A water level early warning method and system for a ship lift of a hydropower station
By using a multivariate nonlinear regression model and data preprocessing, the problem of water level prediction for ship lifts in hydropower stations under shedding and load shedding conditions was solved, achieving high-precision early warning and safe and efficient equipment operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THREE GORGES JINSHAJIANG CHUANYUN HYDROPOWER DEV CO LTD
- Filing Date
- 2026-03-23
- Publication Date
- 2026-06-05
AI Technical Summary
Existing water level early warning methods lack sufficient prediction accuracy under transient conditions such as ship lift shedding and load shedding at hydropower stations. They cannot accurately capture nonlinear dynamic characteristics, resulting in poor model stability, weak adaptability to operating conditions, and an inability to achieve an optimal balance between safety and efficiency.
A multivariate nonlinear regression model is adopted. Through data preprocessing (centralization and natural logarithm transformation) and differential model construction, combined with a four-level verification process, an early warning method adapted to the machine cut-off and load shedding conditions is constructed, including data acquisition, variable preprocessing, model fitting and early warning output.
It achieves high-precision prediction of water level fluctuation and over-limit duration, significantly improving prediction accuracy and model generalization ability, reducing the frequency of auxiliary gate operation, extending equipment life and reducing energy consumption.
Smart Images
Figure CN122149589A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of water level early warning technology for ship lifts in hydropower stations, specifically to a method and system for water level early warning under special transient conditions such as ship lift shedding and load shedding in hydropower stations. In particular, it relates to a method for fitting early warning indications of water level fluctuation and over-limit duration based on variable characteristic preprocessing and multivariate nonlinear regression. Background Technology
[0002] The ship lift is a crucial navigation facility at a hydropower station, and maintaining a stable water level in its lock chamber is essential for its safe and efficient operation. During transient conditions such as turbine shedding (sudden load reduction) or load relinquishment (sudden loss of load), the downstream flow rate changes drastically, causing significant fluctuations in the ship lift lock chamber water level within a short period. If the water level fluctuation (… The duration of excessive or oversized (e.g., exceeding 0.5 meters) Excessive length will seriously affect navigation safety and may trigger frequent opening and closing of auxiliary lock chambers, exacerbating equipment wear and tear.
[0003] Existing water level early warning and prediction methods mostly adopt linear regression models. The core structure uses the initial water level and flow rate variation as key input variables. By collecting historical measured water level and flow rate data, a prediction formula is constructed using a simple linear fitting method. Its working principle is based on the linear relationship between variables to predict the water level fluctuation after the change of operating conditions, thereby providing a reference for the operation of auxiliary gate chambers and gates.
[0004] However, this existing technology has the following technical defects and limitations: Prediction accuracy is severely insufficient: Shedding and load shedding conditions are essentially strong nonlinear transient processes, and the relationship between water level response and the initial water level and flow rate change is not a simple linear one. Existing linear models cannot capture these complex nonlinear dynamic characteristics, leading to poor goodness of fit (…). ) is usually below 0.9, for and The predicted values deviate significantly from the actual values, making it difficult to meet the needs of high-precision early warning.
[0005] The variable handling method is crude: existing technologies directly input raw data into the model, ignoring the characteristics of the data itself. On the one hand, the flow amplitude ( The data range can be extremely large (e.g., from hundreds to thousands). Large values will dominate model training, drowning out the influence of other variables; on the other hand, the initial water level ( The water level has a clear physical dimension and a sensitive threshold effect, and directly using the original values cannot highlight its key role in sudden changes in water level. This crude treatment leads to poor model stability and weak generalization ability.
[0006] The existing model has a simple structure and poor adaptability to different operating conditions: it adopts a fixed linear model structure, fails to distinguish the differences in hydraulic response mechanisms under two different disturbance source conditions, namely turbine shedding and load shedding, and does not consider the acceleration of flow rate change. This is a key transient characteristic. Therefore, the model's predictive ability rapidly declines and its adaptability is poor for operating conditions outside the training data, especially for new water level-flow combinations.
