A method, system and device for early warning of auxiliary equipment of a hydropower station and a storage medium

Abstract

The application provides a kind of hydroelectric station auxiliary equipment early warning method, system, equipment and storage medium, belong to hydroelectric station early warning technical field, including through sensor acquisition hydroelectric station in auxiliary equipment's pressure, vibration, current, flow and equipment surface temperature data, and collect auxiliary equipment related pipeline's working condition parameter and environmental parameter, will be collected data according to working condition parameter and environmental parameter generates real-time working condition label, the working condition label at least contains unit load grade, environmental temperature interval and equipment running time length section, will real-time working condition label establish structured threshold parameter library, built-in threshold adjustment model based on optimization XGBoost algorithm, output corresponding working condition under each working condition parameter Dynamic safety interval.The technical scheme of the application aims to realize the double reduction of hydroelectric station auxiliary equipment early warning false alarm rate and missed alarm rate, and improve the operation and maintenance efficiency and equipment safety operation guarantee ability.

Description

Technical Field

[0001] This invention relates to the field of early warning technology for hydropower stations, specifically to an early warning method, system, equipment, and storage medium for auxiliary equipment in hydropower stations. Background Technology

[0002] Auxiliary equipment in hydropower stations (such as cooling water system pumps, lubrication system oil pumps, and compressed air system air compressors) are the "nerve endings" ensuring the safe and stable operation of the generating units. Their failure can lead to the shutdown of main equipment, causing significant economic losses. While early warning technology for auxiliary equipment in hydropower stations has reached a certain level of sophistication, many technical bottlenecks still exist in practical applications, making it difficult to meet the requirements for high-precision and high-reliability early warning. Specific problems are as follows: Current technologies for monitoring and early warning of auxiliary equipment in hydropower stations mostly rely on fixed thresholds or designated safety ranges. However, these thresholds cannot adapt to load changes and environmental fluctuations. For example, rising cooling water temperatures in summer can lead to false alarms, and gradual faults, such as filter blockage causing a slow decrease in flow, may go undetected. Moreover, a single monitoring dimension cannot comprehensively reflect the equipment status, easily resulting in "missed alarms." For instance, a pipeline leak may only manifest as a pressure drop without a significant temperature change. These issues directly lead to a persistently high false alarm rate, severely impacting operational efficiency and the reliability of equipment safety monitoring. Summary of the Invention

[0003] The main objective of this invention is to provide an early warning method for auxiliary equipment in hydropower stations, which aims to reduce both the false alarm rate and the missed alarm rate of auxiliary equipment in hydropower stations, and improve operation and maintenance efficiency and the ability to ensure safe operation of equipment.

[0004] To achieve the above-mentioned objectives, the first aspect of the present invention provides a method for early warning of auxiliary equipment in hydropower stations: Includes the following steps: The system collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment within the hydropower station using sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The collected data is used to generate real-time operating condition tags based on operating condition parameters and environmental parameters. The operating condition tags include at least the unit load level, ambient temperature range, and equipment operating duration. A structured threshold parameter library is established using real-time operating condition labels, and a threshold adjustment model based on the optimized XGBoost algorithm is built in to output the dynamic safety range of each operating condition parameter under the corresponding operating condition. Monitor the changing trends of various operating parameters, and generate a trend warning signal when the parameter drifts towards the boundary of the corresponding dynamic safety range for N consecutive cycles. The credibility of the trend warning signals corresponding to each working condition parameter is fused using DS evidence theory or weighted voting mechanism, and the final warning result is output in the form of visualization or audible and visual alarm.

[0005] In one possible implementation, the collected data is preprocessed, the preprocessing including: The Kalman filter algorithm is used to filter out random noise in pressure and flow parameters, and the 3σ criterion is used to remove abnormal jump values ​​in temperature and vibration parameters. Missing data are filled in using adjacent period linear interpolation.

[0006] In one possible implementation, the rules for generating real-time operating condition labels include: Unit load levels are classified according to the percentage of rated load; Ambient temperature ranges are divided according to climatic characteristics; The operating time of equipment is divided according to its life cycle; The label format is uniformly set to "load level-temperature range-running duration".

[0007] In one possible implementation, the structured threshold parameter library construction process includes: Collect historical operating data of auxiliary equipment for a preset year, covering normal data and labeled fault data under different operating condition labels; The distribution characteristics of each parameter are statistically analyzed according to the working condition label, and the pre-set information interval is used as the basic safety range to eliminate fault abnormal values ​​within the interval. Each tag-parameter entry is associated with a data source, a verification accuracy greater than a preset value, and an update timestamp. It supports quick retrieval by tag and presets the corresponding time.

[0008] In one possible implementation, the optimization method of the optimized XGBoost algorithm includes: A coupling coefficient for operating conditions is introduced, which is calculated using the Pearson correlation coefficient between unit load and ambient temperature, reflecting the interaction between the two on the parameters; The learning rate, preset tree depth, and preset number of leaf nodes of XGBoost were optimized using the particle swarm optimization algorithm, and the model prediction error was less than the preset value. An incremental learning mechanism is built, which fine-tunes the model monthly using new data, eliminating the need for full retraining.

[0009] In one possible implementation, the output logic of the dynamic security interval includes: First, the parameter library is called to output the basic safe range of the corresponding label. Then, the XGBoost model is optimized and combined with the current parameter fluctuation variance to dynamically correct the basic range value. The correction magnitude is positively correlated with the variance of parameter fluctuations, and the final interval must cover the normal operating data of the preset value.

[0010] In one possible implementation, the trend warning signal includes: When a parameter directly exceeds the dynamic safety range, an over-limit warning signal is generated, which has a higher priority than a trend warning signal for the same parameter. Over-limit warnings must specify the over-limit amplitude, i.e., the absolute value, the proportion of the interval width, and the duration of the over-limit.

