A control system and method for an intelligent gate of an ecological discharge flow of a hydropower station

Abstract

This invention discloses a control system and method for an intelligent gate for ecological outflow in a hydropower station, relating to the field of intelligent control technology for hydropower stations. The method includes: real-time acquisition of ecological hydraulic operation data, preprocessing the data, and using an LSTM model to predict the downstream ecological water demand curve; combining the ecological water demand prediction curve with the preprocessed ecological hydraulic operation data, and using a fuzzy PD+I control algorithm for rule-based reasoning to obtain a preliminary set of gate control parameters; comparing the gate's safe operation constraints using numerical constraint analysis, performing boundary mapping and smoothing adjustments to obtain a set of constraint-compliant gate control parameters; and performing rapid simulation verification using a digital twin model based on the constraint-compliant gate control parameter set and the ecological water demand prediction curve to obtain a verified gate control parameter set. This invention achieves efficient unification of ecological protection goals and gate operation safety through the fusion of prediction curves and real-time data using the fuzzy PD+I control algorithm.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for hydropower stations, and in particular to a control system and method for an intelligent gate for ecological outflow in a hydropower station. Background Technology

[0002] Current hydropower station ecological flow control technology has gradually evolved from the traditional static threshold method to intelligent control systems. Modern systems generally employ multi-source hydrological sensor networks to collect parameters such as flow rate and water level in real time, and apply prediction methods such as time series analysis and machine learning to calculate ecological water demand. In terms of control strategies, existing technologies have been upgraded from basic proportional control to intelligent algorithms such as fuzzy control and adaptive control.

[0003] Traditional control methods are difficult to effectively coordinate the relationship between dynamic prediction of ecological water demand and multi-constraint safety control of gates. When the output of the prediction model is used to directly guide the gate regulation, the lack of an adaptive mapping mechanism between the prediction results and control parameters, as well as the lack of real-time verification of constraints such as mechanical limits and hydraulic fluctuations, can easily lead to deviations between control commands and execution effects. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a control method for intelligent gates for ecological outflow in hydropower stations to solve the problem of the lack of an adaptive mapping mechanism between gate control parameters.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a control method for an intelligent gate for ecological outflow in a hydropower station, comprising: real-time acquisition of ecological hydraulic operation data, preprocessing the data, and using an LSTM model to predict the downstream ecological water demand curve; combining the ecological water demand prediction curve with the preprocessed ecological hydraulic operation data, and using a fuzzy PD+I control algorithm to perform rule reasoning to obtain a preliminary control parameter set for the gate; comparing the gate's safe operation constraints using a numerical constraint analysis method, performing boundary mapping and smoothing adjustments to obtain a constraint-compliant gate control parameter set; based on the constraint-compliant gate control parameter set and the ecological water demand prediction curve, performing rapid simulation verification using a digital twin model to obtain a verified gate control parameter set; and using a multi-objective dynamic optimization method to fuse gate control commands and generate gate execution strategies to obtain the optimal intelligent gate control scheme.

[0008] As a preferred embodiment of the control method for the intelligent gate for ecological discharge flow of hydropower stations described in this invention, the ecological hydraulic operation data includes water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data.

[0009] The preprocessing includes filtering and denoising, time synchronization, missing value imputation, and normalization.

[0010] As a preferred embodiment of the control method for the intelligent gate of ecological discharge flow in hydropower stations described in this invention, the specific steps for using an LSTM model to predict the downstream ecological water demand curve are as follows:

[0011] By sliding and stitching the preprocessed historical ecohydraulic operation data into an input sequence in chronological order, a multivariate time series input matrix is ​​generated.

[0012] The optimal time step and prediction length are analyzed by using a dynamic sliding window mechanism to generate the structural configuration parameters of the LSTM model. The Adam optimization algorithm is then used to train the LSTM model, resulting in an LSTM prediction model that fits the historical ecological flow evolution pattern.

[0013] Based on the preprocessed ecological hydraulic operation data, the downstream ecological water demand prediction curve is obtained by sequential inference through the LSTM prediction model.

[0014] As a preferred embodiment of the control method for the intelligent gate for ecological outflow of hydropower stations described in this invention, the method involves combining the ecological water demand prediction curve and preprocessed ecological hydraulic operation data, and obtaining the preliminary control parameter set for the gate through rule reasoning using a fuzzy PD+I control algorithm. The specific steps are as follows:

[0015] Based on the ecological water demand prediction curve and the preprocessed ecological hydraulic operation data, ecological regulation data is obtained by fusing them through time alignment and feature normalization fusion methods, and fuzzification processing is performed using fuzzy membership rules to generate fuzzy ecological regulation data.

[0016] The gate control rules are matched and reasoned through a fuzzy rule base to obtain fuzzy gate control quantities. The PD+I defuzzification method is then used for quantification calculation to generate a preliminary gate control parameter set including gate opening set value, gate opening speed set value, and gate pre-pressure differential set value.

[0017] As a preferred embodiment of the control method for the intelligent gate for ecological outflow of hydropower stations described in this invention, the steps of comparing the gate's safe operation constraints using numerical constraint analysis, performing boundary mapping and smoothing adjustments to obtain a set of constraint-compliant gate control parameters are as follows.