[0007] Limited engineering guidance value: Due to large prediction errors, existing methods cannot reliably predict the specific time window when water levels exceed the safety threshold. This leads to a conservative approach to operation and regulation: either auxiliary gates are put into operation too early and too frequently to ensure safety, increasing equipment wear and energy consumption; or the risk of water levels exceeding limits is faced due to delayed early warnings, making it impossible to achieve an optimal balance between safety and efficiency.
[0008] The core reason for the above problems is that the existing technology has not been specifically designed for the transient water flow characteristics under the conditions of machine shedding and load shedding. It only uses the linear prediction logic of ordinary conditions, which fails to match the nonlinear response law under complex conditions and lacks in-depth optimization processing of variable characteristics. Summary of the Invention
[0009] The purpose of this invention is to provide a water level early warning method and system for ship lifts in hydropower stations, addressing the aforementioned problems.
[0010] The technical solution of the present invention is as follows: A water level early warning method for a ship lift in a hydroelectric power station includes the following steps: Data acquisition: Collect initial water levels under historical operating conditions, verification operating conditions, and newly added operating conditions. Flow rate fluctuation Variable flow rate Measured water level fluctuation within 20 minutes Measured value of the duration of water level exceeding the limit Actual measured data; Variable preprocessing: For the initial water level Centralized processing, i.e., computation ,in This refers to the water level change sensitivity threshold determined through cluster analysis of historical data; For flow rate variation Perform the natural logarithmic transformation, i.e., calculate ; Preserve Variable Flow The original measured value; Early warning index fitting: Input the preprocessed variables into a multiple nonlinear regression model of the following form, and fit them respectively. and : , , in, , , , , , These are the model coefficients obtained through least squares optimization. , For constant terms; Warning output: Output obtained through fitting and ,when More than 0.5 meters or When the preset threshold is exceeded, an early warning is triggered to guide gate control.
[0011] The above method achieves high-precision transient condition early warning. Through a complete technical chain of "centralized processing → logarithmic transformation → nonlinear regression fitting," a systematic early warning method adapted to nonlinear transient processes such as load shedding / load reduction is constructed for the first time, enabling precise monitoring of water level fluctuations. and exceeding the time limit Prediction accuracy ( Compared to linear models ( This has led to a breakthrough improvement. The process is standardized and highly practical for engineering. The four clearly defined steps (data acquisition, preprocessing, fitting, and output) constitute a replicable and implementable standardized operating procedure, which facilitates its promotion and application in different hydropower stations, transforming complex nonlinear analysis into a stable engineering solution.
[0012] Furthermore, the construction and validation of the multivariate nonlinear regression model includes a four-level validation process: First-level historical data fitting: Use historical operating condition data of machine cut-off BE and load shedding 3-5 to train the model and determine the initial coefficients; Second-level validation with new data: The generalization ability of the model is tested using new working condition data, including a water level of 268.5m and a flow rate of 225sfh. Third-level residual analysis: Calculates the difference between the measured values and the fitted values to ensure... The prediction error is less than or equal to 0.1m. The error is less than or equal to 5 minutes; Fourth-level goodness-of-fit determination: based on the model's goodness of fit. As a standard for model qualification.
[0013] The above methods ensured the model's extremely high reliability and generalization ability. A rigorous quality control system was established through a four-level progressive verification process: historical data fitting, new scenario validation, residual analysis, and goodness-of-performance assessment. This system not only guaranteed the model's ability to interpret historical data but also effectively prevented overfitting through new scenario testing, enabling the model to be reliably applied to unknown scenarios. Its generalization ability was significantly better than models that had not undergone systematic validation. Quantitative model evaluation and admission criteria were established, and specific standards were defined. , Prediction error less than or equal to 0.1m A quantization pass threshold of less than or equal to 5 minutes is set. This provides an objective and consistent "hard metric" for model development, acceptance, and iterative optimization, ensuring the minimum performance of the final deployed model.
[0014] Furthermore, the water level change sensitive threshold The specific method for determining the threshold is as follows: analyze the correlation characteristics between the initial water level and the water level variation in historical data, and determine the inflection point of water level change as the sensitive threshold through cluster analysis. .