[0011] A hydropower station auxiliary equipment early warning system, comprising: The multi-sensor acquisition module collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment in the hydropower station through sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The operating condition label generation module generates real-time operating condition labels from the collected data based on operating condition parameters and environmental parameters. The operating condition labels include at least the unit load level, ambient temperature range, and equipment operating duration. The threshold parameter library construction module establishes a structured threshold parameter library based on real-time operating condition labels, and has a built-in threshold adjustment model based on the optimized XGBoost algorithm to output the dynamic safety range of each operating condition parameter under the corresponding operating condition. The early warning signal generation module monitors the changing trends of various operating parameters. When a parameter drifts towards the boundary of the corresponding dynamic safety range for N consecutive cycles, it generates a trend early warning signal. The early warning result generation module uses DS evidence theory or weighted voting mechanism to fuse the credibility of trend early warning signals corresponding to each working condition parameter, outputs the final early warning result, and finally outputs it in the form of visualization or audible and visual alarm.

[0012] A computer device, comprising: A memory that stores computer-readable instructions; A processor, which, when executing the computer-readable instructions, implements the steps of the early warning method for auxiliary equipment of a hydropower station as described above.

[0013] A computer-readable storage medium storing computer-readable instructions, which, when executed by a processor, implement the steps of the above-described early warning method for auxiliary equipment of a hydropower station.

[0014] Beneficial effects of this invention: This invention utilizes a multi-sensor acquisition module to comprehensively depict the full-scenario operation of auxiliary equipment through a multi-dimensional acquisition system covering core parameters of auxiliary equipment (pressure, vibration, current, flow rate, equipment surface temperature), pipeline operating parameters, and environmental parameters. This avoids the problem of missed hidden faults caused by single-dimensional monitoring and provides comprehensive and high-quality basic data support for subsequent analysis. The operating condition label module generates accurate real-time operating condition labels based on unit load level, ambient temperature range, and equipment operating duration, clearly defining different operating scenarios and providing a scientific, scenario-based basis for dynamic threshold adjustment, solving the core pain point of mismatch between traditional fixed thresholds and operating condition changes. The threshold parameter library construction module, through structured storage of operating condition-threshold mapping relationships and a built-in optimized XGBoost algorithm threshold adjustment model, outputs a dynamic safety range adapted to real-time operating conditions, inheriting the statistical reliability of historical data while also providing... The system responds to the coupling effect of parameter fluctuations and operating conditions, significantly reducing the risk of false alarms caused by load fluctuations and environmental changes. The early warning signal generation module accurately captures the precursors of gradual faults such as filter blockage and bearing wear by monitoring the drift trend of parameters towards the boundary of the dynamic safety range over N consecutive cycles, achieving an upgrade from "passive over-limit alarm" to "active trend early warning," allowing maintenance personnel sufficient time to handle the situation. The early warning result issuance module uses DS evidence theory or a weighted voting mechanism to fuse the credibility of multi-parameter early warning signals, effectively filtering out misjudgments caused by single parameter interference and improving the reliability of early warning decisions. At the same time, it outputs early warning results in the form of visualization and hierarchical audible and visual alarms, clarifying the fault level, abnormal parameters, and investigation direction, significantly shortening the fault location and handling time. Ultimately, this achieves a dual reduction in the false alarm rate and missed alarm rate of auxiliary equipment early warnings in hydropower stations, and a comprehensive improvement in maintenance efficiency and equipment safety assurance capabilities. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the method in Example 1; Figure 2 Here is a flowchart of the early warning method in Example 1; Figure 3 This is a flowchart of the working condition labeling process for Example 1; Figure 4 Flowchart for constructing the structured threshold parameter library in Example 1; Figure 5The flowchart shows the optimization method of the XGBoost algorithm in Example 1. Figure 6 This is a flowchart of the output logic of the dynamic security zone in Example 1; Figure 7 Here is a flowchart of the parameter monitoring process for Example 1; Figure 8 This is a system structure block diagram of Example 2; Figure 9 This is a block diagram of the computer device structure in Example 3. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] Example 1 This invention proposes an early warning method for auxiliary equipment in hydropower stations; Reference Figures 1 to 7 This includes the following steps: The system collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment within the hydropower station using sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The collected data is used to generate real-time operating condition tags based on operating and environmental parameters. The operating condition tags include at least the unit load level, ambient temperature range, and equipment operating duration.

[0019] A structured threshold parameter library is established using real-time operating condition labels, and a threshold adjustment model based on the optimized XGBoost algorithm is built in to output the dynamic safety range of each operating condition parameter under the corresponding operating condition.

[0020] The credibility of the trend warning signals corresponding to each working condition parameter is fused using DS evidence theory or weighted voting mechanism, and the final warning result is output in the form of visualization or audible and visual alarm.

[0021] The DS evidence theory or weighted voting mechanism is used to fuse multi-parameter early warning signals, assigning weights based on parameter monitoring accuracy and fault correlation (e.g., vibration 0.3, pressure 0.25, current 0.2), and combining this with the credibility of the early warning signal to calculate a comprehensive decision result. This mechanism can effectively distinguish between "true drift" and "disruptive fluctuations"—for example, when vibration parameters drift in a single cycle due to pipeline resonance, the comprehensive credibility does not reach the threshold and no early warning is triggered; however, when vibration, pressure, and flow all drift simultaneously, the comprehensive credibility is significantly improved, and an early warning is triggered quickly. This improves multi-parameter fusion, increases early warning accuracy, and reduces the false alarm rate caused by single-parameter interference.

[0022] refer to Figure 2 The collected data will be preprocessed. The core logic of the adjacent period linear interpolation method is: when the collected data of a certain period is missing due to outlier removal, such as the jump value removed by the 3σ criterion or the temporary failure of the sensor, the parameter values ​​of the two adjacent valid periods before and after the data point are used to calculate the estimated value of the missing data through linear fitting, so as to ensure the continuity and integrity of the data sequence.

