A method and system for predicting faults of a high-voltage power supply of an electric precipitator of a thermal power plant

Abstract

The application relates to the technical field of fault prediction, and provides a fault prediction method and system for a high-voltage power supply of an electric dust removal device of a thermal power plant, which comprises the following steps: obtaining inherent electrical pulsation information generated by a high-voltage power supply under a specific operating state, and establishing a reference mode of the inherent electrical pulsation information; continuously obtaining real-time electrical pulsation information of the high-voltage power supply; comparing the real-time electrical pulsation information with the reference mode to identify interference information introduced by a data acquisition link; determining an interference mode of the data acquisition link according to the interference information; processing operating information transmitted by the data acquisition link according to the interference mode to obtain pure operating information; and identifying early fault information of the high-voltage power supply body based on the pure operating information. The application can effectively distinguish the interference information introduced by the data acquisition link from the early fault information of the power supply body, thereby improving the accuracy of fault prediction, avoiding false positives, and ensuring the timely discovery of real faults.

Description

Technical Field

[0001] This application relates to the field of fault prediction technology, and more specifically, to a method and system for predicting faults in the high-voltage power supply of electrostatic precipitators in thermal power plants. Background Technology

[0002] In modern thermal power plants, electrostatic precipitators and their high-voltage power supply equipment are crucial for meeting environmental emission standards. These high-voltage power supply devices operate for extended periods in harsh environments characterized by high temperatures, high dust levels, high humidity, and mechanical vibration, leading to continuous wear and aging of internal insulation materials, electronic components, and data acquisition parts. Traditional maintenance methods struggle to detect problems in their early stages, easily resulting in sudden equipment shutdowns, impacting environmental compliance, and increasing operating costs. To address this issue, power plants have introduced fault prediction systems based on operational data analysis, aiming to identify potential problems in advance.

[0003] However, in practical applications, power plants have encountered a complex problem: highly sensitive fault prediction methods struggle to distinguish between genuine, weak fault characteristic signals caused by early performance degradation of internal components of the power supply and interference signals introduced by data acquisition link deterioration, which are highly similar in characteristics, when processing sensor data with weak noise or drift. For example, intermittent noise on the data path may resemble electromagnetic radiation characteristics generated by partial discharge in certain frequency components, or be mathematically indistinguishable from the minute waveform distortion caused by increased reverse leakage current in the silicon stack. This confusion leads to frequent warnings or anomaly reports from the prediction system, but on-site investigations often fail to find obvious anomalies, significantly consuming maintenance resources and reducing the maintenance team's trust in the system. Summary of the Invention

[0004] This application provides a method and system for predicting high-voltage power supply failures in electrostatic precipitators for thermal power plants, in order to solve at least one of the above-mentioned technical problems.

[0005] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application discloses a method for predicting high-voltage power supply failures in electrostatic precipitators for thermal power plants, including: Acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions, and establish a reference pattern for this inherent electrical pulsation information; Continuously acquire real-time electrical pulsation information of the high-voltage power supply; compare the real-time electrical pulsation information with the reference mode to identify interference information introduced by the data acquisition link; Based on this interference information, the interference mode of the data acquisition link is determined; Based on this interference pattern, the operational information transmitted through the data acquisition link is processed to obtain clean operational information. Based on this pure operational information, early fault information of the high-voltage power supply body can be identified.

[0006] Secondly, this application also discloses a fault prediction system for high-voltage power supply of electrostatic precipitators in thermal power plants, comprising: The information acquisition module is used to acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions, and to establish a reference pattern for the inherent electrical pulsation information. The real-time monitoring module is used to continuously acquire real-time electrical pulsation information of the high-voltage power supply; The comparison and identification module is used to compare the real-time electrical pulsation information with the reference mode to identify interference information introduced by the data acquisition link; The mode determination module is used to determine the interference mode of the data acquisition link based on the interference information. The information processing module is used to process the operational information transmitted through the data acquisition link according to the interference mode in order to obtain clean operational information. The fault identification module is used to identify early fault information of the high-voltage power supply body based on the clean operating information.

[0007] Compared with the prior art, this application has at least the following beneficial effects: This application can significantly improve the accuracy of fault prediction, avoid the waste of operation and maintenance resources and the reduction of trust in the system caused by false alarms, and at the same time ensure that true early fault signals can be detected and handled in a timely manner, thereby effectively improving the operational reliability and environmental compliance of high-voltage power supply equipment for electrostatic precipitators in thermal power plants. Attached Figure Description

[0008] Figure 1 A flowchart illustrating a method for predicting high-voltage power supply failures in electrostatic precipitators for thermal power plants, provided in this application; Figure 2 This application provides a structural schematic diagram of a high-voltage power supply fault prediction system for electrostatic precipitators in thermal power plants. Detailed Implementation

[0009] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0010] like Figure 1 As shown in the embodiment of this application, a method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants is proposed, including: Acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions, and establish a reference pattern for the inherent electrical pulsation information; Continuously acquire real-time electrical pulsation information of the high-voltage power supply; Real-time electrical pulsation information is compared with a reference pattern to identify interference introduced by the data acquisition link; Based on the interference information, determine the interference mode of the data acquisition link; Based on the interference pattern, the operational information transmitted through the data acquisition link is processed to obtain clean operational information. Based on clean operational information, early fault information of the high-voltage power supply body can be identified.

[0011] In this application, the high-voltage power supply specifically refers to a DC high-voltage power supply used in electrostatic precipitator systems. Its main function is to provide a high-voltage electrostatic field for the electrostatic precipitator to adsorb dust particles in the flue gas. This high-voltage power supply typically includes core components such as a transformer, rectifier, and control unit, and its operating status directly affects the efficiency of the electrostatic precipitator.

[0012] Electrical ripple information refers to the periodic or non-periodic fluctuations in electrical quantities such as voltage and current generated during the operation of a high-voltage power supply. This ripple information includes various aspects such as the operating status of the power supply itself, the health status of internal components, and external interference. For example, a normal power supply output may have inherent ripple, while internal insulation degradation or partial discharge may generate specific high-frequency pulse signals.

[0013] The data acquisition link refers to the data transmission path from the sensors of the high-voltage power supply to the fault prediction system, including sensors, signal conditioning circuits, analog-to-digital converters, communication cables, and data transmission protocols. The performance of this link directly affects the quality and accuracy of the acquired electrical pulsation information.

[0014] Inherent electrical ripple information refers to the stable and repeatable electrical ripple characteristics of a high-voltage power supply under normal and healthy operating conditions, determined by its own design and operating principles. These characteristics can serve as a benchmark for assessing the health of the power supply and identifying external interference.

