A control method for automatically judging power deviation alarm of hydroelectric generating set and correcting deviation

Abstract

The application discloses a kind of control methods of automatic determination of hydroelectric generating set power deviation alarm and deviation correction, it is related to the technical field of hydroelectric generating set automatic control and condition monitoring, wherein the method comprises: collecting the multi-source operation data of hydroelectric generating set, and obtaining clean data set after rejecting abnormal data;According to the clustering division of operating condition parameter, the power reference model of multi-source data fusion is built for various operating conditions;Based on historical power deviation, determine dynamic alarm threshold, and compare real-time deviation to alarm in stages;After fault alarm, automatically troubleshoot fault location, generate control parameters to carry out closed-loop correction until power meets the standard.The application is realized by software programming combined with industrial control hardware, can be directly embedded in existing hydroelectric generating set control system, without reconstructing hardware architecture, the transformation cost is reduced by more than 40%, suitable for various hydroelectric generating sets, with wide engineering application prospect and popularization value.

Description

Technical Field

[0001] This invention relates to the field of automatic control and status monitoring technology for hydropower units, and in particular to a control method for automatically detecting and correcting power deviation alarms in hydropower units. Background Technology

[0002] As the core power source and main force for frequency regulation and peak shaving in the power system, the power output stability of hydropower units directly determines the grid frequency, voltage quality, and power supply reliability. In existing technologies, a closed-loop power regulation system is constructed through the coordinated operation of guide vane opening adjustment, excitation system control, and speed regulation system.

[0003] However, existing power deviation control methods directly employ fixed threshold alarms and single PID correction, without considering differences in operating conditions or the fusion of multi-source data. This may lead to false alarms triggered by normal fluctuations under low load conditions, missed alarms due to threshold redundancy under high load conditions, or unit oscillations caused by fixed correction parameters, thus affecting control stability and equipment safety. Furthermore, existing technologies lack a connection between alarms and correction, intelligent fault location and hierarchical closed-loop control, and manual intervention is delayed, making it difficult to meet the integrated requirements of "accurate judgment, fast response, and stable control" for hydropower units. Summary of the Invention

[0004] The main objective of this invention is to provide a control method for automatically detecting and correcting power deviations in hydropower units.

[0005] Another objective of this invention is to provide a control device that automatically detects and corrects power deviations in hydropower units.

[0006] The third objective of this invention is to provide an electronic device.

[0007] A fourth objective of this invention is to provide a non-transitory computer-readable storage medium.

[0008] To achieve the above objectives, a first aspect of the present invention proposes a control method for automatically detecting and correcting power deviations in hydropower units, comprising:

[0009] S1. Collect multi-source operation data of hydropower units, and perform outlier removal and preprocessing on the multi-source operation data to obtain a clean data set; S2, based on the operating condition feature parameters in the clean data set, clusters and identifies the operating conditions of the hydropower unit, and constructs an adaptive power standard model based on multi-source data fusion for each operating condition cluster. S3, calculates the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and triggers alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold. S4, when an alarm signal requiring correction is triggered, intelligently locates the cause of the fault, calculates the correction control parameters based on the location result, and executes hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

[0010] Optionally, multi-source operating data of the hydropower unit is collected, and outlier removal and preprocessing are performed on the multi-source operating data to obtain a clean data set, including: The MAD and moving average combined algorithm is used to remove outliers from the multi-source operating data. The quality of the head measurement values ​​from the centralized control, plant measurement, and manual measurement channels is judged, and the channel with the best quality is selected as the current head value to obtain the first intermediate data. The first intermediate data is filled with missing values ​​using local density interpolation to obtain a clean dataset.

[0011] Optionally, based on the operating condition characteristic parameters in the clean data set, the operating conditions of the hydropower units are clustered and identified, and an adaptive power standard model based on multi-source data fusion is constructed for each operating condition cluster, including: An improved KSQDC algorithm was used to cluster the working condition characteristic parameters in the cleaning dataset to obtain multiple working condition clusters. For each operating condition cluster, a power standard model is constructed using an improved kernel extreme learning machine algorithm. The improved kernel extreme learning machine algorithm introduces a local density weight factor, and the kernel function is a Gaussian kernel.

