Water turbine operation evaluation method and system based on multi-source data fusion

Abstract

The invention discloses a water turbine operation evaluation method and system based on multi-source data fusion, and relates to the technical field of water turbine evaluation. A water turbine operation evaluation system based on multi-source data fusion comprises a data acquisition module, a physical constraint module, a pressure wave lag compensation module, a sediment degradation module, a characteristic relation module and an evaluation result module. According to the method, under the condition of corrected water head and efficiency after lag compensation and sediment compensation, the ideal characteristic relation of guide vane opening-active power is fitted based on stable operation historical data, deviation comparison is carried out on an actual operation track and an ideal characteristic curve, and abnormity is identified by using the amplitude, symbol and time sequence characteristics of the deviation, so that the accuracy of the method is improved. And online fine diagnosis of the states of the guide vane and the servo system is realized.

Description

Technical Field

[0001] This invention relates to the field of hydro turbine evaluation technology, and in particular to a hydro turbine operation evaluation method and system based on multi-source data fusion. Background Technology

[0002] As the core equipment of a hydropower station, the operating condition of the turbine directly affects the safety and stability of the unit and the economy of the power station. In order to improve the level of precision in operation management, some projects have begun to try to use mechanism models or data-driven models to analyze and evaluate the unit status. However, overall, it still mainly relies on single-point monitoring, experience judgment and decentralized index evaluation, which makes it difficult to form a unified quantitative description of the turbine's operating status.

[0003] Under complex hydraulic structures and variable operating conditions, existing assessment methods exhibit differences in timestamps, sampling frequencies, and measurement accuracy for observation data from different sources. Furthermore, sensor noise levels, reliability, and long-term stability vary, making unified processing difficult within a single time scale and credibility framework. Asynchrony or contradictions frequently occur between multi-source data. Existing assessment methods often rely solely on statistical characteristics or local empirical curves, lacking mechanisms for consistency verification and state correction by incorporating multiple variables into a unified physical constraint model. For the propagation lag of head signals, changes in pressure wave characteristics, and the cumulative degradation effects of sediment on unit efficiency reduction and increased vibration amplitude in long-distance water diversion systems, existing methods lack explicit modeling. This makes it difficult to distinguish performance changes caused by sediment degradation, control system anomalies, or operating condition fluctuations, limiting the accuracy and interpretability of operational assessment results. Summary of the Invention

[0004] This invention proposes a turbine operation evaluation method that fully considers the propagation delay of the water head in the water intake system, the long-term impact of sediment conditions on efficiency and vibration, and the dynamic characteristics of the guide vane / servo system. It integrates multi-source observation data under a unified time reference and credibility weight, constructs a physical constraint model based on the hydraulic and mechanical mechanisms of the turbine, and continuously performs physical consistency correction and updates on the state vector representing the unit state. To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for evaluating the operation of a hydro-turbine based on multi-source data fusion is proposed. This method constructs a twin of the hydro-turbine and represents its operating state using state vectors. The state vectors are continuously corrected and updated, including: Collect observational data including head, flow rate, unit output, guide vane opening command and feedback, vibration, servo hydraulic pressure, electrical parameters and sediment-related data; perform synchronous calibration on the observational data and assign confidence levels as input for the twin; A physical constraint model is constructed based on the hydraulic and mechanical mechanisms of water turbines. The state vector is then physically consistent and updated based on observation data and corresponding confidence levels, and physical residuals are generated. Based on the structure of the water diversion system, water characteristics, and physical residuals, the propagation delay of the water head is determined, the hysteresis compensation is performed on the components related to the water head in the state vector, and the pressure wave model parameters of the twin are adaptively adjusted. A sediment degradation model is constructed to compensate for the efficiency and vibration-related components in the state vector and predict the performance degradation trend. The vibration changes caused by sediment are fed back to the physical constraint model. Based on the state vector, an ideal characteristic relationship between guide vane opening and power is generated; based on the ideal characteristic relationship, deviation comparison is performed to identify anomalies in the guide vane or servo system, and the guide vane and servo-related parameters in the state vector are updated. Based on the state vector update, the output includes runtime evaluation results, anomaly alarms, historical state backtracking, and performance prediction information.

[0005] As a preferred technical solution of the present invention, the synchronous calibration of the observation data includes: establishing a unified time reference by offset correction and interpolation of the original timestamps of different data sources; achieving consistent sampling frequency for different observation data by upsampling, downsampling or sliding window interpolation based on the unified time reference; and using operational events such as sudden changes in guide vane opening, changes in unit output, sudden changes in head or load shedding as synchronous trigger points to perform event alignment processing on the time series of different types of observation data.

[0006] As a preferred technical solution of the present invention, the allocation of credibility includes: calculating the initial credibility of each type of observation data based on the noise level, signal fluctuation degree, long-term stability and historical reliability of the corresponding sensor; and dynamically adjusting the initial credibility when the observation data experiences short-term interruption, abnormal amplitude fluctuation or significant inconsistency with adjacent data sources.

[0007] As a preferred technical solution of the present invention, the construction of the physical constraint model includes: determining the physical correlation variables between head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters based on the design parameters and operating characteristics of the turbine; establishing constraint equations describing the relationships between head and flow rate, flow rate and unit output, guide vane opening and runner hydrodynamic action, and unit output and electrical parameters based on hydraulic and mechanical transmission mechanisms; and identifying and correcting the parameters in the constraint equations using historical operating data or experimental calibration data of the turbine to form a physical constraint model that performs physical consistency correction on the state vector.

[0008] As a preferred technical solution of the present invention, the generation of physical residuals includes: using a physical constraint model to calculate the physical prediction quantities corresponding to the head, flow rate, unit output, guide vane opening, servo hydraulic pressure and electrical parameters based on the current state vector; comparing the physical prediction quantities with the corresponding observation data that have been synchronously calibrated; obtaining the deviation values ​​of each physical quantity according to predetermined rules; and normalizing and classifying the deviation values ​​according to different physical links of hydraulic, mechanical and electrical energy conversion to form a physical residual vector indicating the consistency of each physical link.

[0009] As a preferred technical solution of the present invention, the hysteresis compensation includes: determining the equivalent hydraulic path length between the head measuring point and the turbine based on the pre-acquired structural parameters of the water diversion system, and obtaining the initial pressure wave propagation characteristics in combination with the calibrated water body characteristic parameters; correcting the initial pressure wave propagation characteristics using the phase offset or time deviation characteristics related to the head in the physical residual, and estimating the propagation delay of the head data; performing time axis translation or resampling processing on the synchronously calibrated head data based on the estimated propagation delay, writing the corrected head data into the head-related component in the state vector, and completing the hysteresis compensation for the head component; and continuously statistically analyzing the correspondence between the physical residual and the propagation delay estimate under multiple operating conditions, and adaptively adjusting the wave velocity and attenuation coefficient of the pressure wave model in the twin.