[0015] The above methods improve the model's adaptability and objectivity, and data-driven cluster analysis is used to determine key parameters. This replaces the traditional method that relies on human experience for setting parameters. This allows the model to automatically identify and adapt to water level response sensitivities specific to a power plant or during its operation, enhancing the method's universality and objectivity. It also amplifies the nonlinear effects of key variables: determined based on real data. Centralized processing ( This allows for more precise alignment of the "inflection point" in water level changes, thereby enabling subsequent secondary terms... It captures nonlinear responses more effectively, improving the sensitivity and accuracy of early warnings.
[0016] Furthermore, in the early warning output, differentiated control strategies are adopted for different operating conditions: Under machine switching conditions, priority is given to flow rate variation. Predictive coefficients guide flow control; Under load shedding conditions, priority is given to basing decisions on the initial water level. and Flow Rate Variable The predictive coefficients guide water level control and flow regulation.
[0017] The above method achieves a precise closed loop from prediction to safety control, directly and automatically linking early warning with specific engineering control actions (gate opening and closing). An early warning is immediately triggered when the predicted value exceeds the clearly defined safety threshold of 0.5 meters, significantly shortening the response time from risk identification to control initiation and improving the automation level of emergency response. It also optimizes equipment operation and maintenance strategies based on precise... It is predicted that "on-demand control" can be implemented to avoid unnecessary or premature operation of the auxiliary gate chamber. It is expected to reduce the frequency of auxiliary gate chamber operation by more than 50%, effectively reducing equipment mechanical wear and energy consumption, and extending equipment service life.
[0018] Furthermore, the initial water level Centralized processing can be replaced by exponential function transformation. ,in This is a coefficient adjusted according to the characteristics of the watershed.
[0019] Furthermore, the aforementioned flow rate variation The natural logarithmic transformation can be replaced by standardization.
[0020] Furthermore, the optimization of model coefficients using the least squares method can be replaced by parameter optimization using the gradient descent method.
[0021] The above methods enhance the flexibility and robustness of the technical solution, clarifying that the initial water level can be replaced by an exponential function, the flow rate variation can be replaced by standardization, and the optimization algorithm can be replaced by gradient descent. These alternatives provide alternative paths that adapt to different data characteristics, computing resources, or watershed characteristics, ensuring the feasibility and robustness of the core invention in different application scenarios.
[0022] This application also includes a water level early warning system for a ship lift in a hydropower station, used to implement a water level early warning method for a ship lift in a hydropower station, including: The data acquisition module is used to collect the initial water level under historical operating conditions, verification operating conditions, and newly added operating conditions. Flow rate fluctuation Variable flow rate , and Actual measured data; The variable preprocessing module is used to process the initial water level. Centralized processing and traffic amplitude adjustment Perform natural logarithmic transformation while preserving flow rate variation. The original value; The model building and processing module is used to build and run the multivariate nonlinear regression model to fit the data. and ; The parameter optimization module is used to optimize the coefficients of the model using the least squares method or the gradient descent method. The model validation module is used to execute the four-level validation process to ensure model accuracy; The early warning output module is used to output the fitting results and trigger an early warning when the conditions are met.
[0023] Furthermore, the model validation module is configured to sequentially perform historical data fitting, new data validation, residual analysis, and goodness-of-fit determination, and to... , Prediction error less than or equal to 0.1m An error of 5 minutes or less is used as the threshold standard for passing the verification.
[0024] This application also includes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a water level early warning method for a ship lift in a hydroelectric power station.
[0025] Compared with existing technologies, the advantages of this invention are: The prediction accuracy is significantly improved: the goodness of fit of the multivariate nonlinear regression model of this invention is significantly enhanced. , The prediction error is less than or equal to 0.1m. The error is less than or equal to 5 minutes, compared to existing technologies ( The prediction accuracy has been greatly improved, and it can still maintain reliable prediction even under extreme conditions (such as high water level shut-off conditions), providing accurate data support for gate regulation.
[0026] Enhanced adaptability to different operating conditions: Through variable preprocessing optimization and differentiated model structure design, the model can be adapted to different machine switching, load shedding and new operating conditions, and its generalization ability is significantly better than the existing linear model with fixed structure.
