A fault early warning method and system for rectification of a direct current submerged arc furnace

Abstract

The present application belongs to the technical field of fault detection, and specifically relates to a fault early warning method and system for rectification of a direct current (DC) electric arc furnace, which comprises obtaining electrical parameters, temperature parameters and insulation parameters at multiple monitoring positions in a rectification system of the DC electric arc furnace, and constructing a multi-dimensional feature parameter set; the multi-dimensional feature parameter set is preprocessed to obtain a standardized feature parameter set; the standardized feature parameter set is input into a preset fault location model, the fault location model is constructed based on a mapping relationship between fault features of each component of the rectification system and a multi-parameter abnormal mode, and outputs a fault type and a fault location of a hidden fault; according to the fault type and the fault location, a preset fault level determination rule is combined to determine an early warning level of the hidden fault, and a response strategy corresponding to the early warning level is executed. The present application solves the problems of difficulty in locating hidden faults and low efficiency caused by one-by-one checking in the prior art.

Description

Technical Field

[0001] This invention belongs to the field of fault detection technology, specifically a fault early warning method and system for DC submerged arc furnace rectification. Background Technology

[0002] The statements in this section merely refer to the background art related to this invention and do not necessarily constitute prior art.

[0003] The DC submerged arc furnace is a core piece of equipment in ferroalloy smelting. Its rectifier system is responsible for converting AC power into DC power and stably supplying it to the smelting process. The rectifier system consists of a rectifier transformer, a thyristor module, a smoothing reactor, a cooling system, and electrical connection components. During operation, it is subjected to high voltage, high current, and strong electromagnetic interference, making it prone to various electrical faults.

[0004] Existing fault warning technologies mostly target explicit faults (such as thyristor breakdown, short circuits, and open circuits), triggering alarms by detecting sudden changes in parameters such as voltage and current. However, in actual operation, latent electrical faults are more common, such as abnormal increases in thyristor junction temperature, subtle increases in contact resistance, slow deterioration of insulation performance, and subtle blockages in cooling pipes. These faults initially show no obvious parameter changes and are difficult to identify through conventional methods, but they gradually accumulate and worsen, eventually leading to explicit faults, causing rectifier system shutdowns and smelting interruptions.

[0005] In early warning schemes for latent faults, while existing technologies attempt to improve identification capabilities through multi-parameter monitoring, fault location remains challenging. When a latent fault occurs, multiple electrical and temperature parameters often exhibit abnormal fluctuations simultaneously, with these parameters coupling and overlapping characteristics. Existing methods can only determine the general range of the fault, failing to pinpoint its exact location and type. This necessitates that maintenance personnel still need to inspect each component individually when faced with early warning information, resulting in low fault handling efficiency. Summary of the Invention

[0006] This invention provides a fault early warning method and system for DC submerged arc furnace rectifiers, aiming to solve the core problem of difficulty in locating latent electrical faults in DC submerged arc furnace rectifier systems. It provides a hierarchical early warning method and system that can accurately decouple fault sources from complex features of multi-parameter coupling and quickly determine the specific location and type of latent faults. This overcomes the shortcomings of low fault investigation efficiency and difficulty in implementing early detection and treatment due to the inability to accurately locate faults, and realizes the leap from general alarm to precise location of latent faults.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: The first aspect of this invention discloses a fault early warning method for the rectification of a DC submerged arc furnace, comprising the following steps: The electrical, temperature, and insulation parameters of multiple monitoring locations in the DC submerged arc furnace rectifier system are acquired to construct a multi-dimensional feature parameter set. The multi-dimensional feature parameter set is preprocessed to obtain a standardized feature parameter set; The standardized feature parameter set is input into the preset fault location model. The fault location model is constructed based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal mode, and outputs the fault type and fault location of the latent fault. Based on the fault type and location, and combined with the preset fault level judgment rules, the warning level of the latent fault is determined, and the response strategy corresponding to the warning level is executed.

[0008] Furthermore, the multi-dimensional feature parameter set also includes environmental parameters and equipment status parameters; Electrical parameters include at least one of the following: three-phase voltage imbalance on the rectifier input side, DC voltage ripple coefficient on the rectifier output side, thyristor conduction angle deviation, bridge arm current sharing rate, and smoothing reactor current harmonic distortion rate. Temperature parameters include at least one of the following: junction temperature of the thyristor module, winding temperature of the rectifier transformer, temperature of the smoothing reactor, temperature of the electrical connection point, and temperature difference between the inlet and outlet water of the cooling system. Insulation parameters include at least one of the following: insulation resistance of rectifier transformer, insulation resistance of thyristor module, insulation resistance of DC bus, and dielectric loss factor of cable insulation layer; Environmental parameters include at least one of the following: internal temperature of the rectifier cabinet, humidity, and dust concentration; The equipment status parameters include at least one of the following: thyristor trigger pulse amplitude deviation, cooling fan speed deviation, and rectifier transformer noise amplitude.

[0009] Further preprocessing includes: The 3σ criterion was used to remove outliers from the original data. Wavelet transform algorithm is used to denoise the data after outlier removal; The min-max normalization method is used to map the denoised data to a set range; Linear interpolation was used to fill in missing values ​​in the standardized data.

[0010] Furthermore, before inputting the standardized feature parameter set into the preset fault location model, a feature fusion step is also included: The weights of each standardized characteristic parameter are determined by the analytic hierarchy process (AHP), and the comprehensive fault characteristic value is calculated by weighted summation. The comprehensive fault feature value is compared with the preset identification threshold. If the comprehensive fault feature value is greater than the identification threshold, it is determined that there is a hidden fault, and the subsequent steps of inputting into the preset fault location model are executed.