[0023] Select the collected data of a certain working condition parameter (such as vibration amplitude) for M consecutive cycles as the sample set (M≥50, to ensure that the data conforms to the normal distribution assumption, M is taken as 100 cycles by default in this invention).

[0024] Mean μ: The arithmetic mean of all data in the sample set (reflecting the central tendency of the data).

[0025] ; xi represents the data collected in the i-th cycle; In outlier detection and handling, for each data point x collected in real time, the absolute value of its deviation from the mean μ, |x-μ|, is calculated. If |x-μ|>3σ, the data is determined to be an abnormal jump value, such as the instantaneous peak value generated by the vibration sensor due to a collision, and this data is directly discarded. Then, the data gaps caused by the discarding are filled in by linear interpolation of adjacent periods to ensure data continuity.

[0026] The data preprocessing process effectively filters noise, removes outliers, and completes missing data, providing high-quality data support for subsequent analysis and thus improving data reliability.

[0027] refer to Figure 3 Real-time operating condition label generation rules.

[0028] This rule clearly divides the load into four levels based on the rated load percentage: "Low (0-30%), Medium (30%-70%), High (70%-100%), and Rated (100%±5%)", accurately capturing the parameter variation characteristics of different load ranges. Compared to the existing technology's coarse division that "only distinguishes between high and low loads," it can reduce the matching error of the safe range of load-related parameters, providing a core scenario basis for dynamic threshold adjustment. The climatic characteristic division into "Low Temperature (≤0℃), Normal Temperature (0-35℃), and High Temperature (>35℃)" allows the threshold model to specifically correct parameter deviations caused by ambient temperature. This prevents environmental interference from being misjudged as equipment failure. The life cycle is divided into three segments, allowing the threshold parameter library to be updated with safe ranges related to equipment aging stages, avoiding the defect of using new equipment thresholds to judge the condition of old equipment.

[0029] The label format "load level-temperature range-running time segment" provides a unified data interface for multiple modules: the multi-sensor acquisition module can store parameters by label classification, and the threshold parameter library can quickly retrieve the corresponding basic safety range by label. The standardized standard definition format improves modular data interaction and retrieval, thereby enabling faster processing efficiency of the retrieved modules. This allows for adaptation to real-time early warning, making the issuance of early warning signals faster and more accurate. Under the iteration of a large amount of real-time data, modular data processing can remain stable.

[0030] This generation rule, through "precise segmentation + standard format," upgrades the early warning system from "static threshold judgment" to "scenario-based intelligent early warning." Using this rule, the identification of load fluctuations, environmental temperature variables, and gradual faults in new and old equipment can be effectively improved. At the same time, standardized operating condition labels reduce the training costs for operation and maintenance personnel, and enable faster confirmation of the fault location by combining early warning signals, as well as tracing the cause and cycle of the fault. This provides core technical support for the "safe operation and efficient maintenance" of hydropower station auxiliary equipment.

[0031] refer to Figure 4 , Construction of a structured threshold parameter library.

[0032] After classifying and statistically analyzing the distribution characteristics of each parameter according to the operating condition labels and calculating the preset confidence intervals, the core operation of "removing fault outliers within the intervals" is to perform targeted cleaning based on the labeled fault data labels. The specific process is as follows: First, retrieve the original labeling information (normal data / fault data) for all data falling within the preset confidence interval. Second, only remove samples labeled as "fault data" (this type of data is mostly pseudo-normal data caused by latent faults or sensor anomalies, such as minor equipment leakage but parameters not exceeding the statistical interval). Finally, retain all samples labeled as "normal data" within the interval, recalculate the preset confidence interval, and form the final "clean" basic safety range.

[0033] In this embodiment, the verification accuracy specifically refers to the fault differentiation accuracy of the "label-parameter basic safety range". Its calculation logic and verification method are as follows: 1. Formula definition: Validation accuracy = (Number of normal data samples within the interval + Number of faulty data samples outside the interval) / Total number of validation samples × 100%; 2. Validation dataset: Select historical data from the same period that have no overlap with the modeling data (accounting for 20% of the total historical data), including normal samples and typical fault samples (such as parameter samples corresponding to bearing wear and pipeline leakage) for each parameter under this label. 3. Judgment rules: If a normal sample falls within the basic safety range and a faulty sample exceeds the basic safety range, it is judged as "accurately distinguished"; otherwise, it is judged as "misclassified". 4. Compliance Requirements: For each "Label-Parameter" entry, the verification accuracy must be ≥ the preset value. If the accuracy is not met, the amount of historical data under that label must be expanded, the confidence interval must be recalculated and verified until the accuracy requirement is met.

[0034] It is worth noting that due to the uncertainty and scarcity of extreme weather (such as continuous heavy rainstorms not seen in historical preset years causing ambient humidity to exceed warning values, or short-term sharp cooling causing temperatures to drop sharply within 2 hours exceeding warning values), when such unconventional extreme weather occurs, it is necessary to first quantitatively compare the key indicators of the extreme weather (such as extreme temperature values, duration, and rate of temperature and humidity change) with data from similar extreme weather events in previous years. If the amplitude of the core indicator is less than the preset fluctuation value, the basic safety range corresponding to the historical similar operating condition label is used, and dynamic correction is performed by optimizing the operating condition coupling coefficient of the XGBoost model. If the amplitude is greater than the preset fluctuation value, an emergency data supplementation mechanism is activated—the operating condition parameters of the equipment under the extreme condition are collected in real time and stored in batches of 5 minutes as a "temporary emergency dataset." At the same time, the basic safety range of adjacent operating condition labels in the parameter library, such as "low load - low temperature - stable period," is called, combined with the equipment's design rated parameters, such as the allowable operating pressure range under extreme temperatures provided by the manufacturer, to set temporary dynamic thresholds to avoid warning failure due to data loss.