[0015] A reference pattern is a mathematical model or set of features built upon inherent electrical ripple information, used to characterize the normal electrical behavior of a high-voltage power supply under specific operating conditions. This pattern can be a statistical model, spectral characteristics, time-domain waveform template, etc.

[0016] Interference refers to signals not generated by the power supply itself, which are introduced by the degradation of the data acquisition link itself (such as poor contact, electromagnetic interference, insulation aging, etc.) and superimposed on the real electrical pulsation signal. This interference may be similar to early fault signals in some characteristics, thus leading to misjudgment.

[0017] The implementation environment of this application is typically the central control room or equipment monitoring center of a thermal power plant, which is equipped with a data acquisition system, a data processing server, and a fault prediction software platform. Sensors convert the electrical signals from the high-voltage power supply into electrical signals, which are then transmitted to the server for processing and analysis via a data acquisition link.

[0018] The following is a further detailed description of the high-voltage power supply fault prediction method for electrostatic precipitators in thermal power plants proposed in this application: First, the method involves acquiring the inherent electrical ripple information generated by the high-voltage power supply under specific operating conditions and establishing a reference pattern for this inherent electrical ripple information. In practice, this can be achieved by long-term operational monitoring of the power supply equipment after it has been put into operation or after a major overhaul, ensuring it is in a healthy state. For example, a high-precision oscilloscope or data acquisition instrument can be used to collect voltage and current waveform data at key points such as the power supply output, rectifier bridge arm, and control circuit. After processing these data using signal processing techniques such as Fourier transform and wavelet analysis, features such as energy distribution, harmonic content, and ripple coefficient in different frequency ranges can be extracted. These features are considered the fingerprint of the power supply in a healthy state and can serve as inherent electrical ripple information. Subsequently, this feature data can be used to establish a reference pattern. For example, the mean and standard deviation of these features can be calculated, or a cluster center in a multi-dimensional feature space can be constructed, or even a normal pattern recognition model based on a neural network can be trained.

[0019] Secondly, this method continuously acquires real-time electrical pulsation information of the high-voltage power supply. This is typically achieved by installing sensors (such as Hall current sensors, high-voltage probes, etc.) at key monitoring points of the power supply. These sensors convert analog electrical signals into electrical signals, which are then converted from analog to digital and preliminarily processed by a data acquisition module. For example, instantaneous values ​​of voltage and current can be acquired every few seconds or minutes and digitized. These real-time data streams are continuously transmitted to the data processing unit for subsequent analysis.

[0020] Next, the real-time electrical pulsation information is compared with a reference model to identify interference introduced by the data acquisition link. In this step, the real-time electrical pulsation information is compared with a pre-established reference model. For example, the spectral characteristics of the real-time electrical pulsation information can be compared with the spectral characteristics of the reference model, and their Euclidean distance, correlation coefficient, or similarity can be calculated. If there is a significant difference between the real-time information and the reference model, it indicates a possible anomaly. This difference could be caused by a power supply failure or by interference introduced by the data acquisition link. To distinguish between these two cases, the characteristics of the difference can be further analyzed. For example, if the difference is mainly reflected in high-frequency random noise components and does not have the typical characteristics of a power supply failure, it is more likely to be interference introduced by the data acquisition link.

[0021] Then, based on the interference information, the interference mode of the data acquisition link is determined. Once the interference information introduced by the data acquisition link is identified, it needs to be analyzed in depth to determine its specific interference mode. For example, by performing spectral analysis on the interference information, it can be identified whether there is periodic noise at a specific frequency, which may indicate the presence of electromagnetic interference. Alternatively, by analyzing the time-domain characteristics of the interference information, it can be identified whether there are intermittent spikes or drifts, which may be related to poor sensor contact or line aging. These analytical results can help to build an interference mode library, which contains different types of interference and their corresponding characteristics. For example, high-frequency random noise mode, power frequency harmonic interference mode, intermittent pulse mode, etc., can be defined.

[0022] Subsequently, based on the interference pattern, the operational information transmitted through the data acquisition link is processed to obtain clean operational information. Once the interference pattern of the data acquisition link is determined, appropriate signal processing techniques can be used to eliminate or suppress this interference. For example, if a power frequency harmonic interference pattern is identified, a notch filter can be used to filter out harmonic components of specific frequencies. If a high-frequency random noise pattern is identified, a low-pass filter or wavelet denoising algorithm can be used to smooth the signal. Through these processes, clean operational information that more closely approximates the true state of the power supply can be extracted from the original, interfered operational information.

[0023] Finally, based on the clean operating information, early fault information of the high-voltage power supply body is identified. After obtaining clean operating information, this information can be used to accurately identify early faults of the high-voltage power supply body. For example, features can be extracted from clean voltage and current waveforms, such as partial discharge pulse counts, insulation resistance change trends, and increases in silicon stack reverse leakage current. These features can be compared with preset fault thresholds or fault mode libraries. When one or more features exceed the normal range and match known early fault modes, a fault warning can be issued. Since the operating information at this time has eliminated interference from the data acquisition link, the accuracy of fault identification will be greatly improved, avoiding false alarms caused by interference.

[0024] This application effectively solves the problem of distinguishing between real faults and data link interference in traditional fault prediction methods by introducing a mechanism for identifying and processing interference information in the data acquisition link. The method first establishes an inherent electrical pulsation reference mode for the high-voltage power supply as a benchmark for healthy operation. Then, it continuously monitors real-time electrical pulsation information and compares it with the reference mode, thereby identifying interference information introduced by the data acquisition link. Through pattern analysis of this interference information, the specific interference type and characteristics can be determined, and the original operating information can be processed in a targeted manner to remove interference and obtain clean operating information. Finally, based on this clean operating information, the system can accurately identify early fault information of the high-voltage power supply itself.

[0025] In some embodiments, the above-described method for predicting high-voltage power supply failures in electrostatic precipitators for thermal power plants further includes: The real-time electrical pulsation information is compared with the reference mode, and the dynamic transmission characteristics of the data acquisition link are evaluated in real time based on the comparison results. Based on the dynamic transmission characteristics, the operational information transmitted through the data acquisition link is reverse-engineered to obtain clean operational information.