[0011] Furthermore, the fault level determination rules include: The comprehensive fault characteristic value is compared with the preset first-level warning threshold and second-level warning threshold; If the comprehensive fault characteristic value is greater than the identification threshold and less than or equal to the first-level warning threshold, it is determined to be a first-level warning; If the comprehensive fault characteristic value is greater than the first-level warning threshold and less than or equal to the second-level warning threshold, it is determined to be a second-level warning; If the comprehensive fault characteristic value is greater than the level 2 warning threshold, it is determined to be a level 3 warning.

[0012] Furthermore, implement response strategies corresponding to the warning level, including: If the warning level is Level 1, then a green warning will be issued, and the corresponding area will be given special attention during routine inspections. If the warning level is Level II, a yellow warning will be issued, triggering an audible and visual alarm and a terminal information push, requiring a comprehensive inspection of the warning area within a specified time. If the warning level is Level 3, a red warning will be issued and an emergency shutdown will be triggered to conduct a comprehensive investigation and handling of the fault.

[0013] Furthermore, the fault location model outputs the fault type and fault location of latent faults based on the abnormality of each characteristic parameter, the fault level, and the equipment layout of the rectifier system. After the fault is handled, the fault data, early warning process, and handling results are reviewed, and the mapping relationship of the fault location model is updated, the fault level judgment rules are adjusted, and the corresponding response strategies are implemented.

[0014] A second aspect of the present invention discloses a fault early warning system for the rectification of a DC submerged arc furnace, comprising: The data acquisition unit is configured to acquire electrical parameters, temperature parameters, and insulation parameters at multiple monitoring locations in the DC submerged arc furnace rectifier system, and construct a multi-dimensional feature parameter set. The data preprocessing unit is configured to preprocess the multi-dimensional feature parameter set to obtain a standardized feature parameter set. The fault location unit is configured to: input a standardized set of feature parameters into a preset fault location model, which is constructed based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal mode, and output the fault type and fault location of the latent fault. The fault early warning unit is configured to determine the early warning level of a latent fault based on the fault type and fault location, combined with preset fault level judgment rules, and execute the response strategy corresponding to the early warning level.

[0015] A third aspect of the present invention discloses a computer program product including computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the aforementioned fault warning method for DC submerged arc furnace rectification.

[0016] A fourth aspect of the present invention discloses an electronic device, including at least one processor and a memory connected to the processor, the memory being used to store a computer program; the processor being used to execute the computer program, enabling the electronic device to implement the above-described fault warning method for DC submerged arc furnace rectification.

[0017] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects: By constructing a set of multi-dimensional feature parameters covering electrical, temperature, and insulation aspects, and establishing a fault location model based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal modes, the fault source can be accurately decoupled from the complex features of multi-parameter coupling, and the specific location and type of latent faults can be quickly determined. This solves the core problem of the difficulty in locating latent faults and the inefficiency caused by the need to check each fault one by one in the existing technology. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 A schematic diagram of a fault early warning process for rectification in a DC submerged arc furnace, provided for one or more embodiments of the present invention; Figure 2 A schematic diagram of the fault early warning system architecture for DC submerged arc furnace rectification provided in one or more embodiments of the present invention. Detailed Implementation

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] As described in the background section, latent faults are characterized by subtle features and are often mixed with fluctuations in operating conditions, making it difficult to determine whether a fault has occurred. Secondly, different severity levels of faults require different handling strategies; otherwise, over-warning or under-warning may occur. Even if a fault is identified, if the specific component and fault type cannot be located, maintenance personnel still need to investigate step-by-step. Existing technologies mostly address the first two issues, lacking effective means to tackle the core challenge of fault localization.

[0023] Therefore, this solution provides a fault early warning method and system for DC submerged arc furnace rectification. By utilizing the differences in the distribution of fault characteristics of various components in the rectifier system in the parameter space, a mapping relationship between fault types and multi-parameter abnormal modes is established. Through hierarchical feature matching, the latent faults can be quickly and accurately located, fundamentally solving the problem of difficult location in latent fault early warning.

[0024] A fault early warning method for the rectifier of a DC submerged arc furnace includes the following steps: The electrical, temperature, and insulation parameters of multiple monitoring locations in the DC submerged arc furnace rectifier system are acquired to construct a multi-dimensional feature parameter set. The multi-dimensional feature parameter set is preprocessed to obtain a standardized feature parameter set; The standardized feature parameter set is input into the preset fault location model. The fault location model is constructed based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal mode, and outputs the fault type and fault location of the latent fault. Based on the fault type and location, and combined with the preset fault level judgment rules, the warning level of the latent fault is determined, and the response strategy corresponding to the warning level is executed.

[0025] like Figure 1 As shown, a fault early warning method for DC submerged arc furnace rectifiers is applied to the rectifier system of a DC submerged arc furnace in ferroalloy smelting. The specific steps are as follows: (1) Construct a latent fault characteristic parameter system that includes five categories: electrical, temperature, insulation, environment, and equipment status, and comprehensively cover the latent fault characteristics of thyristors, rectifier transformers, cooling systems, connecting components, and insulation systems.

[0026] (2) Multiple types of sensors are deployed in key parts of the rectifier system to collect feature parameters in real time at a frequency of 1~10Hz. After outlier removal, wavelet noise reduction, min-max standardization and data completion, a standardized feature parameter dataset is formed.