[0010] As a preferred technical solution of the present invention, the compensation for the efficiency and vibration-related components in the state vector and the prediction of performance degradation trends include: establishing a sediment degradation model based on sediment-related observation data and efficiency changes and vibration responses during historical operation, characterizing the influence of sediment concentration, particle characteristics, and flow velocity on unit efficiency decline and vibration amplitude changes; using the sediment degradation model to calculate the efficiency loss and additional vibration components caused by sediment under the current operating conditions, subtracting or correcting the efficiency loss and additional vibration components from the corresponding efficiency and vibration components in the state vector to obtain the efficiency and vibration characteristics after sediment impact compensation; and based on the cumulative effect of sediment conditions over time, extrapolating the sediment degradation model over multiple operating cycles to predict the efficiency degradation trend and vibration intensification trend in subsequent operating stages, using the vibration characteristic changes caused by sediment to correct the parameters related to vibration response in the physical constraint model.

[0011] As a preferred technical solution of the present invention, the generation of the ideal characteristic relationship between guide vane opening and power includes: selecting the head component, flow component, and efficiency component after hysteresis compensation and sediment degradation compensation, as well as the corresponding guide vane opening component and active power component in the state vector; eliminating operating condition data with abnormal fluctuations; fitting or interpolating the correspondence between guide vane opening and active power based on historical operating data that meets the stable operating conditions of the turbine; and constructing an ideal characteristic curve or function expression form of monotonic change between guide vane opening and active power under given corrected head and efficiency conditions, as the ideal characteristic relationship between guide vane opening and power.

[0012] As a preferred technical solution of the present invention, the deviation comparison based on the ideal characteristic relationship includes: determining the actual guide vane opening trajectory according to the guide vane opening command and guide vane opening feedback; calculating the deviation of the active power under the corresponding operating condition and the ideal characteristic relationship of guide vane opening-power under each operating condition; extracting the amplitude, sign, duration and distribution characteristics of the deviation as it changes with the operating condition; when the deviation shows segmental saturation in the guide vane opening direction or the actual guide vane opening remains basically unchanged when the guide vane opening command changes, it is judged as a guide vane jamming abnormality; when the deviation is mainly manifested as a significant lag interval or hysteresis loop between the guide vane opening feedback signal and the guide vane opening command signal, it is judged as a servo system hysteresis abnormality; and correcting the parameters related to the guide vane opening dynamics and servo response in the state vector according to the determined abnormality type.

[0013] A turbine operation evaluation system based on multi-source data fusion includes: Data acquisition module: Collects observation data including head, flow rate, velocity, guide vane opening command and feedback, vibration, hydraulic pressure, electrical parameters and sediment-related data; performs synchronous calibration on the observation data and assigns a confidence level as input to the twin. Physical constraint module: Based on the hydraulic and mechanical mechanisms of water turbines, a physical constraint model is constructed. Based on the observation data and corresponding confidence level, the state vector is physically consistent, fused and updated, and physical residuals are generated. Pressure wave hysteresis compensation module: Based on the pre-acquired water diversion system structure, calibrated water characteristics and physical residuals, the propagation delay of the water head is determined, the hysteresis compensation is performed on the water head-related components in the state vector, and the pressure wave model parameters of the twin are adaptively adjusted. Sediment degradation module: Constructs a sediment degradation model, compensates for the efficiency and vibration-related components in the state vector and predicts the performance degradation trend, and feeds back the vibration changes caused by sediment to the physical constraint model. Characteristic Relationship Module: Based on the state vector, it generates the ideal characteristic relationship between guide vane opening and power; based on the ideal characteristic relationship, it compares deviations, identifies anomalies in the guide vane or servo system, and updates the guide vane and servo-related parameters in the state vector; Evaluation Results Module: Based on state vector updates, it outputs runtime evaluation results, anomaly alarms, historical state backtracking, and performance prediction information.

[0014] The present invention has the following advantages: This invention constructs a turbine twin represented by a state vector and continuously corrects and updates the state vector, mapping multi-source observation data into the same state space, thus achieving a holistic and quantitative description of the turbine's operating state. By performing offset correction, interpolation, and sampling frequency consistency processing on the original timestamps from different data sources, and by using operating events as synchronization trigger points to achieve event alignment, the time consistency of data fusion is significantly improved.

[0015] This invention achieves adaptive weighted suppression of multi-source measurement quality by allocating initial confidence levels based on the noise level, fluctuation degree, long-term stability, and historical reliability of various types of observation data, and dynamically adjusting the confidence levels when data is interrupted, abnormally fluctuating, or significantly inconsistent with adjacent data sources. This reduces the interference of faulty sensors or abnormal data on state estimation and operational evaluation results.

[0016] This invention establishes a physical constraint model based on hydraulics and mechanical transmission mechanisms, and identifies and corrects the model parameters using historical operating data or experimental calibration data. It performs consistency correction between the state vector and the observed data under physical constraints, thereby improving the fit between the state estimation and the actual operating conditions. The physical constraint model is used to calculate the physical prediction quantity and calculate the deviation with the synchronously calibrated observation data to construct a physical residual vector indicating the consistency degree of different physical links such as hydraulic, mechanical and electrical energy conversion. This enables a segmented quantitative assessment of the health status of each physical link of the unit.

[0017] This invention determines the pressure wave propagation characteristics between the head measuring point and the turbine by combining the structural parameters of the water diversion system and the characteristics of the water body. It estimates the propagation delay by using the phase or time deviation related to the head in the physical residual, performs hysteresis compensation for the head component, and adaptively adjusts the parameters of the twin pressure wave model. This makes the operation evaluation results more consistent with the actual hydraulic process, improving the timeliness and accuracy of the evaluation. By establishing a sediment degradation model, it correlates sediment concentration, particle characteristics, and flow velocity with the decrease in unit efficiency and changes in vibration response, compensates for the efficiency and vibration components in the state vector, and predicts the performance degradation trend based on the time accumulation effect of sediment conditions. This effectively distinguishes long-term performance degradation caused by sediment from short-term anomalies caused by other factors.

[0018] This invention achieves online, precise diagnosis of the guide vane and servo system status by fitting the ideal characteristic relationship between guide vane opening and active power based on stable historical data under corrected head and efficiency conditions after hysteresis compensation and sediment compensation, and by comparing the deviation between the actual operating trajectory and the ideal characteristic curve, and using the amplitude, sign, and timing characteristics of the deviation to identify anomalies. Attached Figure Description

[0019] 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 in the following description are only schematic diagrams of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort. Figure 1 This is a schematic diagram of the structure of a hydro turbine operation evaluation system based on multi-source data fusion, used in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] Example 1: A method for evaluating the operation of a hydro-turbine based on multi-source data fusion, which constructs a twin of the hydro-turbine and represents the operating state of the twin with a state vector, and continuously corrects and updates the state vector, including the following steps: In one embodiment of the present invention, the turbine twin is a digital model structure that maps the actual turbine operation process in real time. It is constructed based on field observation data and physical mechanism models and is used to reproduce the hydraulic, mechanical and electrical responses of the turbine under different operating conditions in virtual space. The state vector is a set of multidimensional variables used to centrally characterize the unit's operating state, including at least the head component, flow component, unit output component, guide vane opening component, servo hydraulic pressure component, electrical parameter component, efficiency component, vibration characteristic component and component reflecting the degree of sediment degradation. The processing results of the observation data in subsequent steps are all written into or updated to this state vector.