[0026] Specifically, comparing real-time electrical pulsation information with a reference pattern aims to capture the impact of the data acquisition link on the electrical pulsation signal at the current moment. This comparison reflects the link's dynamic characteristics such as instantaneous response, attenuation, and phase shift. For example, the dynamic transmission characteristics of the link can be quantified by calculating indicators such as the correlation coefficient, spectral difference, or time-domain waveform deviation between the real-time electrical pulsation information and the reference pattern. Real-time evaluation of the dynamic transmission characteristics of the data acquisition link can be understood as quantifying the degree and manner of its impact on signal transmission under specific operating conditions. This includes, but is not limited to, the dynamic changes in parameters such as link gain, bandwidth, noise level, nonlinear distortion, and latency. In practical applications, reverse recovery processing of the operational information transmitted by the data acquisition link based on dynamic transmission characteristics refers to constructing a reverse model or compensation algorithm using the evaluated dynamic transmission characteristics to eliminate or reduce the distortion and interference caused by the data acquisition link to the operational information. For example, adaptive filters, deconvolution algorithms, or model-based prediction compensation techniques can be used to restore the operational information affected by the link to a near-original, undisturbed state, thereby obtaining clean operational information.

[0027] This application dynamically captures changes in the transmission characteristics of the data acquisition link by comparing real-time electrical pulsation information of the high-voltage power supply with a pre-established reference pattern. This comparison directly reflects the link's real-time impact on the signal, making it possible to evaluate the link's dynamic transmission characteristics. Once the link's dynamic transmission characteristics are quantified in real time, an accurate reverse recovery model can be constructed. This model can specifically compensate and correct the operational information transmitted by the data acquisition link, effectively offsetting the dynamic distortion and interference introduced by the link, thus ensuring that the obtained operational information is highly pure. This real-time evaluation and reverse recovery mechanism enables the system to adapt to instantaneous changes in the data acquisition link's transmission characteristics, avoiding fault prediction errors caused by the link's dynamics.

[0028] In some embodiments, the step of determining the interference mode of the data acquisition link based on interference information includes: Time-domain and frequency-domain characteristic analysis of interference information is performed to obtain the performance characteristics of interference information in different time scales and frequency ranges; The performance characteristics are compared with a preset degradation mode feature library to identify single or compound degradation modes in the data acquisition link. Based on the comparison results, the impact of each degradation mode in a single or combined degradation mode on the real-time electrical pulsation information of the high-voltage power supply is quantified in order to determine the interference mode of the data acquisition link.

[0029] The analysis of interference information in the time and frequency domains involves using signal processing techniques, such as Fourier transform, wavelet analysis, and Hilbert-Huang transform, to convert the acquired interference signal from the time domain to the frequency domain, or to analyze its transient and statistical characteristics within the time domain. The aim is to comprehensively reveal the intrinsic structure and dynamic behavior of the interference signal, such as its frequency components, energy distribution, periodicity, instantaneous amplitude changes, and phase information, thereby providing a rich and detailed data foundation for subsequent interference pattern recognition.

[0030] Performance characteristics can be understood as key parameters or patterns extracted from time-domain and frequency-domain analysis, such as energy peaks at specific frequencies, harmonic content, transient impact characteristics, signal envelope changes, and power spectral density. These performance characteristics are compared with a pre-defined degradation mode feature library. The aim is to use pattern matching algorithms, such as machine learning classifiers, similarity metrics, and neural networks, to associate currently identified interference characteristics with known feature patterns related to data acquisition link degradation. The degradation mode feature library stores the unique time-domain and frequency-domain manifestations of various typical data acquisition link faults (such as poor contact, insulation aging, cable damage, sensor drift, and electromagnetic interference) in electrical pulsation information. Through comparison, single or combined degradation modes that may exist in the current data acquisition link can be identified.

[0031] Quantifying the impact of each degradation mode on the real-time electrical pulsation information of the high-voltage power supply based on comparison results refers to further assessing the degree of contamination or distortion of the normal operation signal of the high-voltage power supply after identifying specific degradation modes. This can be achieved by establishing a mapping relationship between degradation modes and signal distortion, for example, through statistical methods, expert system rules, or simulations based on physical models. The goal is not only to identify what kind of interference it is, but also to quantify its impact, thereby accurately determining the interference mode of the data acquisition link and providing an accurate basis for subsequent recovery of clean operating information.

[0032] This application, through in-depth time and frequency domain analysis of interference information, can capture the essential characteristics of interference signals from multiple dimensions, avoiding misjudgments or information omissions that may occur with single-dimensional analysis. It is precisely this comprehensive feature extraction that enables subsequent comparison with the degradation mode feature library to more accurately identify the specific degradation modes present in the data acquisition link. Based on this, by quantifying the impact of each degradation mode on real-time electrical pulsation information, this scheme can establish a refined description of interference modes, thus laying a solid foundation for subsequent precise processing of the operational information transmitted through the data acquisition link to obtain clean operational information.

[0033] In some embodiments, the step of performing time-domain and frequency-domain characteristic analysis on the interference information to obtain the performance characteristics of the interference information in different time scales and frequency ranges includes: The steps for performing time-domain and frequency-domain characteristic analysis on interference information to obtain its performance characteristics at different time scales and frequency ranges include: Continuously monitor the operating status of the high-voltage power supply and acquire the characteristics of the inherent electrical pulsation signals generated by the power supply body under different operating conditions; Based on the operating conditions of the high-voltage power supply, adjust the parameters for time-domain and frequency-domain analysis of the interference information; Based on the adjusted parameters, time-domain and frequency-domain analyses are performed on the interference information to extract its performance characteristics within the corresponding time scale and frequency range. The characteristics of the extracted interference information are compared with the characteristics of the inherent electrical pulsation signal and the characteristics of the early fault signal of the high-voltage power supply, respectively, in order to identify the first difference between the interference information and the inherent electrical pulsation signal, and the second difference between the interference information and the early fault signal. Based on the first and second differences, a set of discriminative features is generated.

[0034] The continuous monitoring of the high-voltage power supply's operating status and the acquisition of its inherent electrical pulsation signals under different operating conditions involve continuously collecting operating parameters such as voltage, current, and power, and dynamically identifying the specific operating condition of the current high-voltage power supply by combining historical data or preset models. Based on this, the characteristics of the power supply's inherent electrical pulsation signals, matching the current operating condition, are retrieved or calculated. These characteristics represent the non-fault-related, periodic or random electrical fluctuation patterns generated by the high-voltage power supply's internal structure and operating principle under normal operating conditions. The purpose is to provide an accurate, condition-adaptive benchmark for subsequent interference information analysis, avoiding misjudging electrical fluctuations under normal operating conditions as interference or faults.

[0035] Adjusting the parameters for time-domain and frequency-domain interference analysis based on the operating conditions of the high-voltage power supply refers to dynamically selecting or optimizing the parameters of the time-domain and frequency-domain analysis algorithms used to analyze interference information based on the currently identified operating conditions of the high-voltage power supply. For example, the electrical characteristics of the high-voltage power supply and the frequency distribution of potential interference sources may differ under different loads or output voltages. Therefore, parameters such as the Fourier transform window size, sampling frequency, filter type, or wavelet basis function can be adjusted to ensure that the analysis tool can best capture and characterize the interference features under the current operating conditions. The aim is to improve the accuracy and relevance of interference information analysis.