[0027] (3) The weights are determined by combining the analytic hierarchy process with expert experience, the multi-dimensional parameters are fused, the comprehensive fault feature value is calculated, and it is compared with the preset identification threshold to determine whether there is a hidden fault.

[0028] (4) Based on the magnitude of the comprehensive fault characteristic value, latent faults are divided into three levels: Level 1 mild, Level 2 moderate, and Level 3 severe.

[0029] (5) Green, yellow and red tiered early warnings are activated according to the fault level, and graded response strategies such as key inspections, time-limited maintenance and load reduction shutdown are implemented.

[0030] (6) Locate the fault location and type based on the combination of abnormal parameters, review the data after the fault is handled, optimize the weight and warning threshold, and form a closed-loop management.

[0031] This embodiment can reliably identify typical latent faults such as thyristor aging, hidden blockage of cooling pipes, insulation degradation, and deterioration of bridge arm current sharing, achieving early identification, hierarchical warning, and precise location, ensuring the long-term stable operation of the rectifier system.

[0032] Step 1: Construct a system of latent fault characteristic parameters Based on the structural components (rectifier transformer, thyristor module, rectifier control cabinet, smoothing reactor, cooling system, electrical connection components) and operating conditions of the DC submerged arc furnace rectifier system, characteristic parameters that can characterize latent electrical faults are selected to construct a latent fault characteristic parameter system. These characteristic parameters include electrical parameters, temperature parameters, insulation parameters, environmental parameters, and equipment status parameters, as detailed below: Electrical parameters include: three-phase voltage imbalance on the rectifier input side, DC voltage ripple coefficient on the rectifier output side, thyristor conduction angle deviation, bridge arm current sharing, and smoothing reactor current harmonic distortion rate. These parameters are used to detect latent anomalies in the electrical circuit (such as minor inter-turn short circuits, deteriorated current sharing, and abnormal conduction). Among these, the bridge arm current sharing is primarily used to monitor the deterioration of thyristor bridge arm current sharing, a major cause of latent faults, and to prevent heat loss accumulation due to abnormal current sharing. Temperature parameters: junction temperature of the thyristor module, winding temperature of the rectifier transformer, temperature of the smoothing reactor, temperature of the electrical connection point, and temperature difference between the inlet and outlet water of the cooling system. These parameters are used to detect latent overheating faults in the equipment (such as increased contact resistance, latent blockage of the cooling system, and aging of components). Among them, the junction temperature of the thyristor module is controlled below 65℃ to avoid damage to the components caused by the junction temperature exceeding 125℃. The temperature difference between the inlet and outlet water of the cooling system is used to monitor the latent blockage of the pipeline. Insulation parameters: insulation resistance of rectifier transformer, insulation resistance of thyristor module, insulation resistance of DC bus, and dielectric loss factor of cable insulation layer, used to detect latent faults such as insulation performance degradation; Environmental parameters: internal temperature, humidity, and dust concentration of the rectifier cabinet, used to detect hidden faults caused by environmental factors (such as insulation dampness, poor contact caused by dust accumulation, etc.). Equipment status parameters: thyristor trigger pulse amplitude deviation, cooling fan speed deviation, rectifier transformer noise amplitude, used to detect hidden anomalies in equipment operation.

[0033] Step 2: Feature Parameter Acquisition and Preprocessing Data acquisition devices are installed at key parts of the DC submerged arc furnace rectifier system to collect the characteristic parameters constructed in step 1 in real time, forming a raw parameter dataset. The data acquisition devices include voltage sensors, current sensors, temperature sensors, insulation detectors, temperature and humidity sensors, dust sensors, vibration sensors, and noise sensors. The acquisition frequency is set to 1~10Hz according to the parameter type to ensure that the gradual changes of latent faults can be captured. Among them, core components such as thyristor modules, rectifier transformers, and smoothing reactors adopt multi-point acquisition methods to improve the comprehensiveness of parameter acquisition.

[0034] The original parameter dataset is preprocessed to remove interference signals and outlier data, resulting in a standardized feature parameter dataset. The preprocessing steps include: Outlier removal: The 3σ criterion is used to remove extreme outliers in the original data (such as sudden data changes caused by sensor failure) to avoid outlier data affecting the accuracy of early warning. Noise reduction: Wavelet transform algorithm is used to reduce noise in parameters containing interference (such as current and voltage signals) and filter out interference signals caused by electromagnetic interference and fluctuations in smelting conditions. Standardization processing: The min-max standardization method is used to map feature parameters of different dimensions to the [0,1] interval, eliminating the influence of dimensions and facilitating subsequent feature fusion and fault identification; Data completion: For missing parameter data, linear interpolation is used to complete the data to ensure the integrity of the dataset.

[0035] Step 3: Fusion and Identification of Latent Fault Features An optimized fusion algorithm based on the analytic hierarchy process (AHP) is adopted to fuse the preprocessed standardized feature parameters to obtain comprehensive fault feature values. Combined with the Delphi expert meeting mechanism, the weights of each feature parameter are determined, and industry experience is fully incorporated to improve the rationality of feature fusion and avoid the problem of misjudgment by a single parameter.