[0035] Furthermore, when unconventional extreme weather lasts for ≥2 hours, the system automatically triggers an incremental learning process: merging the "temporary emergency dataset" with historical data, removing faulty and outlier values, and adding a dedicated operating condition label for "load level - unconventional extreme temperature range - operating duration." The system recalculates the preset confidence intervals for each parameter under this label as the basic safety range, verifies the accuracy and update timestamps of recent extreme weather data sources, and supplements it to the structured threshold parameter library. This not only solves the problem of missing thresholds caused by the scarcity of extreme weather data, but also continuously enriches the operating condition coverage of the parameter library through real-time data supplementation and incremental learning. This ensures that when similar extreme weather occurs again, accurate thresholds can be directly matched, guaranteeing the continuity and reliability of early warnings and completely avoiding the technical blind spot of having no data reference for extreme operating conditions.

[0036] After classifying and statistically analyzing the distribution characteristics of each parameter according to the operating condition labels and calculating the preset confidence intervals, the core operation of "removing fault outliers within the intervals" is to perform targeted cleaning based on the labeled fault data labels. The specific process is as follows: First, retrieve the original labeling information (normal data / fault data) for all data falling within the preset confidence interval. Second, only remove samples labeled as "fault data" (this type of data is mostly pseudo-normal data caused by latent faults or sensor anomalies, such as minor equipment leakage but parameters not exceeding the statistical interval). Finally, retain all samples labeled as "normal data" within the interval, recalculate the preset confidence interval, and form the final "clean" basic safety range.

[0037] This operation does not remove normal outliers within the range, but rather filters contaminated data through fault labels, ensuring that the baseline is constructed solely from real, normal operating data. This avoids hidden fault data skewing the threshold range and improves the reliability of the baseline. In this solution, the verification accuracy specifically refers to the fault differentiation accuracy of the "label-parameter basic safety range," and its calculation logic and verification method are as follows: 1. Formula definition: Validation accuracy = (Number of normal data samples within the interval + Number of faulty data samples outside the interval) / Total number of validation samples × 100%; 2. Validation dataset: Select historical data from the same period that have no overlap with the modeling data (accounting for 20% of the total historical data), including normal samples and typical fault samples (such as parameter samples corresponding to bearing wear and pipeline leakage) for each parameter under this label. 3. Judgment rules: If a normal sample falls within the basic safety range and a faulty sample exceeds the basic safety range, it is judged as "accurately distinguished"; otherwise, it is judged as "misclassified". 4. Compliance Requirements: For each "Label-Parameter" entry, the verification accuracy must be ≥ the preset value. If the accuracy is not met, the amount of historical data under that label must be expanded, the confidence interval must be recalculated and verified until the accuracy requirement is met.

[0038] When the duration of unconventional extreme weather is ≥2 hours, the system does not directly use 24 sample points to calculate the pre-set confidence interval, but instead initiates "small sample fusion + iterative optimization", the specific process of which is as follows: 1. Small sample preprocessing: For the 24 temporary emergency samples collected within 2 hours, Kalman filtering was used to remove noise and the 3σ criterion was used to remove transient interference samples to ensure that the proportion of effective samples is greater than or equal to the preset value; 2. Transfer learning fusion: The preprocessed small samples are fused with historical data of adjacent working condition labels in the parameter library (such as pressure and temperature data under the label "low load-low temperature-stable period") through transfer learning, and the distribution characteristics of historical data are used to make up for the deficiency of insufficient small sample size. 3. Temporary Interval Labeling and Iterative Optimization: Based on the fused data, a temporary pre-set confidence interval is calculated and labeled as "to be verified" to be added to the parameter library; when similar extreme weather occurs again, the new samples are continuously integrated into the labeled dataset. Every time 100 valid samples are accumulated, the confidence interval is recalculated and updated until the sample size is ≥500, forming a stable formal basic safety interval; 4. Constraints: The correction range of the temporary interval shall not exceed ±10% of the adjacent working condition label interval to avoid serious interval distortion caused by small samples.

[0039] In this scheme, the 15% amplitude threshold is the basic reference value. Different amplitude judgment criteria are set for different parameter types. The specific division logic is as follows: Amplitude calculation method: Amplitude = |Current extreme weather key indicator value - Historical average value of similar extreme weather indicators| / Historical average value of similar extreme weather indicators × 100%.

[0040] Differentiation threshold setting: For highly sensitive parameters (such as equipment surface temperature and vibration amplitude), the amplitude threshold is lowered to 8%—even small fluctuations in these parameters can affect equipment safety. For example, when the ambient temperature fluctuation reaches 8%, the lubricating oil viscosity has changed by more than 10%, and dynamic correction needs to be initiated.

[0041] For moderately sensitive parameters (such as pressure and current), the basic amplitude threshold of 15% is used. Such parameters have strong anti-interference ability, and fluctuations within the 15% amplitude can be adapted through model correction.

[0042] For low-sensitivity parameters (such as flow rate and pipeline leakage), the amplitude threshold is increased to 20%—the normal fluctuation range of such parameters is relatively large, and the 20% amplitude can still be covered by adjacent operating condition ranges.

[0043] To avoid data version inconsistencies when updating thresholds without system downtime, a "dual-version parallel + atomic switching" update mechanism is adopted. The specific process is as follows: 1. The system has two built-in parameter library instances: "Running Version" and "Version to be Updated". The "Running Version" supports real-time monitoring and early warning queries, while the "Version to be Updated" is used for incremental learning and interval updates. 2. The fusion of new data with temporary emergency datasets and the recalculation of confidence intervals are all performed in the "to be updated version" and do not affect the normal retrieval and threshold output of the "running version"; 3. When the "version to be updated" completes its iteration and the verification accuracy is greater than or equal to the preset value, the system performs an atomic switching operation during the interval between parameter acquisition and early warning judgment (≤10ms) to upgrade the "version to be updated" to the new "running version". The old version is retained for 7 days as a backup. If an abnormal early warning occurs after the switch, it can be rolled back to the historical stable version with one click. 4. All threshold ranges are associated with version numbers and effective timestamps, and the threshold version is simultaneously marked in the warning results, making it easier for operations and maintenance personnel to trace data consistency issues.