[0012] Optionally, a dynamic alarm threshold is calculated based on historical power deviation data for each operating condition cluster, and different levels of alarm signals are triggered based on the comparison between real-time power deviation and the dynamic alarm threshold, including: Statistically analyze the power deviation sequence during normal operation of each operating condition cluster, and use... The criteria calculate the base threshold and take the maximum value of the base threshold and 0.5% of the rated power as the dynamic alarm threshold; Power deviation is calculated in real time. A level 1 alarm is triggered when the power deviation is between the dynamic alarm threshold and 1.2 times the dynamic alarm threshold for 10 seconds. A level 2 alarm is triggered when the power deviation is between 1.2 times the dynamic alarm threshold and 2 times the dynamic alarm threshold for 30 seconds. A level 3 alarm is triggered when the power deviation is greater than or equal to 2 times the dynamic alarm threshold.

[0013] Optionally, when an alarm signal requiring correction is triggered, the cause of the fault is intelligently located, and correction control parameters are calculated based on the location results. Hierarchical closed-loop correction control is then executed until the power deviation recovers to below the dynamic alarm threshold, including: The decision tree algorithm is used to intelligently locate the cause of the fault and output the top 3 most likely causes of the fault. The target control parameters are solved in reverse based on the cause of the fault, and the improved PID control algorithm is used to perform corrective control in stages, wherein the control step size and control cycle are selected according to the power deviation range. The system monitors power deviation in real time and stops corrective action when the power deviation returns to below the dynamic alarm threshold and remains below it for 10 seconds.

[0014] To achieve the above objectives, a second aspect of the present invention provides a control device for automatically judging and correcting power deviation of hydropower units, comprising: The preprocessing module is used to collect multi-source operating data of the hydropower unit, and to remove outliers and preprocess the multi-source operating data to obtain a clean dataset; The identification module is used to cluster and identify the operating conditions of hydropower units based on the operating condition feature parameters in the clean dataset, and to build an adaptive power standard model based on multi-source data fusion for each operating condition cluster. The alarm module is used to calculate the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and to trigger alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold. The correction module is used to intelligently locate the cause of the fault when an alarm signal requiring correction is triggered, calculate the correction control parameters based on the location result, and execute hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

[0015] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0016] To achieve the above objectives, a third aspect of this application provides an electronic device, including a processor and a memory; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, for implementing the control method for automatically judging and correcting power deviation alarms of hydropower units as described in the first aspect embodiment.

[0017] To achieve the above objectives, the fourth aspect of this application provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the control method described in the first aspect embodiment for automatically determining and correcting power deviation alarms of hydropower units.

[0018] The embodiments of the present invention have the following beneficial effects: This invention is implemented through software programming combined with industrial control hardware. It can be directly embedded into the existing hydropower unit control system without reconstructing the hardware architecture, reducing the transformation cost by more than 40%. It is applicable to all types of hydropower units and has broad engineering application prospects and promotional value. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a control method for automatically judging and correcting power deviation of hydropower units, provided in an embodiment of the present invention; Figure 2 A flowchart of redundant logic provided for embodiments of the present invention; Figure 3 A circuit diagram of the dynamically adaptable P parameter in the PID control parameters of the monitoring system provided in this embodiment of the invention; Figure 4 This is a structural diagram of a control device for automatically judging and correcting power deviation of hydropower units, provided in an embodiment of the present invention. Detailed Implementation

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] The following describes, with reference to the accompanying drawings, a control method and apparatus for automatically judging and correcting power deviation of hydropower units according to an embodiment of the present invention.

[0023] Example 1 This invention provides a control method for automatically detecting and correcting power deviations in hydropower units, such as... Figure 1 As shown, the method includes the following steps: S1 collects multi-source operating data of the hydropower unit, and performs outlier removal and preprocessing on the multi-source operating data to obtain a clean dataset.

[0024] In order to obtain accurate and complete unit operation data to support subsequent modeling and control, this application collects four types of unit parameters from multiple sources with redundancy, and generates a clean dataset through head selection, composite algorithm denoising and interpolation to fill gaps.

[0025] In this embodiment, various measurement sensors already deployed at the unit site, the power plant's original SCADA monitoring system, and external third-party online monitoring devices are integrated to form multiple parallel redundant data acquisition links. Relying on the mode of mutual redundancy and backup of multiple acquisitions, various operational measurement data of the unit are collected synchronously around the clock without interruption, avoiding the drawbacks of key data interruption or data distortion caused by single-point acquisition device failure or line abnormality.