[0036] The specific steps are as follows: Constructing the judgment matrix: Based on the importance of each feature parameter for the identification of latent faults, and combined with expert experience, construct a feature parameter weight judgment matrix; Consistency check and weight calculation: Perform a consistency check on the judgment matrix, adjust any judgment matrices that do not meet the requirements, and calculate the weight values ​​of each feature parameter; Feature fusion: By multiplying each standardized feature parameter by its corresponding weight and summing the results, a comprehensive fault feature value is obtained. The formula is as follows: ,in For comprehensive fault characteristic values, For the i-th standardized feature parameter, The weight of the i-th feature parameter, This represents the number of feature parameters. The comprehensive feature value T is used to determine the presence of a fault and classify the warning level, but not for fault location. After determining that a latent fault exists, the system retains a multi-dimensional anomaly vector of all feature parameters and inputs this multi-dimensional vector into the fault location model for fault type and location identification. That is, the single value T is responsible for screening and classification, while the multi-dimensional feature vector is responsible for location; the two functions are independent.

[0037] During fault type and location identification, the following are included: Initial fault diagnosis stage: All characteristic parameters of electrical, temperature, insulation, environment, and equipment status are weighted and integrated to ensure that no weak or hidden faults are missed; Fault location phase: Select corresponding feature parameter subsets according to different fault types to achieve accurate matching; Among them, all characteristic parameters such as bridge arm current equalization, three-phase voltage imbalance, and DC voltage ripple coefficient are calculated from the original sensor data according to existing standards and industry-standard formulas.

[0038] The analytic hierarchy process (AHP) in this solution uses the industry-standard Saaty 1-9 scale to construct the judgment matrix. This scale is a known, mature, and universal system, requiring no customization. Scoring is performed by a fixed expert group consisting of rectifier design experts, senior on-site maintenance engineers, and electrical fault early warning research experts. To address inconsistencies in scores from multiple experts: extreme outliers are first removed, and the remaining scores are then merged using a geometric mean to obtain a unified scale; next, a consistency check of the judgment matrix is ​​performed. If the check fails, the experts make minor adjustments until it meets the standard, ultimately resulting in unique and reproducible parameter weights.

[0039] A preset latent fault identification threshold is set, and the comprehensive fault characteristic value is compared with the identification threshold to determine whether there is a latent electrical fault in the rectifier system. If the comprehensive fault characteristic value is greater than the identification threshold, a latent fault is determined to exist; if the comprehensive fault characteristic value is less than or equal to the identification threshold, no latent fault is determined. The identification threshold is calibrated based on historical fault data, expert experience, and on-site operating conditions, and can be dynamically adjusted according to the operating status of the rectifier system.

[0040] In this scheme, a latent fault refers to an intermediate state in which the parameters of the rectifier system exceed the normal random fluctuation range but do not reach the conventional protection explicit fault tripping threshold. The relevant parameters deteriorate slowly and gradually without sudden changes. In the short term, it does not affect production operation, but if it continues to develop, it will lead to serious faults such as breakdown, short circuit, and shutdown. This is defined as a latent fault.

[0041] Normal range: The parameter fluctuates randomly within the range of ±3σ of the baseline value, without a monotonic degradation trend; Latent fault range: Parameters > 3σ upper limit and < explicit fault protection threshold, showing a slow unidirectional deviation trend; Obvious fault range: When parameters exceed the protection threshold, an alarm or tripping is directly triggered. Boundary quantification is clear and can be determined online.

[0042] This solution does not rely on manually labeled latent fault samples. Instead, it collects long-term normal steady-state operation data of the equipment, establishes normal reference ranges and fluctuation ranges for each parameter, and extracts the historical data of the entire process from latent slow deterioration to eventual manifest fault. The period of slow deviation before the fault trip and the period without triggering conventional protection are used as reference samples for latent faults. Based on the upper limit of 3σ fluctuation under normal operating conditions as the base threshold, the identification threshold is jointly determined by combining the inflection point of latent fault deviation and on-site operating condition calibration. The threshold is supported by measured data and can be reliably set without the need for manually labeled samples.

[0043] Step 4: Hidden Faults Stratification and Classification Based on the severity, development speed, and impact on the operation of the rectifier system of latent faults, and considering the continuous requirements of ferroalloy smelting, latent electrical faults are divided into three levels, each corresponding to a different fault grade and risk level, as detailed below: Level 1 Latent Fault (Mild Risk): In the early stage of the fault, only a single characteristic parameter shows a slight abnormality, and the comprehensive fault characteristic value is between the identification threshold and the Level 1 warning threshold. The fault has no obvious development trend and will not affect the normal operation of the rectifier system. Such as slight increase in contact resistance, slight deterioration of insulation performance, slight scaling of the cooling system, etc. If this type of fault is handled in time, further deterioration can be avoided. Level 2 latent fault (moderate risk): The fault has developed, multiple characteristic parameters are abnormal, and the comprehensive fault characteristic value is between the Level 1 and Level 2 warning thresholds. The fault shows a slow development trend, and long-term operation may affect the stability of the rectifier system, such as continuous increase in thyristor junction temperature, increased bridge arm flow rate deviation, and abnormal difference in inlet and outlet water temperatures of the cooling system. This type of fault requires close monitoring and timely intervention measures. Level 3 latent fault (severe risk): The fault has accumulated to a critical state, multiple characteristic parameters show obvious abnormalities, the comprehensive fault characteristic value is greater than the level 2 warning threshold, the fault development speed is accelerated, and it may cause a visible fault in the short term, leading to the shutdown of the rectifier system, such as severe aging of thyristors, slight short circuit between winding turns, and severe latent blockage of cooling pipes; such faults require immediate shutdown and handling to avoid the fault from expanding.