[0018] Based on the preliminary set of control parameters, the numerical constraint analysis method is used to compare each parameter with the gate's safe operation constraints to obtain the difference between the control parameters and the safety boundary.

[0019] By using the boundary mapping method, the initial control parameter set of the gate that exceeds the limit is compressed and stretched to the range allowed by the gate's safe operation constraints, thus generating the initial boundary adjustment control parameter set;

[0020] By using a smoothing adjustment algorithm to adjust the rate of change of the initial boundary control parameter set, continuous optimization is performed to eliminate abrupt changes, thus obtaining the control parameter set for the constrained compliant gate.

[0021] As a preferred embodiment of the control method for the intelligent gate for ecological outflow of hydropower stations described in this invention, the method involves: based on the constrained and compliant gate control parameter set and the ecological water demand prediction curve, rapid simulation verification is performed using a digital twin model to obtain a verified gate control parameter set. The specific steps are as follows.

[0022] Based on the historical constraint compliance gate control parameter set and the historical ecological water demand prediction curve, the digital twin model is trained by supervised learning method to fit parameters.

[0023] Based on the set of constrained and compliant gate control parameters and the ecological water demand prediction curve, the driving data is obtained by combining the parameters through feature alignment and time series fusion. The trained digital twin model is then used for rapid dynamic simulation to simulate the flow change process after the execution of the constrained and compliant gate control parameters, and the simulated flow response results are obtained.

[0024] By comparing the deviation analysis method with the ecological water demand prediction curve, the flow satisfaction and flow-pressure difference stability index are calculated to obtain the performance evaluation results. A predefined threshold judgment strategy is adopted to analyze the positive and negative tolerance of the predicted demand and the stability of the pressure difference fluctuation, and to obtain the verification gate control parameter set.

[0025] As a preferred embodiment of the control method for the intelligent gate for ecological discharge flow in hydropower stations described in this invention, the optimal intelligent gate control scheme is obtained by fusing gate control commands and generating gate execution strategies through a multi-objective dynamic optimization method. The specific steps are as follows:

[0026] Based on the verification of the gate control parameter set and the ecological water demand prediction curve, the objective function is optimized by a multi-objective weight allocation method to obtain the multi-objective optimization parameter set;

[0027] A set of gate control instructions with optimal solution candidates is obtained by searching the solution space and selecting the Pareto front through a dynamic multi-objective optimization algorithm.

[0028] By using the strategy fusion method, under the condition of satisfying the multi-objective optimization parameter set, the optimal solution candidates are screened and normalized and synthesized to obtain a unified scheduling execution instruction set;

[0029] The optimal intelligent gate control scheme is generated by converting the data to the standard Modbus RTU format through format mapping and communication packaging methods.

[0030] Secondly, this invention provides a control system for an intelligent gate for ecological outflow in a hydropower station, comprising a prediction module, a decision module, a constraint module, a simulation module, and an optimization module. The prediction module collects ecological hydraulic operation data in real time and performs preprocessing; it uses an LSTM model to predict the downstream ecological water demand curve. The decision module combines the ecological water demand prediction curve with the preprocessed ecological hydraulic operation data and uses a fuzzy PD+I control algorithm to perform rule reasoning to obtain a preliminary set of gate control parameters. The constraint module compares the gate's safe operation constraints using numerical constraint analysis, performs boundary mapping and smoothing adjustments, and obtains a set of constraint-compliant gate control parameters. The simulation module performs rapid simulation verification based on the constraint-compliant gate control parameter set and the ecological water demand prediction curve using a digital twin model to obtain a verified gate control parameter set. The optimization module uses a multi-objective dynamic optimization method to fuse gate control commands and generate gate execution strategies to obtain the optimal intelligent gate control scheme.

[0031] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the control method for the intelligent gate for ecological outflow of hydropower station as described in the first aspect of the present invention.

[0032] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the control method for the intelligent gate for ecological outflow of hydropower station as described in the first aspect of the present invention.

[0033] The beneficial effects of this invention are as follows: By using a multivariate time series input and a dynamic sliding window mechanism through an LSTM model, the nonlinear evolution law of ecological hydraulic data is captured, generating a high-precision water demand prediction curve, which provides a reliable target benchmark for the control algorithm; the fuzzy PD+I control algorithm, based on the fusion of the prediction curve and real-time data, enables the gate control command to not only accurately track changes in ecological water demand and suppress flow fluctuations, but also adaptively adjust the control intensity to avoid mechanical overshoot, providing a highly stable and low-disturbance control basis for subsequent constraint verification and multi-objective optimization, and achieving an efficient unity between ecological protection goals and gate operation safety. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating the control method for intelligent gates used for ecological outflow at hydropower stations.

[0036] Figure 2 A schematic diagram of the control system for an intelligent gate for ecological outflow at a hydropower station.

[0037] Figure 3 A flowchart for outputting the ecological water demand prediction curve.

[0038] Figure 4 This is a flowchart for outputting and verifying the gate control parameter set. Detailed Implementation

[0039] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0040] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0041] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0042] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a control method for an intelligent gate for ecological outflow of a hydropower station, comprising the following steps:

[0043] S1. Real-time collection of ecological hydraulic operation data and preprocessing.

[0044] S1.1 Eco-hydraulic operation data includes water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data.