[0026] In this embodiment, four categories of differentiated parameters are used to complete the full-coverage data collection. Among them, the electrical control parameters include the measured active power P, reactive power Q, stator voltage U, stator current I, and excitation current. Operating condition control parameters include upstream water level. Downstream water level Guide vane opening D, axial-flow propeller unit-specific blade opening Unit speed Environmental and power grid parameters include power grid frequency. Grid voltage deviation The equipment status parameters include the displacement feedback of the guide vane actuator. Excitation system output feedback Data collection is completed from multiple dimensions, including the operation of the generating unit itself, the operating conditions of the external power grid, and feedback from on-site equipment.

[0027] In this embodiment, given that the head value directly affects the unit output calculation result, this application specifically designs a three-level redundant architecture with three parallel measurement channels: centralized control, plant measurement, and manual measurement. The quality judgment and optimal switching logic of the three head data sources is as follows: Figure 2 As shown, the system compares the fluctuation amplitude, jump frequency and continuity of the three sets of data in real time, and automatically selects the head measurement result with the best reliability to participate in subsequent calculations, which greatly reduces the acquisition error of key parameters.

[0028] In this embodiment, a composite operation method combining the MAD algorithm and the moving average algorithm is selected to fuse the same source measured data returned by two sets of heterogeneous sensors. With the help of a four-dimensional reliable judgment model, parameter range verification and adjacent sampling point difference constraint are carried out in sequence to accurately filter out abnormal data caused by component failure, strong electromagnetic interference in the factory area, and instantaneous line disturbance. After the initial screening, the first intermediate data is generated.

[0029] In this embodiment, the application continues to search for missing sampling points within the first intermediate data, and uses a local density interpolation algorithm to fill in the missing data items caused by various reasons. After filling in the missing data, the dataset no longer has data gaps or abnormal values, forming a standardized clean dataset and transferring the data to the downstream S2 step.

[0030] In this embodiment, four types of operating parameters of the unit are collected redundantly by multi-source equipment, and a clean data set is obtained by relying on the three-way head selection logic and composite preprocessing algorithm, which lays the foundation for the subsequent construction of an adaptive power standard model based on multi-source data fusion.

[0031] S2, based on the operating condition feature parameters in the clean dataset, clusters and identifies the operating conditions of the hydropower unit, and constructs an adaptive power standard model based on multi-source data fusion for each operating condition cluster.

[0032] In order to obtain accurate baseline power data for each operating condition to support subsequent anomaly detection, this application uses clustering to divide the operating condition into clusters and models based on improved KELM clustering, with an incremental update mechanism to achieve adaptive optimization of the model.

[0033] In this embodiment of the application, core operating condition characteristic parameters that can characterize the unit's operating status are extracted from the clean data set, specifically including the upstream and downstream water level difference. Guide vane opening D, unit speed Grid frequency In addition to reactive power Q, this application uses the aforementioned multidimensional feature parameters as the basis for clustering, and employs an improved KSQDC algorithm to intelligently and automatically cluster massive historical and real-time operating data of generating units. This can accurately distinguish the operating characteristics of generating units under different head conditions, different load states, and different grid environments, thereby dividing multiple clusters of generating unit operating conditions with independent features and clear operating condition boundaries. This enables precise identification and classification of the complex and variable operating conditions of hydropower units.

[0034] In this embodiment, dynamic adaptive matching adjustment of control parameters is achieved for different head operating ranges and different load adjustment ranges of the unit. The correspondence between the dynamically adapted proportional parameter P in the PID control parameters of the monitoring system and the changing operating conditions is as follows: Figure 3 As shown, this application, through a parameter adaptive adaptation mechanism based on operating condition classification, enables the control parameters to always fit the real-time operating characteristics of the unit, greatly improving the convergence effect and fitting accuracy of subsequent model training, while optimizing the response speed and stability of unit power control, providing a solid technical basis for subsequent accurate modeling and intelligent closed-loop control.

[0035] In this embodiment of the application, to avoid the problem that a unified model cannot adapt to different working conditions, this application addresses each divided independent working condition cluster. Each model has its own dedicated adaptive power standard model for specific operating conditions. This application strictly limits the input and output logic of each model, taking into account the upstream and downstream water level differences. Guide vane opening D, unit speed Grid frequency Reactive power Q is used as a multidimensional input feature of the model, and the standard active power value is used as the sole output target of the model, thereby constructing an accurate power mapping function relationship under operating conditions. It can accurately solve the standard power target value of the unit under different complex operating conditions.