[0022] Step S1: Collect observation data including head, flow rate, unit output, guide vane opening command and feedback, vibration, servo hydraulic pressure, electrical parameters and sediment-related data; perform synchronous calibration on the observation data and assign confidence level as input for the twin. In one embodiment of the present invention, the head observation data includes upstream water levels collected by level gauges installed in the forebay or upstream reservoir, downstream water levels collected by level gauges installed in the tailrace or downstream channel, and time series of total head and net head calculated based on the upstream and downstream water levels. The sampling period is set within the range of seconds to reflect the energy supply conditions of the turbine under different operating conditions. The flow observation data includes the unit overflow time series collected by flow meters installed at the pressure pipe or spiral casing of the unit. When necessary, it combines the equivalent flow information derived from the head and unit output to describe the water conveyance capacity of the intake system and the unit loading degree. The unit output observation data includes the unit active power time series recorded by the monitoring system, and reactive power and shaft power are derived by combining electrical parameters when necessary to characterize the result of water energy to electrical energy conversion. The guide vane opening command and feedback observation data include the guide vane opening command signal output by the speed governor and the guide vane opening feedback signal collected by the guide vane stroke sensor. Both are normalized to a relative opening of 0-100%, used to characterize the control commands and actual responses of the guide vane actuator. Vibration observation data includes vibration signals collected by vibration sensors installed in key components such as the main shaft bearing, thrust bearing, and machine base. Statistical calculations within a time window are used to extract vibration characteristic quantities such as the root mean square value of vibration displacement or acceleration and the amplitude of the main frequency band, used to reflect the operational stability of the unit's mechanical components. Servo hydraulic pressure observation data includes the pressure in the main hydraulic station and the time series of pressure in the guide vane servo cylinder cavity, used to reflect the pressure supply of the servo system and the stress state of the actuator. Electrical parameter observation data includes terminal voltage, current, frequency, power factor, and power quality indicators extracted when needed, used to construct the physical correspondence between the power conversion process and the unit's output. Sediment-related observation data includes time series of sediment concentration, representative particle size and particle size distribution obtained from online sediment concentration monitoring devices or water sample analysis, flow velocity information corresponding to the inlet location, and cumulative sediment flux within a certain operating cycle, which are used to characterize the long-term impact of sediment conditions on unit efficiency and vibration.

[0023] The aforementioned observation data are collected by sensors and monitoring devices deployed at the hydropower station site and uploaded in a unified data interface format to form a raw multi-source observation data set, which serves as the basis for subsequent synchronous calibration and reliability allocation.

[0024] The synchronous calibration of the observation data includes: establishing a unified time reference by offset correction and interpolation of the original timestamps from different data sources; achieving consistent sampling frequency for different observation data by using upsampling, downsampling, or sliding window interpolation based on the unified time reference; and using operational events such as sudden changes in guide vane opening, changes in unit output, head jumps, or load shedding as synchronization trigger points to perform event alignment processing on the time series of different types of observation data.

[0025] In one embodiment of the present invention, different types of observation data are recorded with original timestamps by different acquisition devices during the acquisition process, resulting in clock deviations and inconsistencies in the start and end times of recording. By comparing the standard time information recorded by the power monitoring system with the timestamps of each acquisition channel, the original timestamps are offset and corrected. Short-term missing recording points are filled by interpolation to generate a continuous time series. Based on this, a unified time axis is established that spans all observation data. To address the differences in the original sampling periods of observation data such as head, flow rate, vibration, and electrical parameters, a target sampling period is selected based on the unified time axis. Data with sampling periods longer than the target period undergoes upsampling, while data with sampling periods shorter than the target period undergoes downsampling. A sliding window interpolation method is used when necessary to ensure that all types of observation data have a consistent sampling frequency under the unified time axis. To further improve synchronization accuracy, the corresponding event time is identified near operational events such as a step change in guide vane opening command, rapid rise or fall of unit output, significant jump in head, or unit load shedding. This event is used as a trigger for time alignment, and the time series of guide vane opening, head, flow rate, unit output, electrical parameters, and vibration are refined and aligned. This makes the time correspondence of multi-source observations in key dynamic processes more accurate, and the synchronized and calibrated observation data serves as the unified time input for subsequent state vector updates and physical constraint calculations.

[0026] The allocation of credibility includes: calculating the initial credibility for each type of observation data based on the noise level, signal fluctuation, long-term stability, and historical reliability of the corresponding sensor; and dynamically adjusting the initial credibility when the observation data experiences short-term interruption, abnormal amplitude fluctuation, or significant inconsistency with adjacent data sources.

[0027] In one embodiment of the present invention, the initial confidence level is calculated based on the statistical characteristics and maintenance records of each observation channel during the historical operation phase. Measurement points with low noise levels, minimal long-term zero drift, and stable maintenance records are assigned higher confidence levels, while those with large noise fluctuations and frequent fault records are assigned lower confidence levels. Cross-comparison results are introduced among head, flow rate, unit output, and electrical parameters as auxiliary criteria. For example, when the theoretical output derived from head and flow rate maintains long-term consistency with the measured unit output, the confidence level of the output measurement point and related head and flow rate measurement points is increased. During online operation, when an observation channel experiences a short-term data interruption, a sudden change in signal amplitude, or a significant deviation from the trend of other physically related observations, the confidence level of that channel is reduced according to preset rules to weaken its impact on state vector updates and physical constraint model solutions. When the channel recovers stability and regains consistency with other observations in subsequent time periods, its confidence level is gradually restored.

[0028] Step S2: Construct a physical constraint model based on the hydraulic and mechanical mechanism of the turbine, perform physical consistency correction and fusion update on the state vector based on the observation data and corresponding confidence level, and generate physical residuals; In one embodiment of the present invention, the physical constraint model is a set of mechanistic models built based on hydraulic and mechanical transmission mechanisms. It is used to characterize the inherent relationships that physical quantities such as head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters should satisfy under normal operating conditions. The physical constraint model constitutes the core model layer of the turbine twin. Its inputs are the multi-source observation data after synchronous calibration and assignment of confidence in step S1, and the current state vector. The outputs are the predicted values ​​of each physical quantity and the residual quantities reflecting physical consistency. By introducing descriptions of water characteristics and turbine hydrodynamic effects into this model, the key processes of the turbine in water intake, energy conversion, and electrical processes have clear physical orientations within the twin, thus providing a basis for subsequent state vector correction and anomaly identification.