[0045] Specifically, high-precision ultrasonic flow meters are installed at the upstream and downstream sections of the gate to acquire water flow data; pressure sensor arrays are pre-embedded before and after the gate to acquire water level data; differential pressure transmitters are installed on both sides of the gate to acquire upstream and downstream differential pressure data; multi-parameter water quality monitoring buoys are deployed in the downstream area of ​​the spillway, with built-in temperature probes to acquire water temperature data and dissolved oxygen sensors to acquire dissolved oxygen data; water flow data is generated by calculation using ultrasonic echo signals, water level data is obtained by conversion of water pressure detected by pressure sensors, and upstream and downstream differential pressure data is directly measured by differential pressure transmitters; water temperature data is measured by contact measurement using temperature probes, and dissolved oxygen data is acquired through electrode reactions of electrochemical sensors.

[0046] S1.2 Preprocessing includes filtering and denoising, time synchronization, missing value imputation, and normalization.

[0047] Specifically, in the filtering and denoising stage, a Butterworth low-pass filter is applied to the water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data. The cutoff frequency parameter is set to, for example, 0.5Hz. The filtered output value is calculated point by point in time sequence through a sliding window. In the time synchronization stage, the internal clock of each sensor is calibrated using a GPS clock source. Based on the NTP protocol, the timestamps of the water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data are aligned to the same time axis, with a deviation correction threshold of, for example, 1 millisecond. In the missing value filling stage, the missing time points in the water level data are located, and the two adjacent valid data before and after the missing point are extracted. The intermediate value is linearly allocated according to the time interval and filled. In the normalization processing stage, the historical extreme values ​​of the water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data are extracted. For example, the maximum value of the water level data over a historical time span of 30 days is 50 meters and the minimum value is 30 meters. The real-time water level value is mapped to the [0,1] interval according to the formula "(current value-30) / (50-30)".

[0048] S2. Use the LSTM model to predict the downstream ecological water demand curve.

[0049] S2.1 By sliding and stitching the pre-processed historical ecological hydraulic operation data into an input sequence in chronological order, a multivariate time series input matrix is ​​generated.

[0050] Specifically, the process involves extracting water flow, water level, upstream and downstream pressure difference, water temperature, and dissolved oxygen data from historically preprocessed ecological hydraulic operation data; arranging the data points in ascending order by timestamp, with each time point containing five variable values; candidate values ​​for the sliding window length parameter, such as 600, 1800, and 3600 time steps, corresponding to example durations of 10 minutes, 30 minutes, and 60 minutes respectively (one data point per second); sequentially extracting continuous time window data blocks starting from the beginning of the time series, with a window length of, for example, 3600 time steps; each data block containing the values ​​of water flow, water level, upstream and downstream pressure difference, water temperature, and dissolved oxygen within the example 3600 time steps; fixing the sliding step size at 1 time step, sliding the window backward by 1 second when generating a new data block, retaining, for example, the data from the previous 3599 time steps and appending the latest time step data; repeating this process until the entire historical time range is covered; and stacking the new data blocks in chronological order into a three-dimensional array structure with dimensions [total number of samples, time window length, ...]. [Number of variables], in the example, the total number of samples is the length of the original ecological hydraulic operation data minus, for example, 3599; the sliding window length parameter is selected through cross-validation, the candidate values ​​are input into the LSTM model for training and the prediction error of the validation set is calculated, and the parameter corresponding to the minimum error is selected, for example, 3600 time steps to make the validation error less than 0.05; finally, the multivariate time series input matrix is ​​output.

[0051] S2.2 Analyze the optimal time step and prediction length through the dynamic sliding window mechanism, generate the structural configuration parameters of the LSTM model, and train the LSTM model using the Adam optimization algorithm to obtain an LSTM prediction model that fits the historical ecological flow evolution pattern.

[0052] Specifically, the candidate set of time steps includes, for example, 3600, 7200, and 14400 time steps, and the candidate set of prediction length includes, for example, 600 and 1200 time steps; the multivariate time series input matrix is ​​divided into training and validation sets in chronological order, with the training set accounting for, for example, 70% and the validation set accounting for, for example, 30%; for each combination of candidate value of time step and candidate value of prediction length, the number of nodes in the input layer of the LSTM model is configured to be the number of variables in the multivariate time series input matrix, and the number of nodes in the output layer is the candidate value of prediction length;

[0053] Set the LSTM model to have 2 hidden layers, with 64 neurons per layer (e.g.); initialize the LSTM model weight matrix and bias vector; use the Adam optimization algorithm, setting the initial learning rate to, for example, 0.001 and the batch size to, for example, 32; input the training set into the LSTM model for forward propagation to calculate the predicted output, and calculate the difference between the predicted output and the actual ecological flow value using the mean squared error loss function;

[0054] Perform backpropagation to update the LSTM model parameters; calculate the mean squared error of the prediction on the validation set after each training round, and record the parameter combination corresponding to the lowest error value; terminate training when the validation error has not decreased for 10 consecutive rounds using an early stopping mechanism; select the parameter combination with the smallest validation error as the optimal time step and prediction length, for example, a time step of 7200 time steps and a prediction length of 600 time steps; retrain the LSTM model using the complete training data and save the final weight parameters; obtain the LSTM prediction model that fits the historical ecological flow evolution pattern.