[0036] Based on the adjusted parameters, time-domain and frequency-domain analyses are performed on the interference information to extract its characteristics within the corresponding time scale and frequency range. This involves using adaptively adjusted analysis parameters to perform in-depth time-domain and frequency-domain processing on the real-time acquired interference information. Time-domain analysis (such as statistical characteristics and waveform analysis) can reveal the transient characteristics and temporal distribution patterns of the interference signal; frequency-domain analysis (such as spectrum analysis and power spectral density) can reveal the frequency components and energy distribution of the interference signal. These analyses aim to comprehensively and meticulously extract the unique characteristics of the interference information across different time scales and frequency ranges, providing rich data support for subsequent interference identification and pattern determination.

[0037] The extracted interference information's characteristics are compared with the characteristics of inherent electrical pulsation signals and early fault signals of the high-voltage power supply to identify the first difference between the interference information and the inherent electrical pulsation signals. The second difference involves comparing the interference information's characteristics with two key reference features. First, it is compared with the inherent electrical pulsation signal characteristics of the high-voltage power supply under current operating conditions to quantify the difference between the interference information and normal operating signals (the first difference). Second, it is compared with preset signal characteristics representing early faults in the high-voltage power supply to quantify the difference between the interference information and actual fault signals (the second difference). The aim is to accurately distinguish between interference, normal operating fluctuations, and early fault signals, avoiding misjudgments.

[0038] Based on the first and second differences, generating a discriminative feature set involves integrating and optimizing the identified first and second differences to construct a comprehensive feature set capable of effectively distinguishing interference information, inherent electrical pulsation signals, and early fault signals. This set can be a multi-dimensional vector containing parameters that highlight the unique characteristics of the interference while weakening features similar to inherent pulsation and early fault signals. The aim is to provide a highly pure and discriminative feature input for subsequent interference pattern determination, thereby improving the accuracy and robustness of interference pattern recognition.

[0039] This application introduces continuous monitoring of the high-voltage power supply's operating status and adaptively adjusts interference information analysis parameters according to different operating conditions. This enables time-domain and frequency-domain analysis to more accurately capture the interference characteristics introduced by the data acquisition link under the current operating conditions. Because the interference information analysis parameters can be dynamically adjusted according to actual operating conditions, the extracted interference characteristics are more representative and accurate. Based on this, by comparing the extracted interference characteristics with inherent electrical pulsation signal characteristics and early fault signal characteristics, the interference information introduced by the data acquisition link can be effectively distinguished from the normal operating fluctuations of the power supply itself and potential early fault signals. This dual-differentiation identification mechanism fundamentally solves the problem of interference information being confused with inherent pulsation or early fault signals in traditional methods, ensuring the accuracy of subsequent interference mode determination.

[0040] In some embodiments, the step of continuously monitoring the operating status of the high-voltage power supply and acquiring the characteristics of the inherent electrical pulsation signals generated by the power supply itself under different operating conditions includes: Continuously monitor multiple operating parameters of the high-voltage power supply; Calculate multiple operating condition characteristic indicators based on operating parameters; The operating condition characteristic indicators are compared with the preset operating condition interval boundaries to identify the current operating condition; Based on the current operating conditions, retrieve the matching inherent electrical pulsation signal characteristics.

[0041] Specifically, continuous monitoring of multiple operating parameters of a high-voltage power supply refers to the real-time acquisition of key electrical and physical parameters during operation, such as output voltage, output current, frequency, temperature, and load rate, through sensors or existing monitoring systems. These parameters are fundamental data reflecting the current operating status of the high-voltage power supply. Calculating multiple operating condition characteristic indicators based on these parameters can be understood as processing and integrating the collected raw operating parameters to form indicators that comprehensively characterize the operating conditions of the high-voltage power supply. For example, power can be obtained by multiplying voltage and current, or a heat load indicator can be obtained by weighted averaging temperature and load rate. These operating condition characteristic indicators aim to describe the actual operating status of the high-voltage power supply more comprehensively and concisely. In practical applications, the operating condition characteristic indicators are compared with preset operating condition interval boundaries to identify the current operating condition. Specifically, the calculated operating condition characteristic indicators are compared with predefined interval boundaries representing different operating conditions (such as light load, medium load, heavy load, overload, etc.). This comparison allows for accurate determination of the specific operating condition of the high-voltage power supply. For example, when the power index falls within a specific range, it can be identified as a medium-load condition. The purpose is to provide accurate operating condition data for subsequent retrieval of matching inherent electrical pulsation signal characteristics. Furthermore, retrieving matching inherent electrical pulsation signal characteristics based on the current operating condition means that once the current operating condition is identified, the system will retrieve and obtain the inherent electrical pulsation signal characteristics corresponding to that current operating condition from a pre-established feature library. These characteristics are generated when the high-voltage power supply is in a healthy state and operating under specific conditions, representing the normal fingerprint information of the power supply itself under those conditions. The purpose is to provide an accurate benchmark for subsequent interference information analysis.

[0042] This application continuously monitors multiple operating parameters of the high-voltage power supply and calculates operating condition characteristic indicators based on these parameters, thereby accurately identifying the current operating condition of the high-voltage power supply. It is precisely because of this accurate identification of the current operating condition that the system can retrieve the inherent electrical pulsation signal characteristics that precisely match that operating condition from a preset feature library. This method of dynamically matching inherent characteristics based on real-time operating conditions ensures that the reference signal used for subsequent time-domain and frequency-domain analysis of interference information is highly consistent with the actual operating state of the power supply, thus avoiding misjudgments or inaccurate analysis caused by changes in operating conditions.

[0043] In some embodiments, the above-described method for predicting high-voltage power supply failures in electrostatic precipitators for thermal power plants further includes: Real-time verification of inherent electrical pulsation signal characteristics is performed, specifically as follows: Obtain the actual electrical pulsation signal of the high-voltage power supply; Extract multiple features from actual electrical pulsation signals; A multi-dimensional comparison is performed between the retrieved inherent electrical pulsation signal features and multiple features of the extracted actual electrical pulsation signal; The accuracy of the retrieved inherent electrical pulsation signal characteristics is determined based on the relationship between the comparison results and the preset verification threshold. In response to the inaccuracy of the inherent electrical pulsation signal characteristics retrieved, the source of the deviation is determined based on the specific manifestation of the deviation and the changing trend of the current operating parameters; Based on the determination of the source of the deviation, the characteristics of the retrieved inherent electrical pulsation signal are adaptively corrected.