[0036] In this embodiment, an improved kernel extreme learning machine algorithm is used to train and build the power standard model corresponding to each operating condition cluster. To further improve the data fitting accuracy of the unit's normal core operating condition area and weaken the interference of edge abnormal operating condition samples and sparse samples on the model accuracy, this application innovatively introduces a local density weighting factor. , where γ i The Euclidean distance between a single training sample and the center of the corresponding operating condition cluster is used as a weighting factor to strengthen the training weight of core dense operating condition samples, effectively improving the model's fitting accuracy to the unit's normal operating conditions.

[0037] In this embodiment, a Gaussian kernel is uniformly selected as the kernel function for the improved kernel extreme learning machine algorithm, and the kernel parameters are fixed. =0.8, and the number of hidden layer nodes in the model is preset to 128. Furthermore, this application supports iterative optimization and dynamic optimization of the number of hidden layer nodes through a grid search algorithm, which can adaptively adjust the model structure according to the actual operating data characteristics of the unit, further conforming to the actual operating characteristics of the unit.

[0038] In this embodiment, multi-fold cross-validation is used to complete the global iterative optimization of all model parameters. The model parameters are corrected iteratively through repeated training and validation. The overall error of the model test set is strictly constrained to be less than or equal to 0.5%, which ensures the output accuracy and generalization ability of the power standard model for each operating condition cluster from the algorithm level.

[0039] In this embodiment, a complete operating condition-model mapping repository is built and stored, binding and archiving all operating condition clusters with their corresponding dedicated power standard models. During actual unit operation, this application can identify the current operating condition of the unit in real time, automatically retrieve and call the power standard model corresponding to the matching operating condition cluster, and quickly and accurately output the real-time standard power value of the unit under the current operating condition.

[0040] This application is equipped with a complete model adaptive optimization mechanism. As the core function of the control optimization layer of this application, it can achieve long-term accurate model adaptation. Every time 1000 sets of unit normal operation data and unit steady-state operation data after automatic correction are accumulated, the incremental training process of the power standard model of the corresponding operating condition cluster will be started. The internal parameters of the model will be updated in real time to continuously offset the accuracy decay caused by unit equipment aging and operating condition deviation, and maintain the prediction accuracy of the model in the long term.

[0041] In the embodiments of this application, when the newly generated operating data generated by the long-term operation of the unit is clustered to form a new operating condition cluster, and the profile coefficient of the new operating condition cluster is greater than or equal to 0.6, this application can automatically identify the newly generated valid operating condition and autonomously complete the construction and storage of the power standard model and matching parameter system corresponding to the new operating condition. There is no need for manual modeling and parameter configuration, which perfectly realizes the adaptive adaptation capability of the unit to the new unknown operating conditions, and greatly improves the versatility and intelligence of the control system of this application.

[0042] In this embodiment, the operating condition clusters are divided by improving the KSQDC algorithm, a power model is constructed based on the optimized KELM clustering and a dynamic update mechanism is set, and the standard power is output to support subsequent threshold calculation and unit management.

[0043] S3 calculates the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and triggers alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold.

[0044] In order to accurately identify abnormal unit power and manage it in a tiered manner, this application calculates an adaptive dynamic threshold based on clustered historical data, implements a three-level alarm by comparing it with real-time power deviation, and periodically optimizes the threshold parameters.

[0045] In this embodiment of the application, for each cluster of working conditions that has been clustered... Historical operating data of the units under corresponding operating conditions under long-term normal and stable operation were comprehensively collected. Standardized power deviation sequences for each operating condition cluster were obtained through batch statistical calculations. The power deviation calculation method is as follows: This application relies on massive historical deviation sample data to perform threshold calculation, effectively ensuring that the alarm thresholds solved subsequently closely match the actual normal operating fluctuation characteristics of the unit, and avoiding the problem of threshold determination distortion caused by insufficient sample size.

[0046] In this embodiment, the application introduces the 3σ mathematical statistical criterion to perform statistical analysis on the historical power deviation data of each operating condition cluster, and calculates the basic alarm threshold corresponding to each operating condition cluster. ,in This represents the average historical power deviation of the corresponding operating condition cluster. To correspond to the standard deviation of historical power deviations in the operating condition cluster, the basic threshold can accurately cover the normal power fluctuation range during the normal operation of the unit, and statistically distinguish between normal operation fluctuations and abnormal power deviations.