[0044] To achieve complex multi-tag combination retrieval within 50ms, a three-level optimized architecture of "edge-side local caching + distributed indexing + tag precoding" is adopted, with the specific design as follows: 1. Perform numerical pre-encoding on the operating condition labels (load level - temperature range - running time segment), for example, "high load - normal temperature - 1000-10000h" is encoded as "3-2-2" (load level 1-4, temperature range 1-3, running time segment 1-3), converting string matching into numerical comparison; 2. A key-value pair index of "tag code - parameter range" is built using Redis database, and a secondary index is built for each parameter (pressure, vibration, etc.) to support fast filtering by combination of tag + parameter conditions; the index is stored using hash sharding to distribute the retrieval pressure to multiple nodes, and the retrieval response time of a single node is ≤20ms; 3. The parameter range corresponding to high-frequency operating condition tags (accounting for 20% of the total number of tags and with a retrieval frequency of 80%) is cached locally on the edge computing gateway, eliminating the need to retrieve them from the cloud database each time, and the hot spot tag retrieval response time is ≤10ms; 4. During retrieval, first match the local cache → if no match is found, query the distributed index → ​​return the parameter range and synchronize it to the local cache. The entire process takes ≤40ms, with 10ms of redundancy reserved to cope with peak retrieval pressure, ensuring that the total response time is stable at ≤50ms.

[0045] refer to Figure 5 Optimization methods for the XGBoost algorithm include: A coupling coefficient for operating conditions is introduced. The coefficient is calculated using the Pearson correlation coefficient between unit load and ambient temperature, reflecting the interaction between the two on the parameters. The operating parameters of auxiliary equipment in hydropower stations (such as pressure and flow rate) are significantly affected by the coupling effect of unit load and ambient temperature, and the interaction between the two is not a simple additive one. For example, when the unit is operating at high load, the normal range of cooling water pump outlet pressure increases by 0.01 MPa for every 1°C increase in ambient temperature; while when operating at low load, the normal range of pressure increases by only 0.003 MPa for every 1°C increase in ambient temperature. The operating condition coupling coefficient is calculated by using the Pearson correlation coefficient between unit load and ambient temperature, with a value range of [-1, 1], which can quantify the interaction strength between the two: when the correlation coefficient is 0.7, i.e., a strong positive correlation, the coupling coefficient weight increases to 0.3, and the model prioritizes the synergistic effect of the two; when the correlation coefficient is 0.2, i.e., a weak correlation, the coupling coefficient weight decreases to 0.05, and the model focuses on the influence of a single factor on the parameter. Compared to the existing technology's approach of "ignoring working condition interaction and modeling separately," this optimization reduces the error when the dynamic threshold matches the actual working conditions. Under strongly coupled working conditions such as high load-high temperature and low load-low temperature, the false alarm rate of the warning is further reduced.

[0046] The coupling coefficient under operating conditions is dynamically updated based on the random group's operating status. The Pearson correlation coefficient is calculated every 5 minutes. When the unit load suddenly increases from low load to high load and the ambient temperature rises simultaneously, the coupling coefficient increases in real time, and the model responds quickly to the interaction between the two on the parameters. When the load is stable and the temperature fluctuation is small, the coupling coefficient remains stable, ensuring the smoothness of threshold adjustment.

[0047] It is worth noting that the Pearson correlation coefficient is the full name of the Pearson product-moment correlation coefficient, a statistical indicator used to quantify the degree of linear correlation between two continuous variables. Its value ranges from [-1, 1]. Its core function is to determine the strength and direction of the linear relationship between the unit load and the ambient temperature, providing a quantitative basis for calculating the operating condition coupling coefficient.

[0048] The learning rate, tree depth (3-8), and number of leaf nodes (100-500) of XGBoost were optimized using the particle swarm optimization algorithm. The learning rate, tree depth, and number of leaf nodes in the XGBoost algorithm directly affect the model's fitting ability and generalization performance: a learning rate that is too small will lead to slow model convergence and underfitting, while a learning rate that is too large will easily lead to overfitting; a tree depth that is too shallow will fail to capture complex working condition correlations, while a tree depth that is too deep will amplify the influence of noise; a number of leaf nodes that is too few will result in coarse decision boundaries, while a number of leaf nodes will increase computational costs. The Particle Swarm Optimization algorithm simulates the collective intelligent behavior of a flock of birds foraging. Within a preset search space, with a learning rate of 0.01-0.1, a tree depth of 3-8, and 100-500 leaf nodes, iteratively optimizes the hyperparameters to ultimately lock in the globally optimal combination.

[0049] An incremental learning mechanism is built, which fine-tunes the model monthly using new data, eliminating the need for full retraining.

[0050] In existing technologies, model updates require retraining based on all historical data (over 3 years), which is time-consuming and consumes significant computing resources, making it difficult to meet the real-time monitoring needs of hydropower station auxiliary equipment. The incremental learning mechanism of this invention, through "fine-tuning with new data + retention of historical parameters," only requires partial updates to the model each month using newly added operational data, eliminating the need for repeated training on historical data. This results in shorter update times and lower computational resource consumption, thereby significantly improving the engineering efficiency of model maintenance.

[0051] Auxiliary equipment may experience characteristic drift over time, such as increased vibration amplitude due to wear and decreased flow rate due to seal aging. The incremental learning mechanism can capture these slow changes by adding new data each month and dynamically adjust the model parameters so that the threshold range always adapts to the current state of the equipment.

[0052] It is worth noting that during incremental learning, the model only fine-tunes the parameters for the working condition labels corresponding to the newly added data, retaining the best fitting parameters for the historical working conditions, thus avoiding the problem of "decreased adaptability to old working conditions" that may be caused by full retraining.