[0029] The construction of the physical constraint model includes: determining the physical correlation variables between head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters based on the turbine's design parameters and operating characteristics; establishing constraint equations based on hydraulics and mechanical transmission mechanisms to describe the relationships between head and flow rate, flow rate and unit output, guide vane opening and runner hydrodynamic action, and unit output and electrical parameters; and identifying and correcting the parameters in the constraint equations using historical turbine operating data or experimental calibration data to form a physical constraint model that performs physical consistency correction on the state vector.

[0030] In one embodiment of the present invention, the physical variables related to head include the upstream water level of the forebay or upstream reservoir, the downstream water level of the tailrace or downstream river channel, and the total head and net head calculated therefrom. The net head, combined with flow rate and unit efficiency, is used to characterize the water energy input. The flow rate-related variables include the unit overflow rate and the section flow rate introduced when needed, used to connect the water intake link and the turbine runner energy conversion link. The guide vane opening-related variables include guide vane opening commands and feedback. The guide vane opening rate determines the available flow area through the guide vanes, thereby affecting the flow distribution entering the turbine runner. The turbine runner hydrodynamic action is the water flow action. The hydrodynamic effect generated by the runner blades converts the kinetic and potential energy of the water into the mechanical energy of the runner. This hydrodynamic effect is represented in the model by an equivalent force or an equivalent rectangular formula acting on the runner, which is used to establish the relationship between the guide vane opening, flow rate, and unit output. Servo hydraulic related variables include the pressure in the main oil pressure station and the pressure in the servo cylinder cavity, which are used to reflect the stress conditions and response characteristics of the guide vane actuator during different opening changes. Electrical parameter related variables include the terminal voltage, current, frequency, and power factor, which are used to describe the correspondence between the electrical energy output by the generator and the mechanical output of the turbine.

[0031] When establishing the constraint equations, based on the hydraulic energy balance relationship, the net head, flow rate, and unit output power are linked through energy conversion efficiency, thus providing the constraint relationship between head, flow rate, and unit output. Combining the geometric relationship between guide vane structural parameters, guide vane opening, and the flow area through the guide vane cross-section, the runner hydrodynamic action is introduced, incorporating the influence of guide vane opening changes on the inflow rate and runner forces into the model, forming a set of mutually coupled hydraulic-mechanical constraints between guide vane opening, flow rate, and unit output. In the mechanical transmission link, intermediate variables such as rotational speed and shaft power are introduced to link the runner hydrodynamic action with the electrical parameters at the generator end. Constraint equations between unit output and generator terminal voltage, current, and power factor are established through electrical machinery relationships to characterize the physical consistency of the energy conversion link. These constraint relationships are implemented in the model in the form of analytical expressions, empirical fitting functions, or piecewise functions. Parameter identification is performed using historical turbine operating data and experimental calibration data to ensure that the model parameters match the structural and operating characteristics of the specific power station unit. During the parameter identification process, representative operating conditions are selected, and the monitored data such as head, flow rate, unit output, guide vane opening, and electrical parameters are substituted into the constraint relationship. The hydraulic loss coefficient, equivalent hydrodynamic coefficient, mechanical loss coefficient, and power conversion correlation coefficient involved are solved and iteratively corrected until the model output is consistent with the historical observation data within the preset error range, so as to form a physical constraint model suitable for subsequent online correction.

[0032] The generation of physical residuals includes: using a physical constraint model to calculate the physical prediction quantities corresponding to head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters based on the current state vector; comparing the physical prediction quantities with the corresponding observation data that have been synchronously calibrated; calculating the deviation values ​​of each physical quantity according to predetermined rules; and normalizing and classifying the deviation values ​​according to different physical links of hydraulic, mechanical, and electrical energy conversion to form a physical residual vector indicating the degree of consistency of each physical link.

[0033] In one embodiment of the present invention, at each calculation moment, the current head component, flow rate component, guide vane opening component, servo hydraulic pressure component, electrical parameter component, and derived components such as efficiency are read from the state vector and input into the physical constraint model to obtain the predicted head, flow rate, unit output, guide vane hydrodynamic related predicted quantities, servo hydraulic pressure predicted values, and electrical parameter predicted values ​​derived by the model under the current operating conditions. Subsequently, the measured values ​​corresponding to the above predicted quantities at the same time point are extracted from the observation data synchronously calibrated and assigned confidence in step S1. The deviation values ​​of each physical quantity are calculated according to the fixed rule of "measured values ​​minus predicted values" or "predicted values ​​minus measured values". The deviation values ​​are weighted according to the confidence of each observation channel so that the contribution of high confidence observations to the deviation results is higher than that of low confidence observations. For the hydraulic components, head deviation and flow deviation are combined to reflect the physical consistency of the water intake and energy input. For the mechanical components, the unit output deviation, hydrodynamic deviation related to the guide vanes, and shaft vibration characteristic deviation are compiled to characterize the degree of operational deviation of the runner and transmission components. For the electrical energy conversion components, the deviations between the theoretical electrical parameters derived from the unit output and the measured voltage, current, and power factor are summarized to reflect the physical consistency of the generator and its electrical side. By normalizing the deviation values ​​of each component, deviations of different dimensions and magnitudes are compressed to a unified numerical scale and categorized into the three physical components of hydraulic, mechanical, and electrical energy conversion, forming a physical residual vector.

[0034] Step S3: Based on the structure of the water diversion system, water characteristics and physical residuals, determine the propagation delay of the water head, perform hysteresis compensation on the components related to the water head in the state vector, and adaptively adjust the pressure wave model parameters of the twin. In one embodiment of the present invention, the water diversion system structure comprises the hydraulic channel structure information from the upstream head measuring point to the turbine inlet, including the length of the pressure steel pipe, pipe diameter variations, number and arrangement of bends, form of branching pipe sections, and geometric parameters connecting to structures such as the forebay, bends, and spiral casing. This information characterizes the equivalent hydraulic path traversed by the head as it propagates along the water diversion path. The water characteristics are a set of physical parameters of the water within the aforementioned water diversion path, including water temperature, density, and gas content during operation. Water temperature and density affect the compressibility of the water, while gas content affects the effective elastic properties of pressure wave propagation in the water. These water characteristics, together with the geometric parameters of the water diversion path, determine the pressure wave velocity and attenuation pattern of the head propagating in the water diversion channel.

[0035] The hysteresis compensation includes: determining the equivalent hydraulic path length between the head measuring point and the turbine based on the pre-acquired structural parameters of the water diversion system, and obtaining the initial pressure wave propagation characteristics in combination with the calibrated water body characteristic parameters; correcting the initial pressure wave propagation characteristics using the phase offset or time deviation characteristics related to the head in the physical residual, and estimating the propagation delay of the head data; performing time axis translation or resampling processing on the synchronously calibrated head data based on the estimated propagation delay, writing the corrected head data into the head-related component in the state vector, and completing the hysteresis compensation for the head component; and continuously statistically analyzing the correspondence between the physical residual and the propagation delay estimate under multiple operating conditions, and adaptively adjusting the wave velocity and attenuation coefficient of the pressure wave model in the twin.