[0047] In this embodiment, to strictly comply with national industry standards and specifications, and to prevent frequent false alarms and invalid alarms caused by excessively low basic thresholds, this application combines the rated power of the unit. For the base threshold By performing adaptive correction processing, this application sets the dynamic alarm threshold T, which is ultimately used for actual judgment, to... The system's minimum alarm threshold is always no less than 0.5% of the rated power. While ensuring that minor normal fluctuations do not trigger alarms, it can accurately capture abnormal deviations in actual power, taking into account both the unit's operational stability monitoring needs and the system's operational reliability.

[0048] In this embodiment, a dynamic threshold periodic adaptive update mechanism is set. Considering that the performance of the hydropower unit will slowly decline and the hydraulic operating characteristics will slightly deviate during long-term operation, this application automatically retrieves the latest normal steady-state operation data of the unit every 24 hours, re-statistically analyzes the power deviation distribution and iteratively updates the dynamic alarm threshold of each operating condition cluster, so that the threshold judgment standard can continuously adapt to the real-time operating status of the unit and ensure the accuracy and effectiveness of power anomaly judgment in the long term.

[0049] In this embodiment, during normal operation of the unit, the application collects the measured value of the unit's active power in real time, and combines it with the real-time standard active power value output by the S2 model to calculate the current power deviation of the unit in real time. The system compares the real-time calculated power deviation value with the dynamic alarm threshold T to complete the real-time determination of multi-level alarm logic.

[0050] In this embodiment, a three-level differentiated hierarchical alarm mechanism is constructed, and corresponding processing strategies are set for power anomaly deviations of different severity, effectively distinguishing between minor fluctuations, anomalies requiring correction, and emergency fault states.

[0051] In this embodiment, when the real-time power deviation value is within the range of dynamic threshold T to 1.2T, and this abnormal state remains stable for 10 seconds, this application triggers a Level 1 early warning alarm. This level of alarm only records the abnormal state and provides a signal prompt, without initiating any automatic power correction control program, adapting to scenarios with slight instantaneous power deviation of the unit. When the real-time power deviation value is within the range of 1.2T to 2T, and this abnormal state lasts for 30 seconds, this application triggers a Level 2 power correction alarm, determining that the unit has a persistent power deviation problem, and simultaneously activating the closed-loop automatic power correction control process to achieve autonomous repair of the abnormality. When the real-time power deviation value is greater than or equal to 2T, this application immediately triggers a Level 3 emergency fault alarm. This application simultaneously locks the system's automatic power correction control function, actively pushes a shutdown inspection prompt command, avoids safety risks such as equipment damage and fault expansion caused by forced control, and ensures the safe operation of the unit equipment.

[0052] After triggering any level of alarm signal, this application can automatically integrate all-dimensional abnormal information such as the current unit operating condition type, standard power value, measured power value, real-time power deviation value, and the matched fault probability cause. At the same time, it outputs alarm information through a dual early warning method of on-site audible and visual alarm hardware device and ONCALL SMS remote push, realizing the combination of on-site prompts and remote operation and maintenance reminders, ensuring that operation and maintenance personnel can grasp the abnormal operating conditions of the unit as soon as possible.

[0053] In this embodiment, to further optimize the accuracy of alarm judgment, this application relies on the long-term accumulated correction operation logs and alarm history data of the unit, and uses the Q-learning reinforcement learning algorithm to intelligently iteratively optimize the hierarchical alarm thresholds of each operating condition cluster. This application sets the threshold parameter optimization update to be completed every 7 days. Through continuous data learning, the alarm judgment criteria are constantly corrected, further reducing the probability of false alarms and missed alarms, and continuously improving the intelligence level and accuracy of the overall power anomaly monitoring and hierarchical alarm of this application.

[0054] In this embodiment, a dynamic threshold is calculated by combining historical deviations with the 3σ criterion, and the real-time power deviation is compared to trigger alarms in stages and the threshold is periodically optimized, laying the foundation for subsequent graded closed-loop correction control.