[0027] Enhanced regulatory guidance: It can accurately predict the continuous window of water level exceeding the limit, and clarify the regulation logic of prioritizing flow control in the machine-switching operation and prioritizing water level control and flow rate adjustment in the load shedding operation, which can reduce the frequency of auxiliary gate operation by more than 50%.
[0028] The project has significant value: it helps improve the navigation efficiency of the ship lift, reduces the number of times the auxiliary lock chamber is opened and closed to extend the equipment life, and at the same time reduces the navigation safety risks caused by water level exceeding the limit, thus laying a solid foundation for the "safe, smooth and efficient" operation of the ship lift.
[0029] High scalability: The model formula structure is universal and can be transferred to similar hydropower station scenarios. Only local coefficient calibration is required for the measured data of a specific watershed, reducing the cost of technology promotion. Attached Figure Description
[0030] Figure 1 This is a flowchart of the water level early warning method in this application. Detailed Implementation
[0031] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0032] The features and performance of the present invention will be further described in detail below with reference to embodiments.
[0033] Please see Figure 1 A water level early warning method for ship lifts in hydropower stations achieves accurate fitting of early warning indicators under ship cut-off and load shedding conditions through three core steps: variable screening and preprocessing, differential model construction, and parameter optimization and verification. Specifically, the method includes the following steps: Data collection and processing: Integrating measured data from historical operating conditions, verification operating conditions, and newly added operating conditions. Data types include initial water level ( ), flow rate variation ( ), Flow rate variation ( ), Actual measurement and actual measurement This forms a standardized dataset, providing basic data support for model building.
[0034] Variable selection and preprocessing: Initial water level ( ): Centralized processing is adopted ( ),in, The threshold for water level changes is determined through cluster analysis of historical data. This process can eliminate the influence of dimensions and highlight the nonlinear response relationship between the initial water level and the early warning indicators. Flow rate variation ( ): Using natural logarithmic transformation ( ), compressing data scale to avoid large values dominating the model, and adapting The characteristic of a large span; Variable flow rate (VRF) ): Retain the original measured value, whose physical meaning is the acceleration of flow rate change, which can directly reflect the degree of drastic flow rate fluctuation.
[0035] Table 1 lists the variable handling methods
[0036] Differentiation Model Construction: A multivariate nonlinear regression model is adopted, with the following basic structure: ,in, To predict the target, , , The coefficients to be determined are given, and a specific function form is designed to address the response differences under shearing and load shedding conditions: Prediction formula: The nonlinear effects of the initial water level are captured by using a quadratic function; Prediction formula: It adapts to the nonlinear extension effect of rising water levels on the duration of exceeding limits.
[0037] Parameter optimization and validation: The least squares (OLS) method was used to optimize the model coefficients, with the objective of minimizing the sum of squared residuals; this was achieved through historical data fitting, validation with new data, residual analysis, and goodness-of-fit assessment. A four-level verification system ensures model accuracy.
[0038] The initial water level function can be replaced by an exponential function. By adjusting the coefficient It adapts to different watershed characteristics; the parameter optimization algorithm can be replaced with gradient descent to improve the computational efficiency of large-scale datasets; the scale compression processing of flow amplitude can be replaced with standardization processing to adapt to some specific watershed data scenarios.
[0039] Differentiated multivariate nonlinear regression models are designed for machine shedding and load shedding conditions, rather than nonlinear models; the variables are preprocessed to adapt their characteristics (centering, logarithmic transformation) instead of using the original data directly; a four-level validation system is established to ensure model accuracy, and the goodness of fit and prediction error thresholds are clearly defined.
[0040] In another specific embodiment, a water level early warning system for a ship lift in a hydropower station includes: Data acquisition module: Used to collect measured data from historical operating conditions, verification operating conditions, and newly added operating conditions, including initial water level ( ), flow rate variation ( ), Flow rate variation ( ), Actual measurement and actual measurement ; Variable preprocessing module: used to filter and adapt the collected data, including centering, natural logarithm transformation, etc. Model building module: Used to build multivariate nonlinear regression models specifically for machine shedding and load shedding conditions; Parameter optimization module: The least squares (OLS) method is used to optimize the model coefficients; Model validation module: Verifies model accuracy through a four-level process: historical data fitting, new data validation, residual analysis, and goodness-of-fit determination; Early warning output module: Output and The prediction results provide guidance for gate control.