[0036] In one embodiment of the present invention, given the known structural parameters of the water intake system, the hydraulic path from the upstream head measuring point to the turbine inlet is discretized into several equivalent pipe segments based on the layout information of the pressure steel pipe, elbows, branch pipe sections, and connected structures such as the forebay and spiral casing. The geometric lengths of these equivalent pipe segments are accumulated, and the additional hydraulic lengths of local structures are included when necessary, to obtain the equivalent hydraulic path length of the head propagating from the measuring point to the turbine. Then, combining the water body characteristic parameters obtained from on-site calibration or pre-operation tests, physical properties such as water temperature, density, and gas content are substituted into the pressure wave propagation characteristic calculation model to obtain the initial pressure wave propagation velocity and friction loss characteristics under the current water conditions. Based on this, the theoretical propagation delay of the head from the measuring point to the unit's operating location is further derived. The aforementioned initial propagation delay is only based on the design structure and calibrated water body characteristics. After long-term operation or changes in operating conditions, it deviates from the actual propagation behavior. Therefore, it needs to be corrected using the time-series characteristics related to the head in the physical residual.

[0037] During the online operation phase, the head-related physical residuals obtained in step S2 are used to time-match the head predicted by the physical constraint model with the synchronously calibrated measured head sequence. By comparing the waveforms before and after typical events such as head jumps, load shedding, or rapid guide vane changes, the phase shift characteristics of the head component in the residuals are analyzed. Alternatively, the cross-correlation function between the predicted and measured head sequences is calculated to identify the time offset corresponding to the strongest correlation position, which is then used as an estimate of the actual propagation delay. After obtaining the propagation delay estimate of the head data, the head time series synchronously calibrated in step S1 is shifted along the time axis according to the estimated delay. This aligns the head changes originally corresponding to the upstream measuring point with the unit response time in time. Alternatively, in high-precision scenarios, a time-delay-based resampling method is used to interpolate and reconstruct the head data, ensuring that the head sequence is consistent with the flow rate, unit output, and guide vane opening sequences in time. After the lag compensation is completed, the corrected head data is written into the head-related component of the state vector to replace the uncompensated original head value, so that the twin can use the head information that has taken into account the propagation delay in subsequent calculations to solve physical constraints and evaluate performance.

[0038] To improve the consistency of the pressure wave model across different operating conditions, the correspondence between physical residuals and propagation delay estimates under various typical operating conditions is continuously statistically analyzed and summarized. Specifically, this includes recording the theoretical propagation delay calculated based on structural and water characteristics and the actual propagation delay identified based on physical residuals within representative operating conditions. The deviation between the two is calculated, and this deviation is used to determine whether the current pressure wave model parameters match the actual operating conditions. When the deviation between the actual and theoretical propagation delays consistently exceeds a pre-set deviation threshold in the same direction for several consecutive operating conditions, it is considered that the parameters used in the current pressure wave model to calculate the pressure wave propagation velocity and attenuation characteristics are no longer suitable for the existing water conditions or water diversion channel status, and the pressure wave model parameters in the twin need to be adaptively adjusted. During the adaptive adjustment process, the equivalent elastic parameters of the water body, the equivalent density of the water body, or the correction coefficients related to the gas content are updated based on the deviation amplitude between the actual propagation delay and the theoretical propagation delay, thereby indirectly correcting the calculation results of the pressure wave propagation speed. At the same time, the attenuation coefficient in the pressure wave model is corrected according to the attenuation change of the physical residual in different frequency bands.

[0039] Step S4: Construct a sediment degradation model, compensate for the components related to efficiency and vibration in the state vector and predict the performance degradation trend, and feed back the vibration changes caused by sediment to the physical constraint model. The compensation for efficiency and vibration-related components in the state vector and the prediction of performance degradation trends include: establishing a sediment degradation model based on sediment-related observation data and efficiency changes and vibration responses during historical operation, characterizing the influence of sediment concentration, particle characteristics, and flow velocity on unit efficiency decline and vibration amplitude changes; using the sediment degradation model to calculate the efficiency loss and additional vibration components caused by sediment under the current operating conditions, subtracting or correcting the efficiency loss and additional vibration components from the corresponding efficiency and vibration components in the state vector to obtain the efficiency and vibration characteristics after sediment impact compensation; and based on the cumulative effect of sediment conditions over time, extrapolating the sediment degradation model over multiple operating cycles to predict the efficiency degradation trend and vibration intensification trend in subsequent operating stages, using the vibration characteristic changes caused by sediment to correct the parameters related to vibration response in the physical constraint model.

[0040] In one embodiment of the present invention, during the construction phase of the sediment degradation model, a representative historical operating section is selected, and data on sediment concentration, representative particle size, flow velocity, and cumulative sediment flux within that section are collected. Simultaneously, the unit efficiency characteristics and vibration characteristics for the corresponding time period are extracted. Efficiency characteristics include the overall efficiency calculated from head, flow rate, and unit output, as well as efficiency variation curves under different loads. Vibration characteristics include the root mean square value of vibration of the main shaft and base, the vibration amplitude of the main frequency band, and trend indicators over time. By analyzing the variation patterns of efficiency and vibration under different sediment conditions, a degradation relationship model is constructed, with inputs including sediment concentration, particle size, flow velocity, cumulative sediment flux, and operating time, and outputs efficiency loss and additional vibration components. The specific form of the model employs a semi-empirical relationship based on physical experience, assigning greater efficiency loss and vibration increment response to operating conditions with high sediment content, hard particles, or large particle size; or it uses a regression model and an ensemble learning model trained on historical data, learning the correspondence between sediment conditions and efficiency and vibration changes in a large number of historical samples to obtain the mapping relationship used to predict the amount of efficiency loss and additional vibration components. After the model training and calibration are completed, the obtained sediment degradation model is solidified in the model layer of the turbine twin for online calculation of sediment impact.

[0041] During the online operation phase, for each calculation cycle, the sediment concentration, representative particle size, flow velocity, and cumulative sediment flux within the current operating cycle are read from the sediment-related observation data synchronized and calibrated in step S1. These variables, along with auxiliary variables such as the operating time stored in the state vector, are input into the sediment degradation model to obtain the efficiency loss and additional vibration components caused by sediment at that moment. Subsequently, the efficiency and vibration components without sediment compensation are read from the current state vector. The aforementioned efficiency loss is subtracted from the efficiency components or adjusted according to preset correction rules to obtain the corrected efficiency characteristics after removing the influence of sediment degradation. At the same time, the additional vibration components are subtracted from the vibration components or corrected accordingly to obtain vibration characteristics that are closer to the mechanical state of the structure itself. The efficiency and vibration components after sediment impact compensation are written back to the state vector for subsequent physical constraint model solving and the construction of the guide vane-power ideal characteristic relationship, so that the long-term performance degradation caused by sediment under high sediment concentration conditions is distinguished from the performance changes caused by control anomalies or short-term disturbances in the state space.