[0053] Through triple optimization—introducing the coupling coefficient of operating conditions, optimizing hyperparameters through particle swarm optimization, and employing an incremental learning mechanism—the XGBoost algorithm achieves a technological breakthrough in early warning scenarios for hydropower station auxiliary equipment, characterized by "accurate fitting, efficient maintenance, and dynamic adaptation." Compared to the unoptimized XGBoost model, it achieves more accurate predictions, lower false alarm rates, lower false alarm rates, lower model update time, and lower computational requirements, significantly reducing operation and maintenance costs. Simultaneously, the model can adapt to various equipment types, complex operating conditions, and changes in equipment characteristics throughout its entire lifecycle. This expands the applicability of the early warning system from single equipment to all types of auxiliary equipment in hydropower stations, such as cooling systems, lubrication systems, and compressed air systems, providing core technical support for the engineering implementation and large-scale application of early warning systems.

[0054] The optimization methods for the XGBoost algorithm include: introducing a working condition coupling coefficient, using particle swarm optimization for hyperparameter optimization, and constructing a pseudo-incremental learning mechanism based on model retraining and parameter transfer, as detailed below: The operating condition coupling coefficient is used to quantify the impact of the interaction between unit load and ambient temperature on the operating parameters of auxiliary equipment. It does not directly use the Pearson correlation coefficient as a weight, but is constructed through a process of "quantifying the correlation strength using the Pearson correlation coefficient + converting it into model input features via nonlinear mapping." The specific steps are as follows: 1. Select the unit load and ambient temperature operation data of the most recent hour (sample size ≥ 120 sets to ensure statistical validity), calculate the Pearson correlation coefficient r between the two, with a value range of [-1, 1], to determine the strength and direction of the linear correlation between load and temperature; 2. Establish a nonlinear mapping function from Pearson correlation coefficient to operating condition coupling coefficient: ; (Where k is the sensitivity coefficient, calibrated to 5 based on historical data), r Mapping to the [0, 0.5] interval eliminates the interference of negative correlation coefficients on the model, while avoiding statistical fallacies caused by directly assigning weights; 3. Adjust the operating condition coupling coefficient w As an independent feature input to the XGBoost model, it participates in threshold prediction together with features such as unit load, ambient temperature, and equipment operating time. The model automatically learns the influence weight of the coupling coefficient on the parameter safety range through training.

[0055] To ensure compatibility between model execution and coefficient updates, the coupling coefficients are not directly embedded into the XGBoost model structure for real-time updates. Instead, they are implemented using an "offline pre-computation + online table lookup" approach. 1. Offline phase: Based on 3 years of historical data, Pearson correlation coefficients and coupling coefficients under different load-temperature combinations are calculated according to the operating condition labels. A mapping table of "operating condition label-coupling coefficient" is established and stored in a structured threshold parameter library. 2. Online Phase: The working condition label module generates real-time working condition labels every 5 minutes. The system directly retrieves the corresponding coupling coefficients from the parameter library and inputs them into the model. There is no need to calculate the Pearson correlation coefficient in real time during the model inference process. This ensures the dynamic adaptability of the coupling coefficients and avoids conflicts with the batch processing mode of the XGBoost model.

[0056] Through a triple optimization approach—introducing the coupling coefficient of operating conditions, optimizing hyperparameters through particle swarm optimization, and employing a pseudo-incremental learning mechanism—the XGBoost algorithm achieves a technological breakthrough in early warning scenarios for hydropower station auxiliary equipment, achieving "accurate fitting, efficient maintenance, and dynamic adaptation." This significantly reduces operation and maintenance costs. Simultaneously, the model can adapt to multiple types of equipment, complex operating conditions, and changes in equipment characteristics throughout its entire lifecycle. This expands the applicability of the early warning system from single equipment to all types of auxiliary equipment in hydropower stations (such as cooling systems, lubrication systems, and compressed air systems), providing core technical support for the engineering implementation and large-scale application of early warning systems.

[0057] refer to Figure 6 The output logic of the dynamic safety interval includes: First, the parameter library is called to output the basic safe range of the corresponding label. Then, by optimizing the XGBoost model and combining it with the current parameter fluctuation variance, the basic range is dynamically corrected to the preset segment value. The correction magnitude is positively correlated with the variance of parameter fluctuations, and the final interval must cover normal operating data above the preset value.

[0058] In the structured threshold parameter library, the basic safety range corresponding to each operating condition label, such as "high load - normal temperature - 1000-10000h", is constructed based on a pre-set confidence interval of historical data for a preset year, and outliers have been removed. It possesses extremely high statistical reliability. When outputting dynamic safety ranges, the baseline range is first quickly retrieved through the operating condition label to ensure the "scenario adaptability" of the threshold.

[0059] The operating parameters of auxiliary equipment, such as pressure and flow rate, fluctuate dynamically with real-time operating conditions. If a fixed baseline range is still used, false alarms are likely to occur due to normal parameter fluctuations. This logic optimizes the XGBoost model to capture the correlation between parameter fluctuation variance and operating conditions, and dynamically corrects the baseline range values ​​for preset segments.

[0060] For example, when the parameter fluctuation variance is small, such as under stable operating conditions, the fluctuation variance is ≤0.03, and the correction range is ±3%. For example, the pressure base range of 0.4-0.6MPa is corrected to 0.412-0.618MPa to avoid over-correction that leads to threshold distortion.

[0061] When the variance of parameter fluctuations is large, such as under load fluctuations or fluid pulsation conditions, the variance of fluctuation is ≥0.06, and the correction range is ±6%~±8%. For example, the pressure base range is corrected to 0.424-0.648MPa, and the safety range is expanded to accommodate normal fluctuations.

[0062] During the correction process, the optimized XGBoost model incorporates the operating condition coupling coefficient, i.e., the interaction between unit load and ambient temperature, to adjust the correction direction. For example, under high load-high temperature strong coupling conditions, when the fluctuation variance increases, the correction is more biased towards the upper limit of the range to avoid "normal high pressure being misjudged as exceeding the limit".