[0055] S4, when an alarm signal requiring correction is triggered, intelligently locates the cause of the fault, calculates the correction control parameters based on the location result, and executes hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

[0056] In order to accurately eliminate abnormal deviations in unit power and ensure safe operation of equipment, this application relies on decision tree to locate faults, solve control parameters in different scenarios, adopts hierarchical feedforward PID closed-loop correction, and is equipped with fault tolerance and standardized hardware interface.

[0057] In this embodiment, when the system detects the triggering of a secondary correction alarm signal, it immediately initiates an intelligent fault location process. This application constructs a dedicated high-precision fault location model based on a pre-trained decision tree algorithm, selecting multi-dimensional parameters that comprehensively characterize the abnormal state of the unit as model input variables. These parameters include real-time power deviation of the unit, upstream and downstream water level difference, guide vane actuator displacement feedback, excitation current, and grid frequency. Through intelligent computation and probability ranking processing of the fault location model, this application can stably output the top three fault causes by probability, accurately identifying four types of common unit faults: guide vane opening deviation, excitation system abnormality, grid frequency disturbance, and water level imbalance. The overall identification accuracy of the fault location model configured in this application is no less than 90%, providing precise fault tracing basis for subsequent targeted correction and control.

[0058] In this application embodiment, differentiated control parameter inverse solution logic is established for different types of unit fault scenarios. This allows for matching the optimal correction control strategy according to the actual fault type, completely solving the problems of poor targeting, low correction efficiency, and weak adaptability of traditional unified control methods. In this application embodiment, guide vane opening deviation is a high-incidence fault type accounting for more than 60% during the operation of the unit. For this type of fault, this application uses a power standard model for inverse calculation processing, based on the formula... Solve for the target guide vane opening under the current operating conditions of the unit. And combined with real-time guide vane displacement feedback data The optimal adjustment amount of the guide vane opening was accurately calculated. This enables precise correction of guide vane opening deviation.

[0059] In this application embodiment, for power deviation faults caused by excitation system abnormalities, this application relies on the PQ power balance equation. Solve for the target reactive power adapted to the current operating conditions. And combine the real-time reactive power Q of the unit to calculate the precise adjustment of reactive power. This enables precise compensation and control of the excitation conditions.

[0060] To address the issue of generator power fluctuations caused by grid frequency disturbances, this application adopts a coordinated control mode that links the speed regulation system and the excitation system, combined with a joint control strategy of active power fine-tuning and reactive power compensation, to accurately correct the disturbed grid frequency to the 50Hz rated standard frequency, thereby eliminating the interference of grid frequency fluctuations on generator output.

[0061] To address water level imbalance faults caused by upstream and downstream water level differences exceeding the normal range of ±10%, this application prioritizes correcting and optimizing the unit's target load using a correction formula. The unit load is rematched to adapt to the current water level conditions to avoid overload and abnormal stress on the unit equipment caused by forced correction control, thus protecting the safe operation of the unit equipment in all aspects.

[0062] In this application embodiment, in order to adapt to different degrees of power abnormality deviation conditions and avoid the problem of excessive unit operation fluctuation and insufficient correction accuracy caused by a single control parameter, this application adopts an improved PID control algorithm with feedforward compensation function to realize hierarchical closed-loop correction control. Based on the real-time power deviation range, two levels of differentiated correction control are divided to achieve precise matching between the control intensity and the degree of abnormality.

[0063] In this embodiment, when the real-time power deviation of the unit is within a slightly abnormal range of 1.2T to 1.6T, this application initiates the first-level correction control mode, configured with 0.5%. The refined control step size, 500 millisecond control cycle, and 0.6 PID proportional coefficient enable smooth, small-amplitude correction, ensuring stable unit operation. In this embodiment, when the real-time power deviation of the unit is within the moderately abnormal range of 1.6T to 2.0T, this application activates the secondary correction control mode, configured with 1.0%. The enhanced control step size, fast control cycle of 300 milliseconds, and PID proportional coefficient of 0.8 improve the correction response speed and control strength, quickly eliminate large power deviations, and achieve an overall adaptive intelligent control logic where the greater the deviation, the faster the control response and the stronger the control strength.

[0064] In this application embodiment, in order to overcome the inherent defects of response lag and control delay in traditional PID control, this application adds a feedforward compensation mechanism for the rate of change of operating parameters. This application collects dynamic operating parameters such as the rate of change of unit water level difference and the rate of change of grid frequency in real time, predicts the changing trend of unit power deviation in advance, and intervenes in advance to compensate the control amount, effectively offsetting the system control lag and greatly improving the real-time performance and accuracy of the entire correction control.