[0042] Over multiple operating cycles, sediment concentration, particle size, flow velocity, and cumulative sediment flux are continuously recorded at different time points. Combined with efficiency compensation and vibration correction values ​​for the corresponding time periods, this information is input into a sediment degradation model as a time series to extrapolate and predict efficiency degradation and vibration aggravation trends under given sediment conditions in future operating phases. For example, in operating scenarios with prolonged high sediment concentration, the efficiency loss and additional vibration components given by the sediment degradation model show a gradually increasing trend on the time axis. A component representing the degree of sediment accumulation degradation can be added to the state vector to record the long-term performance degradation of the unit relative to its initial state. Based on the predicted efficiency degradation curve and vibration development trend, the operational evaluation results indicate the risk of a significant decrease in unit efficiency or vibration approaching the operating limit within a future time window, providing a reference for optimizing unit economic operation and maintenance plans.

[0043] When feeding back vibration characteristic changes caused by sediment into the physical constraint model, the additional vibration components output by the sediment degradation model are used as additional excitation effects acting on the turbine blades, bearings, and base. Specifically, in the vibration response-related parts of the physical constraint model, the additional vibration components are transformed into correction factors for the structural force balance relationship. By adjusting the vibration-related equivalent damping parameters, stiffness parameters, or hydrodynamic imbalance coefficients, the vibration response predicted by the physical constraint model under sediment-laden conditions is made closer to the observed vibration characteristics after sediment compensation in terms of amplitude and frequency band characteristics.

[0044] Step S5: Based on the state vector, generate the ideal characteristic relationship between guide vane opening and power; compare the deviations based on the ideal characteristic relationship, identify abnormalities in the guide vane or servo system, and update the guide vane and servo-related parameters in the state vector. The process of generating the ideal characteristic relationship between guide vane opening and power includes: selecting the head component, flow component, and efficiency component after hysteresis compensation and sediment degradation compensation, as well as the corresponding guide vane opening component and active power component in the state vector; eliminating operating condition data with abnormal fluctuations; fitting or interpolating the correspondence between guide vane opening and active power based on historical operating data that meets the stable operating conditions of the turbine; and constructing an ideal characteristic curve or function expression for the monotonic change between guide vane opening and active power under given corrected head and efficiency conditions, which serves as the ideal characteristic relationship between guide vane opening and power.

[0045] The deviation comparison based on ideal characteristic relationship includes: determining the actual guide vane opening trajectory according to the guide vane opening command and guide vane opening feedback; calculating the deviation between the active power under the corresponding operating condition and the ideal characteristic relationship between guide vane opening and power under each operating condition; extracting the amplitude, sign, duration, and distribution characteristics of the deviation as it changes with the operating condition; when the deviation shows segmental saturation in the guide vane opening direction or the actual guide vane opening remains basically unchanged when the guide vane opening command changes, it is judged as a guide vane jamming anomaly; when the deviation is mainly manifested as a significant lag interval or hysteresis loop between the guide vane opening feedback signal and the guide vane opening command signal, it is judged as a servo system hysteresis anomaly; and correcting the parameters related to guide vane opening dynamics and servo response in the state vector according to the determined anomaly type.

[0046] In one embodiment of the present invention, the ideal characteristic relationship between guide vane opening and power is the monotonic correspondence that the guide vane opening and the active power of the unit should exhibit under stable operating conditions, given a corrected head and efficiency. It is used to characterize the standard output characteristics of the unit when there are no abnormalities in the guide vane mechanism and servo system, and the effects of sediment and head lag have been compensated.

[0047] When generating the ideal characteristic relationship between guide vane opening and power, operating conditions representing the normal and stable operation of the turbine are selected from the state vector and historical observation data. These include time intervals where the unit load is stable, the vibration level is within the normal range, and the sediment condition does not show any abnormal degradation after compensation in step S4. Within these operating conditions, the head component after hysteresis compensation and the efficiency component after sediment degradation compensation are read from the state vector, and combined with the flow component, guide vane opening component, and active power component at the corresponding time, to form a sample set for fitting. To avoid mixing short-term disturbances, measurement noise, or control anomalies into the ideal characteristic construction process, abnormal operating conditions in the sample set are removed, including periods where there is a significant inconsistency between the guide vane opening command and feedback, periods where active power changes abruptly in a short period of time, and periods where vibration characteristics deviate significantly from normal levels.

[0048] Based on the sample set, a monotonic relationship between guide vane opening and active power is established through fitting or interpolation. Specifically, within a pre-defined guide vane opening interval, the guide vane opening component is used as the independent variable, and the active power component at the corresponding moment is used as the dependent variable. The corresponding relationship is smoothed to ensure that the resulting guide vane opening-power curve maintains a monotonic characteristic of increasing active power as the guide vane opening increases within the considered operating range. In some unit load ranges, if there is a saturation section close to the rated opening interval, the power change trend in this region can be smoothed in the ideal characteristic relationship to ensure that the ideal characteristic curve maintains overall continuity and clear physical meaning. Using the above fitting or interpolation results, a one-dimensional ideal characteristic relationship is formed with guide vane opening as input and active power as output. This relationship is stored as a guide vane opening-power ideal characteristic model within a twin, providing a standard reference for subsequent deviation comparison and anomaly identification.

[0049] When comparing deviations based on ideal characteristic relationships, the actual guide vane opening trajectory is first extracted from the guide vane opening command and guide vane opening feedback time series. This trajectory is a curve formed by arranging the guide vane opening feedback signals in chronological order, with key operational events (such as rapid guide vane opening, rapid closing, or sudden load changes) highlighted to facilitate analysis of deviation behavior under different operating conditions. For each operating point on the trajectory, based on the corrected head and efficiency components in the state vector, the operating condition at that point is assigned to the corresponding ideal characteristic reference condition. The ideal active power value corresponding to the same guide vane opening and reference condition is retrieved from the guide vane opening-power ideal characteristic relationship and compared with the actual active power at that point to calculate the guide vane opening-power deviation. The deviation calculation follows a unified rule, and the magnitude, sign, duration, and distribution of the deviation across the entire guide vane operating range are recorded, thus forming a complete deviation feature set describing the deviation of the guide vane-power behavior from the ideal characteristics under the current operating state.

[0050] Regarding the identification of guide vane jamming anomalies, guide vane jamming anomalies are defined as a state in which, after the speed control system issues a guide vane opening change command, the actual guide vane opening shows significant stagnation or stagnation within a certain opening range, resulting in insufficient active power response to command changes. Specifically, when, within a certain guide vane opening range, the deviation of the actual active power from the ideal power value given by the ideal characteristic relationship is continuously greater than a preset power deviation threshold for a period of time, and the rate of change of the guide vane opening feedback is lower than a preset minimum rate of change threshold, and the above state remains unchanged in multiple consecutive sampling periods, it is determined that there is a segmental saturation phenomenon in that guide vane opening range, i.e., the guide vane is jamming anomaly within that range. For servo system hysteresis anomalies, servo system hysteresis anomalies are defined as a significant time lag or obvious hysteresis characteristic in the response of the guide vane opening feedback to the guide vane opening command, causing the guide vane opening change trajectory to not coincide with the path during command increase and command decrease processes. In specific judgment, by analyzing the time series relationship between the guide vane opening command and the feedback, when the guide vane opening feedback has a long period of stagnation after the command changes or forms an obvious hysteresis loop when the guide vane opening command changes back and forth, and the corresponding guide vane opening-power deviation shows a paired distribution with opposite directions but similar amplitudes in this stage, it is determined that the servo system has a hysteresis abnormality.