[0041] The connection relationships between modules are as follows: the output of the data acquisition module is connected to the input of the variable preprocessing module; the output of the variable preprocessing module is connected to the input of the model building module; the output of the model building module is connected to the input of the parameter optimization module; the output of the parameter optimization module is connected to the input of the model validation module; and the output of the model validation module is connected to the input of the early warning output module, forming a complete data processing and prediction process.
[0042] Implementation and function of each module: Data acquisition module: Acquires data through the existing monitoring system of the hydropower station to ensure data coverage. Flow range The initial water level range and different flow rate scenarios provide comprehensive training and validation data for the model; Variable preprocessing module: Initial water level ( ):calculate ( (As a water level change sensitive threshold), eliminating the influence of dimensions and highlighting the nonlinear response; Flow rate variation ( ):right Take the natural logarithm ( Compressing the data scale helps prevent large values from dominating the model. Variable flow rate (VRF) ): Directly retain the original measured values to reflect the degree of drastic flow fluctuations; Model building module: Based on the fundamental structure of a multivariate nonlinear regression model, construct differentiated prediction formulas: Prediction formula: ; Prediction formula: ; Parameter optimization module: This module aims to minimize the sum of squared residuals. In the algorithm's solution formula , , , , , and the optimal value of the constant term; Model validation module: Historical data fitting: The model was trained using historical operating condition data such as machine cut-off BE and load shedding 3-5, and the initial coefficients were determined; New data validation: The generalization ability of the model was tested using new working condition data such as a water level of 268.5m and a flow rate of 225sfh. Residual analysis: Calculates the difference between the measured values and the fitted values to ensure... The prediction error is less than or equal to 0.1m. The error is less than or equal to 5 minutes; Goodness-of-fit determination: based on As a standard for model qualification; Early warning output module: outputs the prediction results of qualified models ( and The output is sent to the gate control system to guide the gate to open and close as needed.
[0043] The entire workflow and principles: Data Acquisition: Data is collected from three types of operating conditions: historical, verified, and newly added, through the monitoring system. , , Actual measurement Actual measurement Data is used to create standardized datasets. Variable preprocessing: Centralized processing ( ),right Perform natural logarithmic transformation ( ),reserve Original value; Model Construction: Based on the multivariate nonlinear regression structure, a model is established. and The differential prediction formula; Parameter optimization: adopt The algorithm optimizes the formula coefficients and minimizes the sum of squared residuals; Model validation: The model accuracy is ensured by sequentially fitting historical data, validating new data, performing residual analysis, and determining goodness of fit. Early warning output: Output the prediction results of the compliance model to provide quantitative basis for gate control under the conditions of machine shedding and load shedding.
[0044] Functional alternative: Initial water level ( The functional form of ) can be replaced by an exponential function. By adjusting the coefficient It can adapt to the water level response characteristics of different watersheds and can also capture nonlinear relationships. The remaining steps are consistent with the optimal implementation method. Algorithm alternative: Use gradient descent instead of least squares for parameter optimization. This can improve computational efficiency when dealing with large-scale datasets. The optimization objective remains the same: minimizing the sum of squared residuals. The remaining steps are unchanged. Alternative solution for variable handling: flow rate variation ( Standardization can be used to replace natural logarithmic transformation, compress the data scale, and adapt to the specific data distribution scenarios of some watersheds. The processing of other variables and subsequent steps are consistent with the optimal implementation method.
[0045] The essential structure of this application includes: a data acquisition module, a variable preprocessing module (including initial water level centering, flow amplitude compression, and retention of original values for flow rate variation), a differential model construction module, a parameter optimization module, a model verification module (including a four-level verification process), and an early warning output module. The above structure is an indispensable core part for achieving the purpose of this invention. Optional structure: The function form, optimization algorithm, and specific method of traffic amplitude compression can be flexibly replaced according to the actual application scenario without affecting the implementation of core functions.