[0065] In the embodiments of this application, all the correction control commands generated by the calculation of this application are sent to the field actuators such as guide vane actuators and excitation regulators through the standardized communication interface of the equipment. The command output resolution is less than or equal to 0.1%, which can realize precise control with minute precision and greatly improve the accuracy of unit power correction.

[0066] In this embodiment, the application features a comprehensive real-time feedback verification mechanism for the correction effect. The system continuously collects key operating parameters such as the unit's measured active power, guide vane opening feedback data, and excitation output feedback data within each control cycle, monitoring the correction control effect in real time throughout the entire process, forming a complete closed-loop feedback chain. When the real-time power deviation of the unit falls back to within the dynamic alarm threshold T, and this steady-state condition is maintained stably for 10 seconds, the application determines that the automatic correction control task is complete, automatically terminates all control actions, and fully archives and records the control process and effect of this correction.

[0067] In this embodiment, to improve the operational safety and fault tolerance of the entire control system, this application sets up a dedicated fault-tolerant protection mechanism. When the system performs corrective control ten times in a row and runs for a cumulative three to five seconds, if the unit power deviation is still higher than 1.2T, it is determined that the automatic corrective strategy has failed. This application immediately and automatically switches to the backup control strategy with manual priority, and at the same time upgrades the system alarm level to remind the operation and maintenance personnel to intervene and handle the situation, thus completely avoiding equipment failure and operational risks caused by continuous ineffective control.

[0068] In this application embodiment, the application automatically records all control command operation trajectories, real-time feedback data curves of equipment, and power deviation change trends of the unit throughout the entire process, and stores them uniformly in the background control database, providing complete and reliable data support for subsequent incremental model optimization, iterative updates of control parameters, and source analysis of unit faults.

[0069] In this application embodiment, in order to ensure that the entire adaptive power control scheme can be successfully implemented in engineering and achieve seamless integration and compatibility with the existing unit control system, this application has formulated standardized hardware interface and communication protocol specifications.

[0070] In this embodiment, the data input interface is compatible with the mainstream industrial communication protocols Modbus RTU / TCP and IEC61850, and can perfectly adapt to the data docking requirements of field sensors' 4-20mA and 0-10V analog signals and plant SCADA systems.

[0071] The control output interface of this application supports analog signals, digital signals, and the GOOSE communication protocol, enabling precise connection to the unit's guide vane speed control system and excitation regulator to achieve accurate issuance and execution of control commands. The alarm output interface of this application supports both dry contact output and TCP / IP network alarm output, allowing seamless connection to the unit's local monitoring center and ensuring stable transmission of alarm information.

[0072] In this embodiment, the control unit hardware fully meets the harsh operating conditions of industrial sites. The control unit CPU has a main frequency of not less than 1GHz, a running memory of not less than 2GB, and a local storage capacity of not less than 16GB. It can operate continuously and stably for a long time in a wide temperature industrial environment ranging from -20 degrees Celsius to 60 degrees Celsius, which fully guarantees the long-term reliability and stability of the entire adaptive power control system of this application in complex hydropower industrial scenarios.

[0073] In this embodiment, a decision tree model is used to identify the fault type and solve the control quantity in reverse order of the fault. Closed-loop correction is achieved by relying on hierarchical PID with feedforward, and the unit completes autonomous repair of abnormalities by combining fault tolerance mechanism and standardized hardware interface.

[0074] Example 2 This invention provides a control device 10 for automatically detecting and correcting power deviations in hydropower units, such as... Figure 4 As shown, the device includes: The preprocessing module 100 is used to collect multi-source operating data of the hydropower unit, and to remove outliers and preprocess the multi-source operating data to obtain a clean dataset. The identification module 200 is used to cluster and identify the operating conditions of hydropower units based on the operating condition feature parameters in the clean data set, and to build an adaptive power standard model based on multi-source data fusion for each operating condition cluster. The alarm module 300 is used to calculate the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and to trigger alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold. The correction module 400 is used to intelligently locate the cause of the fault when an alarm signal requiring correction is triggered, calculate the correction control parameters based on the location result, and execute hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

[0075] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0076] Example 3 To implement the methods of the above embodiments, the present invention also provides an electronic device, which includes a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the various steps of the methods described above.