[0051] Step S6: Based on the state vector update, output the running evaluation results, anomaly alarms, historical state backtracking and performance prediction information.

[0052] The updated state vector, as a centralized carrier of various observation data processing results and model calculation results, already includes information such as head components, flow components, unit output components, guide vane opening components, servo hydraulic pressure components, electrical parameter components, efficiency components, vibration characteristic components, pressure wave model parameter components, and sediment degradation degree components. It also includes the physical residual vector formed in steps S2 to S5, guide vane and servo-related abnormal parameters, and corrected state quantities after hysteresis compensation and sediment compensation.

[0053] The operational evaluation results provide comprehensive health and performance assessment information for the current or near-real-time operating conditions, encompassing multiple dimensions including hydraulic, mechanical, electrical conversion, and control execution. Specifically, based on head, flow, and efficiency components, hydraulic performance indicators such as water energy utilization efficiency and the degree of deviation of the unit's operating point from the high-efficiency zone are provided; based on vibration characteristic components and servo hydraulic pressure components, mechanical and servo performance indicators such as shaft vibration level, base vibration level, and servo execution stability are provided; based on electrical parameter components and unit output components, electrical performance indicators such as voltage deviation, power factor deviation, and electrical output stability are provided; and based on the normalized values ​​of the physical residual vectors at each physical stage, consistency evaluation indicators such as hydraulic consistency level, mechanical consistency level, and electrical conversion consistency level are provided.

[0054] The abnormal alarms are the results of marking and prompting abnormal states or trend anomalies detected during the operation evaluation process. These include hydraulic, mechanical, and electrical anomaly alarms triggered by physical residual exceeding limits, guide vane jamming anomaly alarms and servo hysteresis anomaly alarms triggered by guide vane opening-power deviation characteristics, and sediment condition exceeding limit alarms given by the sediment degradation model. Specifically: when the head-related or flow-related residuals continuously exceed the preset hydraulic deviation threshold within a certain time window, an abnormal alarm related to hydraulic inconsistency is output; when the output-related residuals and vibration characteristic deviations show that the mechanical components deviate from the normal level, a mechanical abnormal alarm is output; when the deviation between the theoretical electrical parameters derived from the unit output and the measured electrical parameters exceeds the preset range, an energy conversion abnormal alarm is output; when guide vane jamming or servo hysteresis is identified in step S5, the corresponding abnormality type, the guide vane opening range, and the duration are recorded in the abnormal alarm; when the sediment concentration, cumulative sediment flux, or sediment degradation degree component exceeds the safe operating boundary, a sediment condition abnormal alarm is output.

[0055] The historical state backtracking is a process of reproducing and analyzing the unit's state during past operating phases based on the continuous recording of state vectors in the time dimension. Specifically, it includes: archiving the state vector at fixed time intervals or based on operating events (such as load changes, start-up / shutdown, and load shedding events) to form a state vector time series; during historical state backtracking, according to the time interval specified by the user or upper-level application, reading the data for the corresponding time period from the archived state vector series, and reconstructing the evolution trajectory of states such as head, flow rate, unit output, guide vane opening, servo hydraulic pressure, electrical parameters, efficiency, vibration characteristics, and physical residuals within that time period.

[0056] The performance prediction information is a prediction of the unit's efficiency, vibration level, and operational risk trends over a certain future time period based on the current state vector. The performance prediction information relies on the sediment degradation model constructed in step S4, the adaptively adjusted pressure wave model parameters in step S3, and the temporal variation characteristics of the physical residuals to extrapolate the unit's performance evolution under a given scheduling scheme or typical operating strategy. Specifically, it includes: after inputting the expected scheduling conditions or typical operating scenarios for a future period, using the current sediment degradation degree component, sediment concentration trend, and the development direction of cumulative sediment flux, predicting the efficiency loss and additional vibration components for multiple future operating cycles, and forming future efficiency curves and vibration intensity curves accordingly; combining the adaptive results of the pressure wave model parameters, estimating the consistency of future head dynamic response; and based on the historical variation patterns of physical residuals in hydraulic, mechanical, and electrical energy conversion stages, providing a trend judgment on whether the future residual level tends to increase, remain stable, or gradually decrease.

[0057] Example 2: A turbine operation evaluation system based on multi-source data fusion, see [link / reference] Figure 1 As shown, it includes the following modules: Data acquisition module: Collects observation data including head, flow rate, velocity, guide vane opening command and feedback, vibration, hydraulic pressure, electrical parameters and sediment-related data; performs synchronous calibration on the observation data and assigns a confidence level as input to the twin. Physical constraint module: Based on the hydraulic and mechanical mechanisms of water turbines, a physical constraint model is constructed. Based on the observation data and corresponding confidence level, the state vector is physically consistent, fused and updated, and physical residuals are generated. Pressure wave hysteresis compensation module: Based on the pre-acquired water diversion system structure, calibrated water characteristics and physical residuals, the propagation delay of the water head is determined, the hysteresis compensation is performed on the water head-related components in the state vector, and the pressure wave model parameters of the twin are adaptively adjusted. Sediment degradation module: Constructs a sediment degradation model, compensates for the efficiency and vibration-related components in the state vector and predicts the performance degradation trend, and feeds back the vibration changes caused by sediment to the physical constraint model. Characteristic Relationship Module: Based on the state vector, it generates the ideal characteristic relationship between guide vane opening and power; based on the ideal characteristic relationship, it compares deviations, identifies anomalies in the guide vane or servo system, and updates the guide vane and servo-related parameters in the state vector; Evaluation Results Module: Based on state vector updates, it outputs runtime evaluation results, anomaly alarms, historical state backtracking, and performance prediction information.

[0058] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the operation of a hydro-turbine based on multi-source data fusion, characterized in that, A twin of the water turbine is constructed, and its operating state is represented by a state vector. The state vector is continuously corrected and updated, including: Collect observational data including head, flow rate, unit output, guide vane opening command and feedback, vibration, servo hydraulic pressure, electrical parameters and sediment-related data; perform synchronous calibration on the observational data and assign confidence levels as input for the twin; A physical constraint model is constructed based on the hydraulic and mechanical mechanisms of water turbines. The state vector is then physically consistent and updated based on observation data and corresponding confidence levels, and physical residuals are generated. Based on the structure of the water diversion system, water characteristics, and physical residuals, the propagation delay of the water head is determined, the hysteresis compensation is performed on the components related to the water head in the state vector, and the pressure wave model parameters of the twin are adaptively adjusted. A sediment degradation model is constructed to compensate for the efficiency and vibration-related components in the state vector and predict the performance degradation trend. The vibration changes caused by sediment are fed back to the physical constraint model. Based on the state vector, an ideal characteristic relationship between guide vane opening and power is generated; based on the ideal characteristic relationship, deviation comparison is performed to identify anomalies in the guide vane or servo system, and the guide vane and servo-related parameters in the state vector are updated. Based on the state vector update, the output includes runtime evaluation results, anomaly alarms, historical state backtracking, and performance prediction information.

2. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The synchronous calibration of the observation data includes: establishing a unified time reference by offset correction and interpolation of the original timestamps from different data sources; achieving consistent sampling frequency for different observation data by using upsampling, downsampling, or sliding window interpolation based on the unified time reference; and using operational events such as sudden changes in guide vane opening, changes in unit output, head jumps, or load shedding as synchronization trigger points to perform event alignment processing on the time series of different types of observation data.

3. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The allocation of credibility includes: calculating the initial credibility for each type of observation data based on the noise level, signal fluctuation, long-term stability, and historical reliability of the corresponding sensor; and dynamically adjusting the initial credibility when the observation data experiences short-term interruption, abnormal amplitude fluctuation, or significant inconsistency with adjacent data sources.

4. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The construction of the physical constraint model includes: determining the physical correlation variables between head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters based on the turbine's design parameters and operating characteristics; establishing constraint equations based on hydraulics and mechanical transmission mechanisms to describe the relationships between head and flow rate, flow rate and unit output, guide vane opening and runner hydrodynamic action, and unit output and electrical parameters; and identifying and correcting the parameters in the constraint equations using historical turbine operating data or experimental calibration data to form a physical constraint model that performs physical consistency correction on the state vector.

5. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The generation of physical residuals includes: using a physical constraint model to calculate the physical prediction quantities corresponding to head, flow rate, unit output, guide vane opening, servo hydraulic pressure, and electrical parameters based on the current state vector; comparing the physical prediction quantities with the corresponding observation data that have been synchronously calibrated; calculating the deviation values ​​of each physical quantity according to predetermined rules; and normalizing and classifying the deviation values ​​according to different physical links of hydraulic, mechanical, and electrical energy conversion to form a physical residual vector indicating the degree of consistency of each physical link.

6. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The hysteresis compensation includes: determining the equivalent hydraulic path length between the head measuring point and the turbine based on the pre-acquired structural parameters of the water diversion system, and obtaining the initial pressure wave propagation characteristics in combination with the calibrated water body characteristic parameters; correcting the initial pressure wave propagation characteristics using the phase offset or time deviation characteristics related to the head in the physical residual, and estimating the propagation delay of the head data; performing time axis translation or resampling processing on the synchronously calibrated head data based on the estimated propagation delay, writing the corrected head data into the head-related component in the state vector, and completing the hysteresis compensation for the head component; and continuously statistically analyzing the correspondence between the physical residual and the propagation delay estimate under multiple operating conditions, and adaptively adjusting the wave velocity and attenuation coefficient of the pressure wave model in the twin.

7. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The compensation for efficiency and vibration-related components in the state vector and the prediction of performance degradation trends include: establishing a sediment degradation model based on sediment-related observation data and efficiency changes and vibration responses during historical operation, characterizing the influence of sediment concentration, particle characteristics, and flow velocity on unit efficiency decline and vibration amplitude changes; using the sediment degradation model to calculate the efficiency loss and additional vibration components caused by sediment under the current operating conditions, subtracting or correcting the efficiency loss and additional vibration components from the corresponding efficiency and vibration components in the state vector to obtain the efficiency and vibration characteristics after sediment impact compensation; and based on the cumulative effect of sediment conditions over time, extrapolating the sediment degradation model over multiple operating cycles to predict the efficiency degradation trend and vibration intensification trend in subsequent operating stages, using the vibration characteristic changes caused by sediment to correct the parameters related to vibration response in the physical constraint model.

8. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The process of generating the ideal characteristic relationship between guide vane opening and power includes: selecting the head component, flow component, and efficiency component after hysteresis compensation and sediment degradation compensation, as well as the corresponding guide vane opening component and active power component in the state vector; eliminating operating condition data with abnormal fluctuations; fitting or interpolating the correspondence between guide vane opening and active power based on historical operating data that meets the stable operating conditions of the turbine; and constructing an ideal characteristic curve or function expression for the monotonic change between guide vane opening and active power under given corrected head and efficiency conditions, which serves as the ideal characteristic relationship between guide vane opening and power.

9. The method for evaluating the operation of a hydro-turbine based on multi-source data fusion according to claim 1, characterized in that, The deviation comparison based on ideal characteristic relationship includes: determining the actual guide vane opening trajectory according to the guide vane opening command and guide vane opening feedback; calculating the deviation between the active power under the corresponding operating condition and the ideal characteristic relationship between guide vane opening and power under each operating condition; extracting the amplitude, sign, duration, and distribution characteristics of the deviation as it changes with the operating condition; when the deviation shows segmental saturation in the guide vane opening direction or the actual guide vane opening remains basically unchanged when the guide vane opening command changes, it is judged as a guide vane jamming anomaly; when the deviation is mainly manifested as a significant lag interval or hysteresis loop between the guide vane opening feedback signal and the guide vane opening command signal, it is judged as a servo system hysteresis anomaly; and correcting the parameters related to guide vane opening dynamics and servo response in the state vector according to the determined anomaly type.

10. A turbine operation evaluation system based on multi-source data fusion, characterized in that, The system applies a multi-source data fusion-based turbine operation evaluation method as described in any one of claims 1 to 9, including: Data acquisition module: Collects observation data including head, flow rate, velocity, guide vane opening command and feedback, vibration, hydraulic pressure, electrical parameters and sediment-related data; performs synchronous calibration on the observation data and assigns a confidence level as input to the twin. Physical constraint module: Based on the hydraulic and mechanical mechanisms of water turbines, a physical constraint model is constructed. Based on the observation data and corresponding confidence level, the state vector is physically consistent, fused and updated, and physical residuals are generated. Pressure wave hysteresis compensation module: Based on the pre-acquired water diversion system structure, calibrated water characteristics and physical residuals, the propagation delay of the water head is determined, the hysteresis compensation is performed on the water head-related components in the state vector, and the pressure wave model parameters of the twin are adaptively adjusted. Sediment degradation module: Constructs a sediment degradation model, compensates for the efficiency and vibration-related components in the state vector and predicts the performance degradation trend, and feeds back the vibration changes caused by sediment to the physical constraint model. Characteristic Relationship Module: Based on the state vector, it generates the ideal characteristic relationship between guide vane opening and power; based on the ideal characteristic relationship, it compares deviations, identifies anomalies in the guide vane or servo system, and updates the guide vane and servo-related parameters in the state vector; Evaluation Results Module: Based on state vector updates, it outputs runtime evaluation results, anomaly alarms, historical state backtracking, and performance prediction information.

Images