[0046] The embodiments described above merely illustrate specific implementation methods of this application, and while the descriptions are detailed and specific, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the technical solution of this application, and these modifications and improvements all fall within the scope of protection of this application.
Claims
1. A water level early warning method for a ship lift in a hydropower station, characterized in that, Includes the following steps: Data acquisition: Collect initial water levels under historical operating conditions, verification operating conditions, and newly added operating conditions. Flow rate fluctuation Variable flow rate Measured water level fluctuation within 20 minutes Measured value of the duration of water level exceeding the limit Actual measured data; Variable preprocessing: For the initial water level Centralized processing, i.e., computation ,in The water level change sensitivity threshold was determined through cluster analysis of historical data; For flow rate variation Perform the natural logarithmic transformation, i.e., calculate ; Preserve Variable Flow The original measured value; Early warning index fitting: Input the preprocessed variables into a multiple nonlinear regression model of the following form, and fit them respectively. and : ， ， in, , , , , , These are the model coefficients obtained through least squares optimization. , For constant terms; Warning output: Output obtained through fitting and ,when More than 0.5 meters or When the preset threshold is exceeded, an early warning is triggered to guide gate control.
2. The water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, The construction and validation of the multivariate nonlinear regression model includes a four-level validation process: First-level historical data fitting: Use historical operating condition data of machine cut-off BE and load shedding 3-5 to train the model and determine the initial coefficients; Second-level validation with new data: The generalization ability of the model is tested using new working condition data, including a water level of 268.5m and a flow rate of 225sfh. Third-level residual analysis: Calculates the difference between the measured values and the fitted values to ensure... The prediction error is less than or equal to 0.1m. The error is less than or equal to 5 minutes; Fourth-level goodness-of-fit determination: based on the model's goodness of fit. As a standard for model qualification.
3. The water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, The water level change sensitive threshold The specific method for determining the threshold is as follows: analyze the correlation characteristics between the initial water level and the water level variation in historical data, and determine the inflection point of water level change as the sensitive threshold through cluster analysis. .
4. The water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, In the early warning output, differentiated control strategies are adopted for different operating conditions: Under machine switching conditions, priority is given to flow rate variation. Predictive coefficients guide flow control; Under load shedding conditions, priority is given to basing decisions on the initial water level. and Flow Rate Variable The predictive coefficients guide water level control and flow regulation.
5. A water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, The initial water level Centralized processing can be replaced by exponential function transformation. ,in This is a coefficient adjusted according to the characteristics of the watershed.
6. A water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, The change in flow rate The natural logarithmic transformation can be replaced by standardization.
7. A water level early warning method for a ship lift in a hydropower station according to claim 1, characterized in that, The optimization of model coefficients using the least squares method can be replaced by parameter optimization using the gradient descent method.
8. A water level early warning system for a ship lift in a hydroelectric power station, characterized in that, A method for implementing a water level early warning system for a ship lift in a hydropower station as described in any one of claims 1-7 includes: The data acquisition module is used to collect the initial water level under historical operating conditions, verification operating conditions, and newly added operating conditions. Flow rate fluctuation Variable flow rate , and Actual measured data; The variable preprocessing module is used to process the initial water level. Centralized processing and traffic amplitude adjustment Perform natural logarithmic transformation while preserving flow rate variation. The original value; The model building and processing module is used to build and run the multivariate nonlinear regression model to fit the data. and ; The parameter optimization module is used to optimize the coefficients of the model using the least squares method or the gradient descent method. The model validation module is used to execute the four-level validation process to ensure model accuracy; The early warning output module is used to output the fitting results and trigger an early warning when the conditions are met.
9. A water level early warning system for a ship lift in a hydropower station according to claim 8, characterized in that, The model validation module is configured to sequentially perform historical data fitting, new data validation, residual analysis, and goodness-of-fit determination, and to... , Prediction error less than or equal to 0.1m An error of less than or equal to 5 minutes is used as the threshold standard for passing the verification.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements a water level early warning method for a ship lift in a hydropower station as described in any one of claims 1-7.