[0044] The first-level and second-level early warning thresholds are determined through historical fault data statistics, fault simulation tests, and expert demonstrations. They can be dynamically calibrated based on the operating years, maintenance conditions, and smelting conditions of the rectifier system to ensure the rationality and pertinence of the hierarchical classification.

[0045] In this scheme, the comprehensive characteristic value T is the overall deviation after multi-parameter normalization. The larger the value, the more serious the deviation of the equipment from normal operating conditions and the higher the risk of failure. Its physical meaning is clear. The thresholds for the three levels are pre-defined by statistical clustering partitioning of a large number of historical T values ​​for normal, slightly deteriorated, moderately deteriorated, and severely deteriorated conditions. Then, engineering fine-tuning is made in combination with smelting production conditions and equipment safety requirements to form the first, second, and third level early warning thresholds. The classification is based on sufficient evidence and is reasonable.

[0046] Step 5: Layered Early Warning and Response Based on the hierarchical results from step 4, the corresponding early warning mechanism is activated, and the corresponding response strategy is executed to achieve hierarchical control of latent faults, avoiding excessive or insufficient early warnings, as detailed below: Level 1 Warning (Mild Risk): The primary warning is activated by issuing a green warning through the rectifier system monitoring terminal, and simultaneously recording the fault characteristic parameters, occurrence time and location. No shutdown is required. Just focus on the corresponding parts during daily inspections, regularly check changes in characteristic parameters, and clean dust and replenish conductive paste in a timely manner to prevent the fault from developing. At the same time, the cooling system's Level 1 heat dissipation mode is activated to ensure stable equipment temperature. Level 2 Warning (Moderate Risk): A moderate warning is activated, issuing a yellow alert and simultaneously sending an alarm signal via audible and visual alarms. Warning information is also pushed to the mobile terminals of maintenance personnel. Maintenance personnel must conduct a comprehensive inspection of the warning area within 24 hours to investigate the cause of the fault and take targeted corrective measures (such as adjusting the thyristor trigger parameters, cleaning cooling pipes, and tightening electrical connections). Simultaneously, the secondary cooling mode of the cooling system is activated to increase heat dissipation, and changes in fault characteristic parameters are monitored in real time. If the parameters continue to deteriorate, the warning level is immediately upgraded. Level 3 Early Warning (Severe Risk): Activate the advanced early warning system, issue a red alert, continuously sound and light alarms, push emergency warning information to the terminals of operation and maintenance personnel and management personnel, immediately trigger the rectifier system to reduce load, complete the shutdown operation within 30 minutes, and allow operation and maintenance personnel to conduct a comprehensive investigation of the fault. Only after thorough repair and passing the test can the rectifier system be restarted. At the same time, record the entire fault process data to provide a basis for subsequent fault analysis and prevention.

[0047] Step 6: Fault Location and Post-mortem Optimization For identified latent faults, the location and type of fault are accurately determined by combining the abnormality of each characteristic parameter, the fault level, and the equipment layout. For example, if the junction temperature of the thyristor rises abnormally and the conduction angle deviation increases, while other parameters are normal, the fault is located as aging of the thyristor module or a latent fault in the trigger circuit. If the inlet and outlet water temperatures of the cooling system are abnormal and the junction temperature of the thyristor rises, the fault is located as a latent blockage fault in the cooling pipe. If the current sharing deviation of the bridge arm is too large, the fault is located as a problem with the installation quality of the thyristor module or an abnormality of the fast-fuse resistor.

[0048] This solution uses a fault knowledge graph and a fuzzy matching inference model to construct a fault location model, including the following steps: 1. Establish a knowledge base linking equipment components, fault types, and characteristic anomaly patterns, and create a dedicated fuzzy membership function for each type of latent fault; 2. Input a multidimensional feature anomaly vector, calculate the matching degree with various types of faults, and output the fault location and type according to the maximum matching degree; 3. When encountering multiple abnormal parameters and overlapping fault characteristics, the model automatically calculates the matching score of each candidate fault, sorts and outputs the main fault and secondary hidden dangers, and has the ability to decouple multiple faults, overcoming the limitations of simple "if...then..." rules; The entire model has complete reasoning logic and matching algorithms.

[0049] After the fault handling is completed, the fault data, early warning process, and handling results are reviewed to analyze the causes, development patterns, and shortcomings of the early warning mechanism of latent faults. The weights of feature parameters, early warning thresholds, and response strategies are optimized, and the latent fault feature database is updated to continuously improve the accuracy and reliability of early warnings. At the same time, the fault handling experience is incorporated into the operation and maintenance specifications to prevent similar latent faults from recurring.

[0050] To further improve the accuracy of early warning and the applicability of the system, the following optimizations will be made: In step 3, the feature fusion algorithm can adopt a hybrid algorithm combining the analytic hierarchy process (AHP) and the backpropagation (BP) neural network. By utilizing the self-learning capability of the BP neural network and combining it with the weight allocation advantage of the AHP, the algorithm can automatically learn the characteristic patterns of latent faults, improve the calculation accuracy of the comprehensive fault feature value, and adapt to the latent fault identification needs under different working conditions.

[0051] In step 5, the response strategy can be dynamically adjusted according to the smelting conditions of the DC submerged arc furnace (such as smelting load and ore type). For example, during the high-load smelting stage, the warning threshold can be appropriately lowered to improve the warning sensitivity and avoid hidden faults from affecting the smelting stability; during the low-load stage, the warning threshold can be appropriately increased to reduce false alarms.