[0044] Specifically, acquiring the actual electrical pulsation signal of a high-voltage power supply refers to the real-time acquisition of the electrical pulsation signal generated by the high-voltage power supply under its current operating state using high-precision sensors or data acquisition equipment. These signals typically contain instantaneous fluctuation information of various electrical parameters such as voltage, current, and frequency. Extracting multiple features of the actual electrical pulsation signal can be understood as preprocessing and feature engineering the acquired raw actual electrical pulsation signal. For example, using methods such as Fourier transform, wavelet analysis, and statistical analysis, characteristic parameters in the time domain, frequency domain, and time-frequency domain, such as mean, variance, peak value, harmonic content, and energy distribution, can be extracted to comprehensively characterize the dynamic characteristics of the signal.

[0045] Multi-dimensional comparison of retrieved inherent electrical pulsation signal features and extracted actual electrical pulsation signal features involves comparing the inherent electrical pulsation signal features retrieved from the feature library that match the current operating conditions with the real-time extracted actual electrical pulsation signal features across multiple dimensions. These dimensions may include, but are not limited to, numerical differences in feature values, trend consistency, and spectral similarity. Comparison methods can employ various mathematical or statistical methods such as Euclidean distance, cosine similarity, and correlation coefficients. The accuracy of the retrieved inherent electrical pulsation signal features is determined based on the relationship between the comparison results and a preset verification threshold. The purpose is to assess the degree of conformity between the retrieved inherent features and the actual operating conditions. The preset verification threshold is an allowable error range determined based on historical data, expert experience, or system design requirements. If the comparison results exceed this threshold, the retrieved inherent electrical pulsation signal features are considered to have a deviation and require correction.

[0046] In response to the judgment that the retrieved inherent electrical pulsation signal characteristics are inaccurate, the source of the deviation is determined based on the specific manifestations of the deviation and the changing trends of the current operating parameters. This means that after identifying the deviation in the inherent electrical pulsation signal characteristics, further analysis is conducted to determine the cause of the deviation. Specific manifestations of the deviation may include persistently high or low characteristic values, periodic fluctuations, and enhancement or weakening of specific frequency components. The changing trends of the current operating parameters reflect changes in the operating environment or load conditions of the high-voltage power supply. By cross-analyzing these two types of information, it is possible to distinguish whether the deviation is caused by sensor failure, environmental interference, power supply aging, or changes in operating conditions. Adaptive correction of the retrieved inherent electrical pulsation signal characteristics based on the determination of the deviation source refers to adopting corresponding correction strategies according to the identified deviation source. For example, if it is sensor drift, the sensor data can be calibrated; if it is power supply aging, the corresponding parameters in the inherent electrical pulsation signal characteristic library can be updated; if it is a change in operating conditions, the feature retrieval logic can be adjusted or new operating condition characteristics can be introduced. This correction process is adaptive, enabling the system to maintain high accuracy under constantly changing operating conditions.

[0047] This application effectively solves the problem of discrepancies between inherent electrical pulsation signal characteristics and actual operating conditions by introducing a real-time verification mechanism. Specifically, by acquiring the actual electrical pulsation signal of the high-voltage power supply in real time and extracting its characteristics, a multi-dimensional comparison is performed with the pre-retrieved inherent electrical pulsation signal characteristics, enabling timely detection of deviations. Furthermore, by analyzing the specific manifestations of the deviation and the changing trends of current operating parameters, the source of the deviation can be accurately determined, thereby avoiding misjudging signal changes caused by non-core fault factors (such as sensor drift and environmental interference) as core faults. Finally, the retrieved inherent electrical pulsation signal characteristics are adaptively corrected based on the source of the deviation, ensuring that the reference mode used for subsequent fault prediction always closely matches the actual operating conditions of the high-voltage power supply, greatly improving the accuracy and reliability of fault prediction.

[0048] Suppose that after a period of operation, the high-voltage power supply of an electrostatic precipitator in a thermal power plant, when comparing its retrieved inherent electrical pulsation signal characteristics with the real-time acquired actual electrical pulsation signal characteristics, finds that a harmonic component at a specific frequency is persistently high, and the comparison result exceeds a preset verification threshold. At this point, the system further analyzes the specific manifestations of the deviation (e.g., the harmonic component is abnormally enhanced under a specific load) and the changing trends of current operating parameters (e.g., the output voltage or current of the high-voltage power supply fluctuates within the normal range, but the ambient temperature is slightly increased). Through cross-analysis, the system determines that the deviation is not caused by an early fault in the power supply itself, but rather by a slight change in the transmission characteristics of the data acquisition link due to the increased ambient temperature. Based on this determination, the system fine-tunes the parameters related to the harmonic component at that frequency in the retrieved inherent electrical pulsation signal characteristics to better match the signal characteristics under the current actual operating environment. Through this adaptive correction, the system avoids misjudging signal changes caused by environmental factors as power supply faults, thereby ensuring the accuracy of fault prediction.

[0049] In some embodiments, the step of determining the source of deviation based on the specific manifestation of the deviation and the changing trend of the current operating parameters in response to the determination that the retrieved inherent electrical pulsation signal characteristics are inaccurate includes: First, in response to the judgment that the retrieved inherent electrical pulsation signal characteristics are inaccurate, the specific manifestations of the deviation are extracted from the actual electrical pulsation signal. This step aims to accurately identify and quantify the parts of the actual electrical pulsation signal that do not match the retrieved inherent electrical pulsation signal characteristics through signal processing techniques. Specifically, methods such as differential analysis, spectrum comparison, and waveform matching can be used to extract key features of the deviation, such as amplitude, frequency, and phase, thereby forming a quantitative description of the deviation phenomenon.

[0050] Secondly, based on multiple operating parameters of the high-voltage power supply, the rate of change of these parameters and their acceleration are calculated. Operating parameters may include, but are not limited to, the output voltage, output current, ambient temperature, and load change rate of the high-voltage power supply. By continuously monitoring and sampling these parameters, their rate of change over time (i.e., the speed of change) and the rate of change of that rate (i.e., acceleration) can be calculated. For example, algorithms such as moving average and Kalman filtering can be used to smooth the raw parameter data, and then the speed and acceleration can be calculated through differential or regression analysis. The current trend of the operating parameters is obtained based on the rate of change, and the curvature change characteristics of this current trend are obtained based on the acceleration. The current trend and the curvature change characteristics are superimposed to generate the trend of the operating parameters. The current trend reflects the immediate direction of change of the operating parameters, such as whether it is rising, falling, or remaining stable. The curvature change characteristics describe the speed of this trend change and the degree of curvature, such as accelerating upwards or decelerating downwards. By superimposing these two, a more comprehensive and dynamic model of the operating parameter change trend can be constructed, which can reflect the short-term and medium-term evolution of the high-voltage power supply's operating conditions.