[0055] It should be noted that the expression for calculating the mean squared error of the predictions on the validation set is:

[0056] ;

[0057] in, It is the mean squared error of the predictions on the validation set. It is the total number of samples in the validation set. It is the index variable of the validation set samples. It is the verification set number The true value of each sample It is the verification set number The predicted value for each sample.

[0058] S2.3. Based on the preprocessed ecological hydraulic operation data, the downstream ecological water demand prediction curve is obtained by sequential inference through the LSTM prediction model.

[0059] Specifically, the process involves acquiring the latest, exemplary 7200-time-step preprocessed water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data, forming an input matrix with dimensions 1×7200×5; loading the input matrix into a trained LSTM prediction model, which has two hidden layers and an exemplary 64 neurons per layer; performing forward computation of the LSTM prediction model, processing the 7200 time-step data step by step; generating an exemplary 600-time-step ecological water demand prediction sequence from the output layer; and arranging the 600 predicted values ​​in chronological order, with the horizontal axis representing the exemplary future 600 seconds and the vertical axis representing the ecological water demand value.

[0060] The horizontal axis time range is from the current moment to an exemplary 600 seconds in the future. When a new exemplary time step monitoring data is added, the earliest time step in the input matrix is ​​removed, and the latest time step is appended to the end, keeping the input matrix dimension at, for example, 1×7200×5. The updated input matrix is ​​re-input into the LSTM prediction model to perform forward computation. Real-time data updates and predictions are performed cyclically, outputting exemplary 600 time step ecological water demand prediction values ​​each time. The continuously output prediction values ​​are connected in chronological order to form a complete time series curve. The curve time resolution is consistent with the data acquisition frequency, with an exemplary one data point per second. Through rolling updates and prediction outputs, the downstream ecological water demand prediction curve is obtained.

[0061] S3. Combine the ecological water demand prediction curve with the pre-processed ecological hydraulic operation data, and use the fuzzy PD+I control algorithm to perform rule reasoning to obtain the preliminary control parameter set of the gate.

[0062] S3.1 Based on the ecological water demand prediction curve and the preprocessed ecological hydraulic operation data, ecological regulation data is obtained by fusing them through time alignment and feature normalization fusion methods, and fuzzification processing is performed using fuzzy membership rules to generate fuzzy ecological regulation data.

[0063] Specifically, the predicted value sequence of the ecological water demand prediction curve is extracted and compared with the pre-processed water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data; the start and end timestamps of the ecological water demand prediction curve are completely consistent with the start and end timestamps of the pre-processed water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data; when the timestamps are completely consistent, the timestamp range of the ecological water demand prediction curve completely overlaps with the timestamp range of the pre-processed ecological hydraulic operation data; Non-overlapping time periods are truncated to retain common time intervals; the predicted ecological water demand is aligned point-by-point with water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data according to timestamps; the maximum and minimum values ​​of historical water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data are calculated; extreme value normalization is performed on the real-time data of water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data, and the normalization formula is the difference between the current value and the historical minimum value divided by the difference between the historical maximum value and the minimum value;

[0064] Normalized water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data are merged into a multidimensional data sequence in chronological order. Fuzzy set partitioning is performed for each parameter. The water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data are divided into five fuzzy sets: "minimal," "small," "medium," "large," and "maximum," based on percentile points such as 20%, 40%, 60%, and 80%. Each fuzzy set corresponds to a triangular or trapezoidal membership function, with the vertex position of the membership function set according to the historical distribution of water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data. The membership degree of the normalized parameter value at each time point within its respective fuzzy set is calculated.

[0065] The fuzzy rule base is formed by manually summarizing hydropower station operation experience, such as "if the water level is low and the ecological water demand prediction curve is high, then the gate opening should be increased"—a set of condition-conclusion logical relationships. Based on the fuzzy rule base, condition combinations are matched point-by-point. The rule form in the fuzzy rule base is "if parameter A is X and parameter B is Y, then output Z". The fuzzy inference engine is executed to perform a weighted average of the results of all activated rules. The membership distribution of fuzzy regulation quantities at each time point is generated. The fuzzy regulation quantity sequence with timestamps is output as fuzzy data for ecological regulation.

[0066] S3.2. The gate control rule conditions are matched and reasoned through the fuzzy rule base to obtain the fuzzy gate control quantity. The PD+I defuzzification method is used to perform quantification calculation to generate a preliminary gate control parameter set including the gate opening degree setting value, the gate opening speed setting value, and the gate pre-pressure difference setting value.