[0052] This solution constructs a multi-dimensional feature parameter system covering electrical, temperature, insulation, environment, and equipment status. Combined with an optimization fusion algorithm based on the analytic hierarchy process, it can capture the gradual changes of latent faults, accurately distinguish latent faults from operating condition fluctuations, and effectively identify various latent electrical faults such as thyristor aging, increased contact resistance, latent blockage of cooling pipes, insulation degradation, and deterioration of bridge arm current sharing. It avoids the accumulation and deterioration of latent faults leading to manifest faults, significantly improves the accuracy of fault identification, and is suitable for the high-load and strong-interference operating conditions of DC submerged arc furnaces. It makes up for the shortcomings of existing technologies that can only identify manifest faults and reduces downtime losses caused by latent faults.

[0053] Latent faults are categorized into three levels, each with a corresponding early warning method and response strategy. This avoids false alarms, over-warnings, or missed alarms, minimizing production interruptions and improving the continuity of ferroalloy smelting while ensuring the safe operation of the rectifier system. Minor faults do not require shutdown and only need routine inspection; moderate faults require timely detection and handling; and severe faults require emergency shutdown. This approach enables tiered management of fault risks, balancing production efficiency and equipment safety, and addresses the issues of poor targeting and lack of tiered management in existing technologies.

[0054] By combining abnormal characteristic parameters, fault level, and equipment layout, it can quickly and accurately locate the specific location and type of latent faults, generate fault location reports and emergency response plans, guide maintenance personnel to quickly troubleshoot and handle faults, shorten fault handling time, and reduce maintenance costs. In particular, it can achieve precise location for parts that are prone to latent faults, such as thyristor components and cooling systems, solving the problems of vague fault location and low maintenance efficiency in existing technologies.

[0055] Through the data storage and review optimization module, the entire fault process data is recorded, the causes of the fault and the deficiencies in early warning are reviewed and analyzed, the weights of characteristic parameters, early warning thresholds and response strategies are optimized, and the operating status of the rectifier system and the changes in smelting conditions are dynamically adapted to continuously improve the accuracy and reliability of early warning. This forms a closed-loop management of "collection-identification-early warning-processing-review-optimization", and at the same time provides data support for the subsequent optimization design and operation and maintenance of the rectifier system, reduces the recurrence of similar hidden faults, extends the service life of equipment, further reduces production and operation and maintenance costs, adapts to the long-term stable production needs of the ferroalloy industry, and can continuously improve the accuracy of early warning by referring to existing technologies.

[0056] After fault handling is completed, a case library is established by collecting data from the entire fault process, early warning records, location results, and actual on-site fault conditions. With the optimization goals of reducing false alarm rates and missed alarm rates, and improving location accuracy, a clear objective function is constructed. An intelligent iterative self-learning optimization algorithm is used to optimize parameters such as AHP weights, identification thresholds, three-level early warning thresholds, and knowledge graph matching weights. After optimization, the algorithm is automatically updated to the system model, forming a closed loop of "collection-early warning-location-handling-review-self-optimization." The optimization algorithm, optimization object, and optimization goal are all clearly defined and achievable.

[0057] Designed specifically for the structural characteristics and operating conditions of DC submerged arc furnace rectifier systems used in ferroalloy smelting, this system takes into account the high-voltage, high-current, and strong electromagnetic interference operating environment. The sensors employ an anti-electromagnetic interference design, and the early warning strategy can be dynamically adjusted according to smelting conditions. It can be directly applied to the upgrade and renovation of existing DC submerged arc furnace rectifier systems without large-scale equipment modifications, resulting in low upgrade costs and easy promotion and application. Simultaneously, the system can be linked with existing central control systems to achieve collaborative management, further enhancing its practicality. It is adaptable to DC submerged arc furnace rectifier systems of different scales and types, with broad application prospects. It can be widely used in the early warning and control of latent faults in DC submerged arc furnace rectifier systems in the ferroalloy smelting industry, possessing significant economic and social benefits. It can reduce economic losses caused by faults, improve production efficiency and product quality, while reducing equipment wear and achieving energy conservation and consumption reduction goals, meeting the green, efficient, and stable development needs of the ferroalloy industry.

[0058] Correspondingly, a fault early warning system for DC submerged arc furnace rectifier, such as Figure 2 As shown, the system includes a data acquisition module, a data preprocessing module, a feature fusion and fault identification module, a hierarchical classification module, an early warning and response module, a fault location module, and a data storage and review optimization module. These modules are electrically connected in sequence to collaboratively achieve accurate identification, hierarchical early warning, location, and optimization of latent faults, as detailed below: (1) Data acquisition module Composed of multiple different types of sensors and data acquisition terminals, these devices are installed in key parts of the DC submerged arc furnace rectifier system to collect various parameters in the latent fault characteristic parameter system in real time, forming a raw parameter dataset. The sensors include voltage sensors (installed on the rectifier input and output sides), current sensors (installed on the bridge arm and smoothing reactor circuit), temperature sensors (installed on the thyristor module, rectifier transformer, electrical connection points, and cooling system inlet and outlet), insulation detectors (connected to the rectifier transformer, thyristor module, and DC bus), temperature and humidity sensors, dust sensors (installed inside the rectifier cabinet), vibration sensors, and noise sensors (installed on the rectifier transformer and smoothing reactor). The data acquisition terminals are used to aggregate the data collected by various sensors and transmit it to the data preprocessing module. Core components such as the thyristor assembly and rectifier transformer employ multi-point sensor arrangements to ensure the comprehensiveness and accuracy of parameter acquisition, adapting to the high-load operating conditions of the DC submerged arc furnace.