[0051] Then, based on pre-defined association rules, a cross-analysis is performed on the specific manifestations of the deviation and the changing trends of the operating parameters. Association rules are knowledge bases built upon historical data, expert experience, or physical models, used to describe the intrinsic relationship between specific types of electrical ripple deviations and the changing trends of specific operating parameters. For example, a rule might indicate that when the output current rises rapidly and the electrical ripple signal exhibits enhanced harmonics at a specific frequency, it may indicate a sudden load change rather than a power supply failure. This cross-analysis aims to evaluate, through logical reasoning or pattern matching, whether the currently observed deviation manifestations and operating condition changing trends conform to known association patterns.

[0052] Finally, based on the analysis results, the source of the deviation is determined. Through the above cross-analysis, the system can distinguish between normal drift of inherent electrical pulsation signal characteristics caused by changes in operating conditions and abnormal deviations caused by faults in the high-voltage power supply itself or the data acquisition link. For example, if the deviation closely matches the specific changing trend of the operating parameters according to the preset operating condition adaptive adjustment rules, the deviation is determined to originate from changes in operating conditions; conversely, if the deviation closely matches the early fault characteristic rules, the deviation is determined to originate from a fault in the power supply itself.

[0053] This application, through refined extraction of deviations in actual electrical pulsation signals, combined with comprehensive modeling of the dynamic changes in high-voltage power supply operating parameters, and intelligent cross-analysis using preset association rules, enables accurate identification of the source of deviations occurring during the verification of inherent electrical pulsation signal characteristics. This method avoids misjudging signal differences caused by changes in normal operating conditions as faults, and also prevents changes in operating conditions from masking the occurrence of true faults.

[0054] In some embodiments, the step of cross-analyzing the specific manifestations of the deviation and the changing trends of the operating parameters according to preset association rules includes: Feature extraction is performed on the specific manifestations of the deviation and the changing trends of the operating parameters to obtain multiple feature parameters; Based on multiple feature parameters, a subset of rules related to the current running state is selected from preset association rules; Prioritize the rules in the rule subset and resolve potential conflicts between the rules; Based on the sorted subset of rules, logical judgments are made on the specific manifestations of the deviation and the changing trends of the operating parameters to draw conclusions about the source of the deviation.

[0055] Feature extraction refers to identifying and quantifying representative and discriminative information from the raw data on specific deviations and trends in operating parameters. For example, for specific deviations, amplitude, frequency, duration, and waveform characteristics can be extracted; for trends in operating parameters, rate of change, acceleration, and periodicity can be extracted. The purpose of extracting these feature parameters is to transform complex raw data into numerical or symbolic representations that are easier to analyze and compare.

[0056] The pre-defined association rules are a knowledge base built upon historical data, expert experience, or machine learning models, used to describe the potential relationships between deviation performance, changes in operating conditions, and sources of deviation. The process of filtering a subset of rules involves matching multiple currently extracted feature parameters with the conditions in the association rules to identify the set of rules most relevant to the current operating state. This effectively narrows the scope of analysis, improving efficiency and accuracy.

[0057] Prioritizing rules within a subset of rules enables decision-making based on their importance or reliability when multiple rules simultaneously meet certain conditions. For example, rules can be ranked according to their confidence level, support level, or weights set by experts. Resolving potential conflicts refers to ensuring the consistency and reasonableness of the analysis results when different rules may point to contradictory sources of bias, using pre-defined conflict resolution strategies (such as majority voting, weighted averaging, or selecting the highest-priority rule).

[0058] Logical judgment based on the sorted subset of rules, examining the specific manifestations of deviations and the changing trends of operating parameters, involves using the filtered and sorted rules, combined with the currently extracted feature parameters, to reason and deduce conclusions about the sources of deviations. This logical judgment can be implemented using expert systems, decision trees, or fuzzy logic, aiming to provide a clear and interpretable indication of the source of deviations.

[0059] This application, through refined feature extraction of the specific manifestations of deviations and the changing trends of operating parameters, transforms complex raw data into structured, quantifiable information, thus laying the foundation for subsequent correlation analysis. By selecting a subset of rules highly relevant to the current operating state from a vast association rule base, the analysis scope can be effectively focused, avoiding interference from irrelevant information and improving the relevance of the analysis. Furthermore, prioritizing the rule subset and resolving potential conflicts ensures that, under the influence of multiple intertwined factors, the system can make judgments based on the most reliable and relevant rules, thereby avoiding misjudgments or uncertainties. Therefore, this solution enables accurate and efficient preliminary judgment of the source of deviations, providing a solid basis for subsequent fault diagnosis and handling.

[0060] In some embodiments, the steps of prioritizing rules in a subset of rules and resolving potential conflicts between rules include: Obtain the stability parameters of the electrical pulsation signal characteristics associated with each rule in the rule subset; Obtain the sensitivity parameters of the rate of change of operating condition parameters associated with each rule in the rule subset; Based on the stability parameter and sensitivity parameter, calculate the feature discrimination strength of each rule according to the feature discrimination strength calculation formula; Based on the feature discrimination strength, the rules in the rule subset are prioritized from high to low, and the conclusion of the rule with the highest feature discrimination strength is selected as the preferred conclusion. When the difference in the feature discrimination strength of two or more rules in the rule subset is less than the preset discrimination strength threshold, the overlap parameter between the electrical pulsation signal features associated with the two or more rules and the early fault signal features of the power supply body is obtained. The overlap parameter is the cosine similarity or the reciprocal of the Euclidean distance between the two feature vectors in the feature space. Based on the overlap parameter from high to low, two or more rules are sorted a second time, and the conclusion of the rule with the highest overlap parameter is taken as the final conclusion after conflict resolution, and a conflict-free preferred rule subset is generated.

[0061] Specifically, the stability parameter refers to the fluctuation range or coefficient of variation of the electrical pulsation signal characteristic associated with the rule under normal operating conditions, reflecting the reliability of the characteristic under stable operating conditions. The smaller the stability parameter, the more stable the characteristic, and the higher its reliability as a judgment criterion. The sensitivity parameter refers to the degree to which the rate of change of the operating condition parameters associated with the rule affects the electrical pulsation signal characteristic, reflecting the rule's responsiveness to changes in operating conditions. The larger the sensitivity parameter, the faster the rule responds to changes in operating conditions, and the higher its effectiveness in judging under dynamic operating conditions.