[0077] Example 4 To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method described in the foregoing embodiments.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the 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.

[0079] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0080] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A control method for automatically detecting and correcting power deviations in hydropower units, characterized in that, include: S1. Collect multi-source operation data of hydropower units, and perform outlier removal and preprocessing on the multi-source operation data to obtain a clean data set; S2, based on the operating condition feature parameters in the clean data set, clusters and identifies the operating conditions of the hydropower unit, and constructs an adaptive power standard model based on multi-source data fusion for each operating condition cluster. S3, calculates the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and triggers alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold. S4, when an alarm signal requiring correction is triggered, intelligently locates the cause of the fault, calculates the correction control parameters based on the location result, and executes hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

2. The method according to claim 1, characterized in that, The process involves collecting multi-source operating data from the hydropower units, performing outlier removal and preprocessing on the multi-source operating data, and obtaining a clean data set, including: The MAD and moving average combined algorithm is used to remove outliers from the multi-source operating data. The quality of the head measurement values ​​from the centralized control, plant measurement, and manual measurement channels is judged, and the channel with the best quality is selected as the current head value to obtain the first intermediate data. The first intermediate data is filled with missing values ​​using local density interpolation to obtain a clean dataset.

3. The method according to claim 1, characterized in that, The operating conditions of hydropower units are clustered and identified based on the operating condition feature parameters in the clean data set, and an adaptive power standard model based on multi-source data fusion is constructed for each operating condition cluster, including: An improved KSQDC algorithm was used to cluster the working condition characteristic parameters in the cleaning dataset to obtain multiple working condition clusters. For each operating condition cluster, a power standard model is constructed using an improved kernel extreme learning machine algorithm. The improved kernel extreme learning machine algorithm introduces a local density weight factor, and the kernel function is a Gaussian kernel.

4. The method according to claim 1, characterized in that, The process involves calculating a dynamic alarm threshold based on historical power deviation data for each operating condition cluster, and triggering alarm signals of different levels based on a comparison between the real-time power deviation and the dynamic alarm threshold, including: Statistically analyze the power deviation sequence during normal operation of each operating condition cluster, and use... The criteria calculate the base threshold and take the maximum value of the base threshold and 0.5% of the rated power as the dynamic alarm threshold; Power deviation is calculated in real time. A level 1 alarm is triggered when the power deviation is between the dynamic alarm threshold and 1.2 times the dynamic alarm threshold for 10 seconds. A level 2 alarm is triggered when the power deviation is between 1.2 times the dynamic alarm threshold and 2 times the dynamic alarm threshold for 30 seconds. A level 3 alarm is triggered when the power deviation is greater than or equal to 2 times the dynamic alarm threshold.

5. The method according to claim 1, characterized in that, When the alarm signal requiring correction is triggered, the cause of the fault is intelligently located, and the correction control parameters are calculated based on the location result. Hierarchical closed-loop correction control is then executed until the power deviation recovers to below the dynamic alarm threshold, including: The decision tree algorithm is used to intelligently locate the cause of the fault and output the top 3 most likely causes of the fault. The target control parameters are solved in reverse based on the cause of the fault, and the improved PID control algorithm is used to perform corrective control in stages, wherein the control step size and control cycle are selected according to the power deviation range. The system monitors power deviation in real time and stops corrective action when the power deviation returns to below the dynamic alarm threshold and remains below it for 10 seconds.

6. A control device for automatically detecting and correcting power deviations in hydropower units, characterized in that, include: The preprocessing module is used to collect multi-source operating data of the hydropower unit, and to remove outliers and preprocess the multi-source operating data to obtain a clean dataset; The identification module is used to cluster and identify the operating conditions of hydropower units based on the operating condition feature parameters in the clean dataset, and to build an adaptive power standard model based on multi-source data fusion for each operating condition cluster. The alarm module is used to calculate the dynamic alarm threshold based on the historical power deviation data of each operating condition cluster, and to trigger alarm signals of different levels based on the comparison results between the real-time power deviation and the dynamic alarm threshold. The correction module is used to intelligently locate the cause of the fault when an alarm signal requiring correction is triggered, calculate the correction control parameters based on the location result, and execute hierarchical closed-loop correction control until the power deviation recovers to below the dynamic alarm threshold.

7. An electronic device, characterized in that, Including processor and memory; The processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method as described in any one of claims 1-5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-5.

Images