[0059] In the data acquisition module, the sensor adopts an anti-electromagnetic interference design to adapt to the strong electromagnetic interference operating environment of the DC submerged arc furnace and ensure the accuracy of parameter acquisition. At the same time, the sensor has a self-diagnostic function, which can monitor its own operating status in real time. If a sensor failure occurs, it will issue a prompt in time to avoid false alarms caused by sensor failure. (2) Data preprocessing module Electrically connected to the data acquisition module, it receives the raw parameter dataset transmitted by the data acquisition module, performs preprocessing operations such as outlier removal, noise reduction, standardization, and data completion to eliminate interference signals and abnormal data, obtains a standardized feature parameter dataset, and transmits it to the feature fusion and fault identification module. This module uses a high-speed data processing chip to ensure the real-time performance of data preprocessing and adapts to the high-frequency parameter acquisition requirements of the rectifier system.

[0060] (3) Feature fusion and fault identification module Electrically connected to the data preprocessing module, it incorporates an optimization fusion algorithm based on the analytic hierarchy process (AHP) and a Delphi expert weight determination model. This algorithm is used to perform feature fusion on a standardized feature parameter dataset and calculate the comprehensive fault feature value. A preset latent fault identification threshold is set, and the comprehensive fault feature value is compared with the threshold to determine if a latent electrical fault exists in the rectifier system. If a latent fault exists, the comprehensive fault feature value and abnormal data of each feature parameter are transmitted to the hierarchical module. If no latent fault exists, the parameter data is transmitted to the data storage module. This module can dynamically adjust the feature parameter weights based on historical data and on-site operating conditions to improve fault identification accuracy and avoid misjudgment based on a single parameter.

[0061] (4) Hierarchical and graded modules Electrically connected to the feature fusion and fault identification module, it has built-in first-level and second-level early warning thresholds. It is used to receive comprehensive fault feature values ​​and abnormal data of various feature parameters. Based on the range of comprehensive fault feature values, the hidden faults are divided into three levels: first-level, second-level, and third-level, to determine the degree of fault risk. The hierarchical classification results are then transmitted to the early warning and response module. This module supports dynamic threshold calibration and can adjust the early warning thresholds in real time according to the rectifier system's operating years, maintenance status, and changes in smelting conditions to adapt to the early warning needs under different operating conditions.

[0062] (5) Early warning and response module Electrically connected to the hierarchical module, it includes an early warning unit, an audible and visual alarm unit, an information push unit, and a response control unit. The early warning unit displays different levels of early warning prompts (green, yellow, and red) on the monitoring terminal. The audible and visual alarm unit issues corresponding audible and visual alarm signals for level two and three warnings. The information push unit pushes early warning information (fault level, fault location, and abnormal parameters) to the terminals of maintenance and management personnel. The response control unit triggers corresponding response strategies (daily inspection, equipment testing, and load reduction shutdown) based on the fault level and controls the operating status of the cooling system and rectifier system to achieve hierarchical fault management and prevent fault escalation or excessive shutdown from affecting production efficiency. This module can be linked with the rectifier system's control system to automatically execute response strategies and improve the timeliness of early warning responses.

[0063] In the early warning and response module, a fault emergency handling plan push function has been added. For different types and levels of hidden faults, corresponding emergency handling steps are pushed to guide operation and maintenance personnel to quickly handle faults and shorten fault handling time. (6) Fault location module Electrically connected to the early warning and response module and the feature fusion and fault identification module, this module receives fault stratification and classification results and abnormal data of various feature parameters. Combined with the equipment layout of the DC submerged arc furnace rectifier system, it accurately locates the specific location and type of hidden faults, generates a fault location report, and transmits it to the maintenance personnel's terminal, providing a basis for fault investigation and handling. This module has a built-in equipment layout database and fault feature matching model, which can quickly match abnormal parameters with fault types and fault locations, improve fault location efficiency, solve the problem of ambiguous fault location in existing technologies, and adapt to the rapid maintenance needs of ferroalloy smelting sites.

[0064] (7) Data storage and review optimization module Electrically connected to all the aforementioned modules, this module stores raw parameter datasets, preprocessed parameter data, comprehensive fault characteristic values, fault stratification and classification results, early warning records, fault location reports, and fault handling records, constructing a latent fault characteristic database. Simultaneously, it performs post-mortem analysis of fault data, early warning processes, and handling results, optimizing characteristic parameter weights, early warning thresholds, and response strategies. The optimized parameters are then updated to the corresponding modules, continuously improving the accuracy and reliability of the early warning system. This module supports data export, facilitating fault analysis and technical summarization by maintenance personnel, and providing data support for subsequent optimization design of the rectifier system. Furthermore, this module can record data related to the installation and maintenance of thyristor components, providing a reference for daily maintenance and reducing latent faults caused by improper installation.

[0065] The system can be linked with the central control system of DC submerged arc furnace to acquire smelting condition data in real time. By combining the condition data with the optimization of early warning strategies, the system can achieve coordinated control of latent fault early warning and smelting condition, thereby further improving the operational stability of the rectifier system.

[0066] Correspondingly, a computer program product includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the aforementioned fault warning method for DC submerged arc furnace rectification.

[0067] Correspondingly, an electronic device includes at least one processor and a memory connected to the processor, the memory being used to store computer programs; the processor is used to execute the computer programs, enabling the electronic device to implement the aforementioned fault early warning method for DC submerged arc furnace rectification.