[0062] Feature discrimination strength is a quantitative assessment of a rule's discriminative ability, taking into account both stability and sensitivity parameters. For example, the formula for calculating feature discrimination strength can be the product of the reciprocal of the stability parameter and the sensitivity parameter, or other weighted combination formulas. Higher feature discrimination strength indicates that the rule has a stronger discriminative ability and higher priority in identifying sources of bias.

[0063] When the discriminative strengths of multiple rules are similar, additional discriminative criteria are needed for more refined priority ranking. In this case, the overlap parameter between the electrical pulsation signal features associated with the rule and the early fault signal features of the power supply itself is obtained. The overlap parameter refers to the degree of similarity or the size of the overlapping region between the electrical pulsation signal features of interest to the rule and the known early fault signal features of the power supply itself in the feature space. The smaller the overlap parameter, the higher the distinguishability between the features associated with the rule and the early fault signal, and the greater its value in eliminating interference from the power supply itself and accurately identifying the source of deviation; therefore, it should be given higher priority.

[0064] This application addresses the potential for insufficient accuracy and misjudgment in rule priority determination by introducing multi-dimensional parameters for rule prioritization. Specifically, firstly, by acquiring the stability parameters of the electrical pulsation signal characteristics associated with the rules, the reliability of the rules under stable operating conditions can be assessed, avoiding misjudgments caused by large fluctuations in the characteristics themselves. Secondly, by acquiring the sensitivity parameters of the rate of change of the operating condition parameters associated with the rules, the responsiveness of the rules to changes in operating conditions can be measured, ensuring that under dynamic operating conditions, rules that are sensitive to changes in operating conditions and can promptly reflect deviations are prioritized. Therefore, the feature discrimination strength calculated based on the stability and sensitivity parameters can comprehensively reflect the reliability and responsiveness of the rules, thereby achieving a preliminary and more scientific prioritization.

[0065] To avoid priority ambiguity caused by similar discrimination strengths when the feature discrimination strengths of rules are close, this application introduces an overlap parameter. By obtaining the overlap parameter between the electrical pulsation signal features associated with the rule and the early fault signal features of the power supply itself, the ability of the rule to distinguish between data acquisition link interference and early faults in the power supply itself can be evaluated. The lower the overlap, the higher the distinguishability between the features of interest to the rule and the fault signal of the power supply itself, thus enabling a more accurate attribution of deviation to the data acquisition link and avoiding misjudging the power supply itself as link interference, or vice versa. This multi-level, multi-dimensional priority ranking mechanism ensures that the most effective and accurate rule can be selected to determine the source of deviation in a complex and ever-changing operating environment, thereby providing a reliable basis for subsequent adaptive correction.

[0066] Suppose that when conducting cross-analysis of the specific manifestations of the deviation and the changing trends of the operating parameters, two rules, A and B, are selected, and their initial feature discrimination strengths are very similar.

[0067] The electrical pulsation signal characteristics associated with Rule A fluctuate little under normal operating conditions, with a stability parameter of 0.05; the rate of change of the associated operating condition parameters has a significant impact on the electrical pulsation signal characteristics, with a sensitivity parameter of 0.8.

[0068] The electrical pulsation signal characteristics associated with Rule B fluctuate slightly more under normal operating conditions, with a stability parameter of 0.08; the rate of change of the associated operating condition parameters also has a significant impact on the electrical pulsation signal characteristics, with a sensitivity parameter of 0.7.

[0069] According to preliminary calculations, the feature discrimination strengths of rule A and rule B are likely to be similar.

[0070] At this point, it is necessary to further obtain the overlap parameter. Assuming the overlap parameter between the electrical pulsation signal characteristics associated with rule A and the early fault signal characteristics of the power supply body is 0.1, it indicates a high degree of distinguishability between them and the power supply body's fault signal. However, the overlap parameter between the electrical pulsation signal characteristics associated with rule B and the early fault signal characteristics of the power supply body is 0.3, indicating a relatively high degree of overlap and low distinguishability.

[0071] Based on the aforementioned overlap parameters, this application prioritizes rule A. This is because, although rule A and rule B have similar overall discriminative strength in terms of stability and sensitivity, rule A demonstrates a higher ability to distinguish between data acquisition link interference and early power supply failures. In this way, even when the rule discriminative strengths are similar, a deeper level of screening can be achieved by introducing the overlap parameter, ensuring that the final rule used to determine the source of deviation has the highest accuracy and reliability, thereby avoiding misjudging power supply failures as link interference, or vice versa.

[0072] like Figure 2 As shown in the figure, this application proposes a high-voltage power supply fault prediction system for electrostatic precipitators in thermal power plants, comprising: Information acquisition module 1 is used to acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions, and to establish a reference pattern for the inherent electrical pulsation information; Real-time monitoring module 2 is used to continuously acquire real-time electrical pulsation information of the high-voltage power supply; The comparison and identification module 3 is used to compare the real-time electrical pulsation information with the reference mode to identify interference information introduced by the data acquisition link; The mode determination module 4 is used to determine the interference mode of the data acquisition link based on the interference information. Information processing module 5 is used to process the operational information transmitted through the data acquisition link according to the interference mode in order to obtain clean operational information. Fault identification module 6 is used to identify early fault information of the high-voltage power supply body based on clean operating information.

[0073] The system in this application establishes an inherent electrical pulsation reference mode as a benchmark for healthy operation through an information acquisition module, and continuously acquires real-time electrical pulsation information through a real-time monitoring module. A comparison and identification module effectively distinguishes between the inherent characteristics of the power supply itself and anomalies introduced by the data acquisition link. A pattern determination module further analyzes these anomalies to identify specific interference patterns. An information processing module then selectively purifies the original operating information based on the identified interference patterns, thereby obtaining clean operating information. Finally, a fault identification module focuses on analyzing this clean power supply operating information, significantly improving the accuracy and reliability of fault identification. This hierarchical diagnostic system architecture not only reduces false alarms and lowers maintenance costs, but more importantly, it ensures that when true early fault signals appear, they can be identified promptly and accurately, thus avoiding the risk of overlooking real faults due to misjudgment and greatly improving the operational reliability and safety of the high-voltage power supply in the electrostatic precipitator system.

[0074] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants, characterized in that, include: Acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions, and establish a reference pattern for the inherent electrical pulsation information; Continuously acquire real-time electrical pulsation information of the high-voltage power supply; Real-time electrical pulsation information is compared with a reference pattern to identify interference introduced by the data acquisition link; Based on the interference information, determine the interference mode of the data acquisition link; Based on the interference pattern, the operational information transmitted through the data acquisition link is processed to obtain clean operational information. Based on clean operational information, early fault information of the high-voltage power supply body can be identified.

2. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 1, characterized in that, Also includes: The real-time electrical pulsation information is compared with the reference mode, and the dynamic transmission characteristics of the data acquisition link are evaluated in real time based on the comparison results. Based on the dynamic transmission characteristics, the operational information transmitted through the data acquisition link is reverse-engineered to obtain clean operational information.

3. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 1, characterized in that, The step of determining the interference mode of the data acquisition link based on the interference information includes: Time-domain and frequency-domain characteristic analysis of interference information is performed to obtain the performance characteristics of interference information in different time scales and frequency ranges; The performance characteristics are compared with a preset degradation mode feature library to identify single or compound degradation modes in the data acquisition link. Based on the comparison results, the impact of each degradation mode in a single or combined degradation mode on the real-time electrical pulsation information of the high-voltage power supply is quantified in order to determine the interference mode of the data acquisition link.

4. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 3, characterized in that, The step of performing time-domain and frequency-domain characteristic analysis on the interference information to obtain the performance characteristics of the interference information in different time scales and frequency ranges includes: Continuously monitor the operating status of the high-voltage power supply and acquire the characteristics of the inherent electrical pulsation signals generated by the power supply body under different operating conditions; Based on the operating conditions of the high-voltage power supply, adjust the parameters for time-domain and frequency-domain analysis of the interference information; Based on the adjusted parameters, time-domain and frequency-domain analyses are performed on the interference information to extract its performance characteristics within the corresponding time scale and frequency range. The characteristics of the extracted interference information are compared with the characteristics of the inherent electrical pulsation signal and the characteristics of the early fault signal of the high-voltage power supply, respectively, in order to identify the first difference between the interference information and the inherent electrical pulsation signal, and the second difference between the interference information and the early fault signal. Based on the first and second differences, a set of discriminative features is generated.

5. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 4, characterized in that, The step of continuously monitoring the operating status of the high-voltage power supply and acquiring the characteristics of the inherent electrical pulsation signals generated by the power supply itself under different operating conditions includes: Continuously monitor multiple operating parameters of the high-voltage power supply; Calculate multiple operating condition characteristic indicators based on operating parameters; The operating condition characteristic indicators are compared with the preset operating condition interval boundaries to identify the current operating condition; Based on the current operating conditions, retrieve the matching inherent electrical pulsation signal characteristics.

6. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 5, characterized in that, Also includes: Real-time verification of inherent electrical pulsation signal characteristics is performed, specifically as follows: Obtain the actual electrical pulsation signal of the high-voltage power supply; Extract multiple features from actual electrical pulsation signals; A multi-dimensional comparison is performed between the retrieved inherent electrical pulsation signal features and multiple features of the extracted actual electrical pulsation signal; The accuracy of the retrieved inherent electrical pulsation signal characteristics is determined based on the relationship between the comparison results and the preset verification threshold. In response to the inaccuracy of the inherent electrical pulsation signal characteristics retrieved, the source of the deviation is determined based on the specific manifestation of the deviation and the changing trend of the current operating parameters; Based on the determination of the source of the deviation, the characteristics of the retrieved inherent electrical pulsation signal are adaptively corrected.

7. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 6, characterized in that, The step of determining the source of deviation based on the specific manifestation of the deviation and the changing trend of the current operating parameters, in response to the determination that the retrieved inherent electrical pulsation signal characteristics are inaccurate, includes: In response to the inaccuracy of the inherent electrical pulsation signal characteristics retrieved, the specific manifestations of the deviation are extracted from the actual electrical pulsation signal; Based on multiple operating parameters of the high-voltage power supply, the rate of change of the operating parameters is calculated, and the acceleration of the operating parameters is calculated. The current trend of the operating parameters is obtained based on the rate of change, and the curvature change characteristics of the current trend are obtained based on the acceleration. The current trend and the curvature change characteristics are superimposed to generate the trend of the operating parameters. Based on the preset correlation rules, cross-analysis is performed on the specific manifestations of the deviation and the changing trends of the operating parameters; Based on the analysis results, determine the source of the deviation.

8. The method for predicting high-voltage power supply faults in electrostatic precipitators of thermal power plants according to claim 7, characterized in that, The step of cross-analyzing the specific manifestations of the deviation and the changing trends of the operating parameters according to preset association rules includes: Feature extraction is performed on the specific manifestations of the deviation and the changing trends of the operating parameters to obtain multiple feature parameters; Based on multiple feature parameters, a subset of rules related to the current running state is selected from preset association rules; Prioritize the rules in the rule subset and resolve potential conflicts between the rules; Based on the sorted subset of rules, logical judgments are made on the specific manifestations of the deviation and the changing trends of the operating parameters to draw conclusions about the source of the deviation.

9. A method for predicting high-voltage power supply faults in electrostatic precipitators for thermal power plants according to claim 8, characterized in that, The steps of prioritizing rules in the rule subset and resolving potential conflicts between rules include: Obtain the stability parameters of the electrical pulsation signal characteristics associated with each rule in the rule subset; Obtain the sensitivity parameters of the rate of change of operating condition parameters associated with each rule in the rule subset; Based on the stability parameter and sensitivity parameter, calculate the feature discrimination strength of each rule according to the feature discrimination strength calculation formula; The rules in the rule subset are prioritized according to the feature discrimination strength from high to low, and the conclusion of the rule with the highest feature discrimination strength is selected as the preferred conclusion. When the difference in the feature discrimination strength of two or more rules in the rule subset is less than the preset discrimination strength threshold, the overlap parameter between the electrical pulsation signal features associated with the two or more rules and the early fault signal features of the power supply body is obtained. The overlap parameter is the cosine similarity or the reciprocal of the Euclidean distance between the two feature vectors in the feature space. Based on the overlap parameter from high to low, two or more rules are sorted a second time, and the conclusion of the rule with the highest overlap parameter is taken as the final conclusion after conflict resolution, and a conflict-free preferred rule subset is generated.

10. A fault prediction system for high-voltage power supply of electrostatic precipitators in thermal power plants, characterized in that, include: The information acquisition module is used to acquire the inherent electrical pulsation information generated by the high-voltage power supply under specific operating conditions and to establish a reference pattern for the inherent electrical pulsation information. The real-time monitoring module is used to continuously acquire real-time electrical pulsation information of the high-voltage power supply; The comparison and identification module is used to compare real-time electrical pulsation information with a reference mode to identify interference information introduced by the data acquisition link; The mode determination module is used to determine the interference mode of the data acquisition link based on the interference information. The information processing module is used to process the operational information transmitted through the data acquisition link according to the interference mode in order to obtain clean operational information. The fault identification module is used to identify early fault information of the high-voltage power supply body based on clean operating information.

Images