[0063] The correction magnitude is positively correlated with the variance of parameter fluctuation. Essentially, it allows the safe range to be dynamically adjusted according to the stability of the parameter: the more stable the parameter, the smaller the correction magnitude and the more accurate the threshold, ensuring that fault signals can be captured in a timely manner; the more unstable the parameter, the larger the correction magnitude and the more lenient the threshold, avoiding normal fluctuations from being misjudged as faults.

[0064] The correction range is strictly limited to the preset range values, which ensures the flexibility of the threshold while avoiding "over-correction leading to threshold runaway".

[0065] The ultimate goal of the dynamic safety range is to cover normal operating data above the preset value. This indicator is based on statistical reliability design: through the basic range, the preset normal data is covered, and with dynamic correction, the superposition of 3% of normal data caused by fluctuations is supplemented, which greatly reduces the probability of normal data being misjudged as abnormal, and significantly reduces the amount of ineffective inspection work of operation and maintenance personnel.

[0066] The final range covers normal data above the preset value, indicating that ≤2% of normal data may be misjudged, while fault data, such as increased vibration amplitude due to bearing wear or pressure drop due to pipeline leakage, usually exceeds this range. This design of "preset value coverage + clear boundaries" improves the distinction between fault signals and normal data, reduces the false alarm rate, and ensures the safe operation of hydropower station auxiliary equipment.

[0067] The output logic of the dynamic safety range addresses the core shortcomings of existing technologies—fixed thresholds, poor scenario adaptability, and imbalance between anti-interference and sensitivity—through a three-layer design: a basic range providing a benchmark, dynamic correction of fluctuation variance, and guaranteed coverage of preset values. Furthermore, this logic requires no manual intervention and can be executed in real-time within the edge computing gateway, adapting to the real-time monitoring needs of hydropower station auxiliary equipment and providing accurate and reliable threshold support for operation and maintenance decisions.

[0068] refer to Figure 7 Trend warning signals include: When a parameter directly exceeds the dynamic safety range, an over-limit warning signal is generated, which has a higher priority than the trend warning for the same parameter. The over-limit warning needs to indicate the over-limit amplitude, i.e., the absolute value, the proportion of the range width, and the duration of the over-limit.

[0069] Over-limit warning signals have higher priority than trend warnings for the same parameters. Essentially, this establishes a correspondence between "urgency level" and "response priority": a parameter directly exceeding the dynamic safety range means the equipment has entered an "abnormal operating state," constituting an urgent fault requiring immediate attention; while trend warnings only reflect parameter drift towards the range boundary. For example, if pressure parameters drift towards the range boundary for several consecutive cycles, it represents a potential risk that has not yet exceeded the limit. By setting priorities, maintenance personnel can quickly focus on over-limit warnings, avoiding delayed emergency fault responses caused by "mixing trend warnings and over-limit warnings."

[0070] By implementing an over-limit early warning rule that combines "priority setting and quantitative information labeling," the technical goals of "prioritizing emergency fault response, accurately quantifying anomalies, and effectively filtering interference signals" have been achieved. Compared to traditional over-limit early warning systems, this rule improves the efficiency of emergency fault response, the accuracy of fault location, and reduces the false alarm rate of instantaneous interference. At the same time, the quantitatively labeled anomaly information (amplitude, percentage, and duration) provides maintenance personnel with a complete decision chain of "fault severity - handling priority - investigation direction," reducing the average fault handling time and significantly improving the maintenance efficiency and safe operation guarantee capability of hydropower station auxiliary equipment.

[0071] This embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0072] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0073] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory, or random access memory.

[0074] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0075] Example 2 This invention also proposes an early warning system for auxiliary equipment in hydropower stations, referring to... Figure 8 include: The multi-sensor acquisition module collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment in the hydropower station through sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The operating condition label generation module generates real-time operating condition labels from the collected data based on operating condition parameters and environmental parameters. The operating condition labels include at least the unit load level, ambient temperature range, and equipment operating duration. The threshold parameter library construction module establishes a structured threshold parameter library based on real-time operating condition labels, and has a built-in threshold adjustment model based on the optimized XGBoost algorithm to output the dynamic safety range of each operating condition parameter under the corresponding operating condition. The early warning signal generation module monitors the changing trends of various operating parameters. When a parameter drifts towards the boundary of the corresponding dynamic safety range for N consecutive cycles, it generates a trend early warning signal. The early warning result generation module uses DS evidence theory or weighted voting mechanism to fuse the credibility of trend early warning signals corresponding to each working condition parameter, outputs the final early warning result, and finally outputs it in the form of visualization or audible and visual alarm.

[0076] The early warning system for auxiliary equipment of a hydropower station provided in Embodiment 2 of the present invention can realize all the processes of the early warning method for auxiliary equipment of a hydropower station in Embodiment 1 above. The functions and technical effects of each module in the early warning system for auxiliary equipment of a hydropower station are the same as those of the early warning method for auxiliary equipment of a hydropower station in Embodiment 1 above, and will not be repeated here.

[0077] Example 3 To address the aforementioned technical problems, this embodiment also provides a computer device. Please refer to [link / reference needed] for details. Figure 9 , Figure 9 This is a basic structural block diagram of the computer device in this embodiment.

[0078] The computer device 3 includes a memory 31, a processor 32, and a network interface 33 that are interconnected via a system bus. It should be noted that only the computer device 3 with components 31-33 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors, embedded devices, etc.

[0079] The computer device 3 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device 3 can interact with the user through a keyboard, mouse, remote control, touchpad, or voice control device.