[0068] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A fault early warning method for the rectification of a DC submerged arc furnace, characterized in that, Includes the following steps: The electrical, temperature, and insulation parameters of multiple monitoring locations in the DC submerged arc furnace rectifier system are acquired to construct a multi-dimensional feature parameter set. The multi-dimensional feature parameter set is preprocessed to obtain a standardized feature parameter set; The standardized feature parameter set is input into the preset fault location model. The fault location model is constructed based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal mode, and outputs the fault type and fault location of the latent fault. Based on the fault type and location, and combined with the preset fault level judgment rules, the warning level of the latent fault is determined, and the response strategy corresponding to the warning level is executed.

2. The fault early warning method for DC submerged arc furnace rectification as described in claim 1, characterized in that, The multi-dimensional feature parameter set also includes environmental parameters and equipment status parameters; Electrical parameters include at least one of the following: three-phase voltage imbalance on the rectifier input side, DC voltage ripple coefficient on the rectifier output side, thyristor conduction angle deviation, bridge arm current sharing rate, and smoothing reactor current harmonic distortion rate. Temperature parameters include at least one of the following: junction temperature of the thyristor module, winding temperature of the rectifier transformer, temperature of the smoothing reactor, temperature of the electrical connection point, and temperature difference between the inlet and outlet water of the cooling system. Insulation parameters include at least one of the following: insulation resistance of rectifier transformer, insulation resistance of thyristor module, insulation resistance of DC bus, and dielectric loss factor of cable insulation layer; Environmental parameters include at least one of the following: internal temperature of the rectifier cabinet, humidity, and dust concentration; The equipment status parameters include at least one of the following: thyristor trigger pulse amplitude deviation, cooling fan speed deviation, and rectifier transformer noise amplitude.

3. The fault early warning method for DC submerged arc furnace rectifier as described in claim 1, characterized in that, Preprocessing includes: The 3σ criterion was used to remove outliers from the original data. Wavelet transform algorithm is used to denoise the data after outlier removal; The min-max normalization method is used to map the denoised data to a set range; Linear interpolation was used to fill in missing values ​​in the standardized data.

4. The fault early warning method for DC submerged arc furnace rectifier as described in claim 1, characterized in that, Before inputting the standardized feature parameter set into the preset fault location model, a feature fusion step is also included: The weights of each standardized characteristic parameter are determined by the analytic hierarchy process (AHP), and the comprehensive fault characteristic value is calculated by weighted summation. The comprehensive fault feature value is compared with the preset identification threshold. If the comprehensive fault feature value is greater than the identification threshold, it is determined that there is a hidden fault, and the subsequent steps of inputting into the preset fault location model are executed.

5. The fault early warning method for DC submerged arc furnace rectifier as described in claim 1, characterized in that, The fault level determination rules include: The comprehensive fault characteristic value is compared with the preset first-level warning threshold and second-level warning threshold; If the comprehensive fault characteristic value is greater than the identification threshold and less than or equal to the first-level warning threshold, it is determined to be a first-level warning; If the comprehensive fault characteristic value is greater than the first-level warning threshold and less than or equal to the second-level warning threshold, it is determined to be a second-level warning; If the comprehensive fault characteristic value is greater than the level 2 warning threshold, it is determined to be a level 3 warning.

6. The fault early warning method for DC submerged arc furnace rectifier as described in claim 1, characterized in that, Implement the response strategy corresponding to the warning level, including: If the warning level is Level 1, then a green warning will be issued, and the corresponding area will be given special attention during routine inspections. If the warning level is Level II, a yellow warning will be issued, triggering an audible and visual alarm and a terminal information push, requiring a comprehensive inspection of the warning area within a specified time. If the warning level is Level 3, a red warning will be issued and an emergency shutdown will be triggered to conduct a comprehensive investigation and handling of the fault.

7. The fault early warning method for DC submerged arc furnace rectifier as described in claim 1, characterized in that, The fault location model outputs the fault type and location of latent faults based on the abnormality of each characteristic parameter, the fault level, and the equipment layout of the rectifier system. After the fault is handled, the fault data, early warning process, and handling results are reviewed, and the mapping relationship of the fault location model is updated, the fault level judgment rules are adjusted, and the corresponding response strategies are implemented.

8. A fault early warning system for DC submerged arc furnace rectifier, characterized in that, include: The data acquisition unit is configured to acquire electrical parameters, temperature parameters, and insulation parameters at multiple monitoring locations in the DC submerged arc furnace rectifier system, and construct a multi-dimensional feature parameter set. The data preprocessing unit is configured to preprocess the multi-dimensional feature parameter set to obtain a standardized feature parameter set. The fault location unit is configured to: input a standardized set of feature parameters into a preset fault location model, which is constructed based on the mapping relationship between the fault characteristics of each component of the rectifier system and the multi-parameter abnormal mode, and output the fault type and fault location of the latent fault. The fault early warning unit is configured to determine the early warning level of a latent fault based on the fault type and fault location, combined with preset fault level judgment rules, and execute the response strategy corresponding to the early warning level.

9. A computer program product, characterized in that, It includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to perform the steps in the fault warning method for DC submerged arc furnace rectification as described in any one of claims 1-7.

10. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, the memory being used to store computer programs; the processor is used to execute the computer programs, enabling the electronic device to perform the steps in the fault early warning method for DC submerged arc furnace rectification as described in any one of claims 1-7.

Images