[0067] It should be noted that the expression for generating the preliminary control parameter set through quantization calculation using the PD+I defuzzification method is as follows:

[0068] ;

[0069] in, This is the gate opening setting value. This is the gate opening speed setting value. This is the gate pre-pressure differential setting value. It is the error between the predicted downstream ecological water demand and the current measured flow rate. yes The proportionality coefficient (exemplary values ​​range from 0.5 to 5.0). yes The scaling factor (example values ​​range from 0.1 to 2.0). yes The proportionality coefficient (exemplary values ​​range from 0.05 to 1.0). yes The differential coefficients (example values ​​range from 0.01 to 1.0). yes The differential coefficients (exemplary values ​​range from 0.001 to 0.5) yes The differential coefficients (exemplary values ​​range from 0.0001 to 0.1) yes The integral coefficient (exemplary values ​​range from 0.001 to 0.5). yes The integral coefficient (example values ​​range from 0.0001 to 0.2). yes The integral coefficient (exemplary values ​​range from 0.00001 to 0.1). It is the change in the flow deviation of downstream ecological water demand. It is the time interval during the downstream ecological water demand sampling process. The flow deviation of downstream ecological water demand By sampling time interval The accumulated points;

[0070] Specifically, the following steps are taken: extract the membership degree distribution of fuzzy control quantities at each time point in the fuzzy data of ecological regulation; match the fuzzy rule base; analyze the degree of matching between the preconditions of each rule in the fuzzy rule base and the current membership degree, and take the minimum value of the membership degree of each condition as the rule activation intensity; perform a weighted average on the rule conclusions of the activated fuzzy rule base as the activation intensity weight, and take the clear value corresponding to the rule conclusion as the control quantity; perform the above fuzzy inference process on the gate opening setpoint, gate opening speed setpoint, and gate pre-pressure difference setpoint respectively to obtain the fuzzy gate control quantity;

[0071] For the gate opening degree, gate opening speed, and gate pre-pressure differential control parameters, set the proportional coefficient range, derivative coefficient range, and integral coefficient range respectively. For example, the proportional coefficient for the gate opening degree setpoint is 0.5 to 5.0, the proportional coefficient for the gate opening speed setpoint is 0.1 to 2.0, and the proportional coefficient for the gate pre-pressure differential setpoint is 0.05 to 1.0. Calculate the proportional term based on the real-time error, where the error is the instantaneous difference between the predicted water demand and the measured flow rate. Calculate the derivative term based on the error change rate, where the change rate is the rate between adjacent time intervals. The ratio of the point error difference to the sampling interval; the integral term is calculated based on the historical cumulative error value, which is the sum of the products of all historical error values ​​and the sampling interval; the proportional term, derivative term, and integral term are multiplied by their corresponding coefficients and then superimposed to obtain the gate opening setting value, the gate opening speed setting value, and the gate pre-pressure differential setting value; the coefficient values ​​are selected within a preset range, for example, the proportional coefficient of the gate opening setting value is 2.0, the derivative coefficient is 0.1, and the integral coefficient is 0.05; the output contains a set of preliminary gate control parameters containing the three setting values.

[0072] S4. By comparing the gate's safe operation constraints using numerical constraint analysis, boundary mapping and smoothing adjustments are performed to obtain a set of control parameters for the constrained gate.

[0073] S4.1 Based on the preliminary control parameter set, the control parameters and safety boundary are compared item by item using numerical constraint analysis and gate safety operation constraints to obtain the difference between the control parameters and the safety boundary.

[0074] Specifically, the safety operation constraints of the gates are extracted from the hydropower station equipment technical manual. For example, the maximum allowable value for the gate opening setting is 80%, the maximum allowable value for the gate opening speed setting is 10% / s, and the maximum allowable value for the gate pre-pressure differential setting is 5m. The gate opening setting value in the initial control parameter set is compared with the maximum allowable value in the constraints. For example, the difference between the gate opening setting value of 85% and the maximum allowable value of 80% is +5%. The gate opening speed setting value is compared with the maximum allowable value. For example, the difference between the gate opening speed setting value of 12% / s and the maximum allowable value of 10% / s is +2% / s. The gate pre-pressure differential setting value is compared with the maximum allowable value. For example, the difference between the gate pre-pressure differential setting value of 6m and the maximum allowable value of 5m is +1m.

[0075] The sign rule for the difference values ​​is that a positive value indicates that the gate exceeds the safety boundary, and a negative value indicates that the gate is below the safety boundary. Record the difference in gate opening setting value, gate opening speed setting value, and gate pre-pressure difference setting value. The gate's safe operation constraints are derived from the "Gate Mechanical Structure Limit Parameter Table" in Chapter 3, Section 5 of the equipment technical manual. The difference between the control parameters and the safety boundary is obtained.

[0076] S4.2. By using the boundary mapping method, the initial control parameter set of the gate that exceeds the limit is compressed and stretched to the range allowed by the gate's safe operation constraints, and an initial boundary adjustment control parameter set is generated.

[0077] Specifically, the maximum permissible value for the gate opening setting is 80% of the example value, the maximum permissible value for the gate opening speed setting is 10% / s of the example value, and the maximum permissible value for the gate pre-pressure differential setting is 5m of the example value; read the difference data between the control parameters and the safety boundary, with the difference sign rule being that a positive value indicates that the parameter exceeds the safety boundary, and a negative value indicates that it does not exceed the safety boundary; when the difference in the gate opening setting value is positive, the gate opening setting value is directly set to the example value of 80%;

[0078] When the difference in the gate opening speed setpoint is positive, the gate opening speed setpoint is linearly compressed according to the excess ratio. The compression range is the current value minus an exemplary 10% / s and then multiplied by an exemplary compression coefficient of 0.8. When the difference in the gate pre-pressure differential setpoint is positive, the gate pre-pressure differential setpoint is linearly stretched to an exemplary 5m. If the difference is negative, the parameter values ​​in the original gate preliminary control parameter set are retained. During the adjustment process, the gate opening degree setpoint is not allowed to exceed 80% of the exemplary value, the gate opening speed setpoint after compression is not allowed to exceed 10% / s of the exemplary value, and the gate pre-pressure differential setpoint after stretching is not allowed to exceed 5m of the exemplary value. The adjusted gate opening degree setpoint, gate opening speed setpoint, and gate pre-pressure differential setpoint are recombined according to the original data structure to generate a preliminary boundary adjustment control parameter set.