[0080] The memory 31 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory such as SD or DX memory, random access memory, static random access memory, read-only memory, electrically erasable programmable read-only memory, programmable read-only memory, magnetic memory, disk, optical disk, etc. In some embodiments, the memory 31 may be an internal storage unit of the computer device 3, such as the hard disk or memory of the computer device 3. In other embodiments, the memory 31 may also be an external storage device of the computer device 3, such as a plug-in hard disk, smart memory card, secure digital card, flash memory card, etc., equipped on the computer device 3. Of course, the memory 31 may include both the internal storage unit and its external storage device of the computer device 3. In this embodiment, the memory 31 is typically used to store the operating system and various application software installed on the computer device 3, such as computer-readable instructions for a hydropower station auxiliary equipment early warning method. In addition, the memory 31 may also be used to temporarily store various types of data that have been output or will be output.

[0081] In some embodiments, the processor 32 may be a central processing unit, a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 32 is typically used to control the overall operation of the computer device 3. In this embodiment, the processor 32 is used to execute computer-readable instructions stored in the memory 31 or to process data, for example, to execute computer-readable instructions for a hydropower station auxiliary equipment early warning method.

[0082] The network interface 33 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 3 and other electronic devices.

[0083] Example 4 This embodiment also provides another implementation method, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the above-described early warning method for auxiliary equipment of a hydropower station.

[0084] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium, such as ROM / RAM, magnetic disk, or optical disk, and includes several instructions to cause a terminal device, such as a mobile phone, computer, server, air conditioner, or network device, to execute the methods described in the various embodiments of this application.

[0085] The above are merely preferred embodiments of the present invention and are not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for early warning of auxiliary equipment in a hydropower station, characterized in that, Includes the following steps: The system collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment within the hydropower station using sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The collected data is used to generate real-time operating condition tags based on operating condition parameters and environmental parameters. The operating condition tags include at least the unit load level, ambient temperature range, and equipment operating duration. A structured threshold parameter library is established using real-time operating condition labels, and a threshold adjustment model based on the optimized XGBoost algorithm is built in to output the dynamic safety range of each operating condition parameter under the corresponding operating condition. Monitor the changing trends of various operating parameters, and generate a trend warning signal when the parameter drifts towards the boundary of the corresponding dynamic safety range for N consecutive cycles. The credibility of the trend warning signals corresponding to each working condition parameter is fused using DS evidence theory or weighted voting mechanism, and the final warning result is output in the form of visualization or audible and visual alarm.

2. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, The collected data is preprocessed, including: The Kalman filter algorithm is used to filter out random noise in pressure and flow parameters, and the 3σ criterion is used to remove abnormal jump values ​​in temperature and vibration parameters. Missing data are filled in using adjacent period linear interpolation.

3. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, Its features are, The rules for generating real-time operating condition labels include: Unit load levels are classified according to the percentage of rated load; Ambient temperature ranges are divided according to climatic characteristics; The operating time of equipment is divided according to its life cycle; The label format should be uniformly set to "load level - temperature range - running time segment".

4. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, The process of constructing the structured threshold parameter library includes: Collect historical operating data of auxiliary equipment for a preset year, covering normal data and labeled fault data under different operating condition labels; The distribution characteristics of each parameter are statistically analyzed according to the working condition label, and the pre-set information interval is used as the basic safety range to eliminate fault abnormal values ​​within the interval. Each tag-parameter entry is associated with a data source, a verification accuracy greater than a preset value, and an update timestamp. It supports quick retrieval by tag and presets the corresponding time.

5. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, The optimization methods for the XGBoost algorithm include: A coupling coefficient for operating conditions is introduced, which is calculated using the Pearson correlation coefficient between unit load and ambient temperature, reflecting the interaction between the two on the parameters; The learning rate, preset tree depth, and preset number of leaf nodes of XGBoost were optimized using the particle swarm optimization algorithm, and the model prediction error was less than the preset value. An incremental learning mechanism is built, which fine-tunes the model monthly using new data, eliminating the need for full retraining.

6. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, The output logic of the dynamic safety zone includes: First, the parameter library is called to output the basic safe range of the corresponding label. Then, the XGBoost model is optimized and combined with the current parameter fluctuation variance to dynamically correct the basic range value. The correction magnitude is positively correlated with the variance of parameter fluctuations, and the final interval must cover the normal operating data of the preset value.

7. The early warning method for auxiliary equipment of a hydropower station according to claim 1, characterized in that, The trend warning signals include: When a parameter directly exceeds the dynamic safety range, an over-limit warning signal is generated, which has a higher priority than a trend warning signal for the same parameter. Over-limit warnings must specify the over-limit amplitude, i.e., the absolute value, the proportion of the interval width, and the duration of the over-limit.

8. A hydropower station auxiliary equipment early warning system, said system being used to implement the hydropower station auxiliary equipment early warning method according to any one of claims 1-7, characterized in that, include: The multi-sensor acquisition module collects data on pressure, vibration, current, flow rate, and surface temperature of auxiliary equipment in the hydropower station through sensors, and also collects operating parameters and environmental parameters of related pipelines of the auxiliary equipment. The operating condition label generation module generates real-time operating condition labels from the collected data based on operating condition parameters and environmental parameters. The operating condition labels include at least the unit load level, ambient temperature range, and equipment operating duration. The threshold parameter library construction module establishes a structured threshold parameter library based on real-time operating condition labels, and has a built-in threshold adjustment model based on the optimized XGBoost algorithm to output the dynamic safety range of each operating condition parameter under the corresponding operating condition. The early warning signal generation module monitors the changing trends of various operating parameters. When a parameter drifts towards the boundary of the corresponding dynamic safety range for N consecutive cycles, it generates a trend early warning signal. The early warning result generation module uses DS evidence theory or weighted voting mechanism to fuse the credibility of trend early warning signals corresponding to each working condition parameter, outputs the final early warning result, and finally outputs it in the form of visualization or audible and visual alarm.

9. A computer device, characterized in that, include: A memory that stores computer-readable instructions; A processor, which, when executing the computer-readable instructions, implements the steps of a method for early warning of auxiliary equipment in a hydropower station as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of a hydropower station auxiliary equipment early warning method as described in any one of claims 1-7.

Images