in the artificial intelligence evaluation framework based on the evaluation data flow of the artificial intelligence evaluation framework in the specified time period; a parameter adjustment control module, configured to perform parameter adjustment operation on the evaluation control unit based on the adjustment instructions of the evaluation control unit; an evaluation effect verification module, configured to quantitatively analyze evaluation effects of the artificial intelligence evaluation framework on the running state of the power equipment.

Images