[0079] S4.3. The change rate of the initial boundary adjustment control parameter set is adjusted by a smoothing adjustment algorithm to continuously optimize and eliminate abrupt changes, thereby obtaining the control parameter set of the constrained and compliant gate.

[0080] Specifically, the process involves extracting time series data of gate opening setpoints, gate opening speed setpoints, and gate pre-pressure differential setpoints from the initial boundary adjustment control parameter set; analyzing the rate of change for each adjacent time point in chronological order, where the rate of change is the difference between the current value and the previous time point divided by an exemplary 1-second time interval; setting the maximum permissible rate of change as follows: the rate of change for gate opening setpoints not exceeding an exemplary 8% / s, the rate of change for gate opening speed setpoints not exceeding an exemplary 5% / s², and the rate of change for gate pre-pressure differential setpoints not exceeding an exemplary 1.5 m / s; and linearly attenuating parameter values ​​exceeding the maximum permissible rate of change, with the adjustment increment being the excess portion multiplied by an exemplary attenuation coefficient of 0.6.

[0081] A sliding window mean filter is applied, with the window length set to an exemplary 10 time points. The parameter values ​​within the window are replaced by the arithmetic mean. Exponential smoothing analysis is performed, with the smoothing coefficient set to an exemplary 0.2. The current smoothed value is equal to the current value multiplied by 0.2 plus the smoothed value of the previous time point multiplied by 0.8. The adjusted parameter values ​​are then subjected to a secondary verification against the maximum allowable value of the gate opening setting (exemplarily 80%), the maximum allowable value of the gate opening speed setting (exemplarily 10% / s), and the maximum allowable value of the gate pre-pressure differential setting (exemplarily 5m) in the gate safety operation constraints. If the verification fails, the attenuation and filtering operations are repeated, and the final parameter values ​​are aligned and combined according to the timestamp to obtain the set of control parameters for the constrained and compliant gate.

[0082] S5. Based on the constrained and compliant gate control parameter set and the ecological water demand prediction curve, a rapid simulation verification is performed using a digital twin model to obtain the gate control parameter set that has passed verification.

[0083] S5.1 Based on the historical constraint compliance gate control parameter set and historical ecological water demand prediction curve, the digital twin model is trained by supervised learning method to fit parameters.

[0084] Specifically, the historical constraint-compliant gate control parameter set includes timestamp-aligned gate opening setting, gate opening speed setting, and gate pre-pressure differential setting; historical ecological water demand prediction curve data within the same time range are extracted, with timestamps aligned with the parameter set; the historical constraint-compliant gate control parameter set is used as input features, and the historical ecological water demand prediction curve data is used as output labels, arranged in chronological order as training samples; the training samples are divided into a training set and a validation set in an exemplary 7:3 ratio.

[0085] Configure the digital twin model structure as an exemplary three-layer fully connected neural network, set the number of input layer nodes to an exemplary 3 corresponding to the gate opening degree setting value, gate opening speed setting value, and gate pre-pressure difference setting value, set the number of hidden layer nodes to an exemplary 64, and set the number of output layer nodes to an exemplary 1 corresponding to the ecological water demand prediction value; initialize the digital twin model weight parameters as an exemplary random normal distribution;

[0086] The Adam optimization algorithm is used with a learning rate of 0.001 (exemplary) and a batch size of 32 (exemplary). The training set is input into the digital twin model to perform forward propagation, and the predicted ecological water demand value is output. The mean square error between the predicted value and the actual value is calculated. The weight parameters are updated by backpropagation. After each batch is trained, the validation set is input into the digital twin model to obtain the validation error. If the validation error does not decrease for 10 consecutive (exemplary) rounds, training is stopped. The weight parameters at the lowest validation error are saved. Training is repeated until the maximum number of iterations (exemplary) of 1000 rounds is reached. The trained digital twin model has the ability to predict ecological water demand based on the gate opening degree setting value, the gate opening speed setting value, and the gate pre-pressure difference setting value, thus obtaining the trained digital twin model.

[0087] S5.2 Based on the set of control parameters for the constrained and compliant gate and the ecological water demand prediction curve, the driving data is obtained by combining the parameters through feature alignment and time series fusion. The trained digital twin model is then used for rapid dynamic simulation to simulate the flow change process after the execution of the control parameters for the constrained and compliant gate, and the simulated flow response results are obtained.

[0088] Specifically, the control parameter set for the constrained and compliant gates is extracted, including gate opening setting values, gate opening speed setting values, and gate pre-pressure differential setting values ​​aligned with timestamps; the pre-processed ecological water demand prediction curve data is extracted, with timestamps aligned with the control parameter set for the constrained and compliant gates; the control parameter set for the constrained and compliant gates and the ecological water demand prediction curve data are matched point by point according to timestamps, and missing time points are filled using linear interpolation. For example, if the gate opening setting value at second t is missing, the average value of the gate opening setting values ​​at second t-1 and second t+1 is used to fill it in;

[0089] The gate opening setting, gate opening speed setting, gate pre-pressure differential setting, and ecological water demand prediction are concatenated in columns to form multi-dimensional driving data. The dimension of the multi-dimensional driving data is time step × 4. The multi-dimensional driving data is input into the trained digital twin model. The structure of the digital twin model is a three-layer fully connected neural network. The number of nodes in the input layer is 4, corresponding to the gate opening setting, gate opening speed setting, gate pre-pressure differential setting, and ecological water demand prediction. The number of nodes in the output layer is 1, corresponding to the flow prediction.

[0090] The digital twin model is propagated forward step by step in chronological order. For example, the driving data is input into the model at second t and the flow prediction value at second t is output. The flow prediction values ​​are arranged in chronological order to form a time series. The time resolution of the time series is consistent with the driving data acquisition frequency, for example, one flow prediction value per second. The flow prediction values ​​are recorded as the simulation flow response results. The simulation flow response results are stored in the simulation result database of the hydropower station's central control server to obtain the simulation flow response results.

[0091] S5.3 By comparing the deviation analysis method with the ecological water demand prediction curve, the flow satisfaction and flow-pressure difference stability index are calculated to obtain the performance evaluation results. A predefined threshold judgment strategy is adopted to analyze the positive and negative tolerance of the predicted demand and the stability of the pressure difference fluctuation, and to obtain the verification gate control parameter set.

[0092] Specifically, by analyzing the acceptance clauses defined in Section 8 of Chapter 2, "Flow Control Acceptance Standards," of the "Operation Procedures for Ecological Regulation of Hydropower Stations," we extract the flow satisfaction thresholds, such as the error range of ±5% for the predicted ecological water demand, the flow pressure difference fluctuation thresholds, such as the flow change between adjacent time points ≤1%, and the compliance rate thresholds, such as ≥90% of time points meeting the requirements, to form a predefined threshold judgment strategy.

[0093] By comparing the simulated flow response results with the ecological water demand prediction curve using the deviation analysis method, the time series of the simulated flow response results and the time series of the ecological water demand prediction curve were extracted. The timestamps of the two series were verified to be completely aligned, and missing time points were filled with the value of the previous time point. The absolute error value between the simulated flow response results and the ecological water demand prediction curve was obtained point by point. The number of time points where the absolute error was within the flow satisfaction threshold range defined in Chapter 2, Section 8 of the "Operating Procedures for Ecological Regulation of Hydropower Stations" was counted, and a flow satisfaction index was generated.

[0094] The system counts the number of time points where the absolute value of the difference between simulated flow response results at adjacent time points exceeds the flow pressure difference volatility threshold, generating a flow pressure difference stability index. Based on a predefined threshold judgment strategy, it verifies whether the flow pressure difference stability index meets or exceeds the compliance rate threshold, whether the flow pressure difference stability index is lower than the maximum volatility threshold, and whether all errors are within the tolerance range. When the flow pressure difference stability index meets or exceeds the compliance rate threshold, the flow pressure difference stability index volatility is lower than the maximum volatility threshold, and all error values ​​are within the tolerance range, all three conditions are considered met. After all three conditions are met, the constraint compliance gate control parameter set is marked as the verification passed gate control parameter set. The verification passed gate control parameter set is then output.

[0095] S6. The optimal intelligent gate control scheme is obtained by fusing gate control commands and generating gate execution strategies through a multi-objective dynamic optimization method.

[0096] S6.1 Based on the verification of the gate control parameter set and the ecological water demand prediction curve, the objective function is optimized by using a multi-objective weight allocation method to obtain the multi-objective optimization parameter set.

[0097] Specifically, based on the gate opening setpoint, gate opening speed setpoint, and gate pre-pressure differential setpoint verified in the gate control parameter set, a parameter sequence aligned with the timestamp of the ecological water demand prediction curve is extracted; the ecological water demand matching degree weight is set to an exemplary 0.6, the gate operation frequency weight is set to an exemplary 0.3, and the energy consumption control weight is set to an exemplary 0.1; the absolute value of the error between the simulated flow response result and the ecological water demand prediction curve at each time point is quantified, and the error value is mapped to the matching degree scoring interval; the number of times the gate opening setpoint changes by more than 2% between adjacent time points is counted and mapped to the operation frequency scoring interval;

[0098] The percentage of time points where the gate opening speed setpoint exceeds the exemplary 5% / s is mapped to an energy consumption scoring interval. The results from the three scoring intervals are linearly weighted to generate a comprehensive score. The gradient descent method is used to iteratively adjust the gate opening degree setpoint, gate opening speed setpoint, and gate pre-pressure difference setpoint. After each adjustment, it is verified whether the parameters meet the maximum opening degree, maximum opening speed, and maximum pre-pressure difference constraints for safe gate operation. If a parameter exceeds the limit, it reverts to the previous valid value. Optimization terminates when the comprehensive score improvement is less than the exemplary 0.1% in 100 consecutive exemplary iterations. The set of parameters that meets the constraints and has the highest comprehensive score is output. This generates a multi-objective optimization parameter set.

[0099] S6.2. A set of gate control instructions with optimal solution candidates is obtained by performing solution space search and Pareto front selection through dynamic multi-objective optimization algorithm;

[0100] Specifically, the Latin hypercube sampling method is used to generate 100 exemplary initial parameter combinations. The parameter range is limited to 0% to 80% of the gate opening setting, 0% / s to 10% / s of the gate opening speed setting, and 0m to 5m of the gate pre-pressure differential setting. The ecological water demand matching score is generated by inversely mapping the absolute error average between the simulated flow response results and the ecological water demand prediction curve. The exemplary error average is 0.5m³ / s, and the score is 0.8. The gate operation frequency score is generated by the reciprocal of the standard deviation of the opening change amplitude at adjacent time points. The exemplary standard deviation is 1.2, and the score is 0.6. The energy consumption control score is generated by the reciprocal of the sum of the squares of the gate opening speed setting. The exemplary sum of the squares is 50, and the score is 0.7.

[0101] After each exemplary 50 iterations, the learning rate decreases exponentially from the exemplary initial value of 0.8 to the exemplary 0.2, and the mutation probability increases linearly from the exemplary 0.1 to the exemplary 0.3; Solution space search: A simulated binary crossover operator is applied to perform crossover operations on the parameter combinations, with the crossover distribution exponent set to the exemplary 20; A multinomial mutation operator is used to perturb the parameters, with the mutation distribution exponent set to the exemplary 15; After each iteration, the parameter combinations are sorted by non-dominated order, and the first exemplary 20% of the non-dominated solutions are retained;

[0102] The congestion distance calculation method is used to screen uniformly distributed solutions, and the congestion threshold is set to an exemplary 0.5. Optimization is terminated when the change in the hypervolume index of the Pareto front solution set is less than 1% in 10 consecutive exemplary iterations. The parameter combination of the Pareto front solution set is input into the digital twin model simulation to verify whether the gate opening setting value exceeds 80% of the exemplary value, whether the gate opening speed setting value exceeds 10% / s of the exemplary value, and whether the gate pre-pressure difference setting value exceeds 5m of the exemplary value. Solutions that exceed the limits are eliminated, and the remaining solution set is marked as the gate control instruction set of the optimal solution candidate.

[0103] S6.3. Using the strategy fusion method, under the condition of satisfying the multi-objective optimization parameter set, the optimal solution candidates are screened and normalized to obtain a unified scheduling execution instruction set.

[0104] Specifically, solutions that satisfy the multi-objective optimization parameter set are extracted from the set of gate control instructions for optimal solution candidates; the ecological water demand matching degree weight is 0.5 (exemplary), the gate operation frequency weight is 0.3 (exemplary), and the energy consumption control weight is 0.2 (exemplary), and the comprehensive scores of the candidate solutions are weighted and ranked; the top 10% of the candidate solutions in terms of comprehensive score are retained, and the gate opening degree setting value does not exceed 80% (exemplary), the gate opening speed setting value does not exceed 10% / s (exemplary), and the gate pre-pressure difference setting value does not exceed 5m (exemplary).

[0105] Candidate solutions are segmented according to an exemplary 60-second time window. Within the window, the gate opening setting, gate opening speed setting, and gate pre-pressure difference setting are respectively subjected to time-weighted averaging based on the scoring ratio. Parameter verification: If the synthesized value exceeds 80% of the exemplary gate opening setting, 10% / s of the exemplary gate opening speed setting, or 5m of the exemplary gate pre-pressure difference setting, it is truncated according to the constraint value. The parameters that pass the verification are arranged into a continuous sequence according to an exemplary 1-second time resolution. A unified scheduling execution instruction set is obtained.

[0106] S6.4. Convert to standard Modbus RTU format through format mapping and communication packaging methods to generate the optimal intelligent gate control scheme.

[0107] Specifically, this involves extracting the gate opening degree setting, gate opening speed setting, and gate pre-pressure differential setting parameter sequences from the unified scheduling execution instruction set; converting floating-point parameter values ​​to 4-byte hexadecimal format according to the IEEE 754 standard, for example, converting the gate opening degree setting of 80.0% to hexadecimal 43A00000; generating a communication frame structure according to the Modbus RTU protocol specification, including the device address code (example 0x01), function code (example 0x10, write multiple registers), start address (example 0x0000), number of registers (example 0x0003), number of bytes (example 0x06), data field (hexadecimal values ​​of the three parameters), and CRC16 checksum; using the CRC-16 algorithm to analyze the checksum of bytes from the device address code to the data field, with an initial value (example 0xFFFF) and a polynomial (example 0x8005); verifying that the receiver's returned response frame has the function code (example 0x10) and that the data field is consistent with the sent frame; and converting the complete Modbus... RTU data frames are arranged by timestamp, with an exemplary time resolution of 1 second, to generate the final optimal smart gate control scheme.

[0108] This embodiment also provides a control system for an intelligent gate for ecological outflow in a hydropower station, including: a prediction module, a decision module, a constraint module, a simulation module, and an optimization module; the prediction module is used to collect ecological hydraulic operation data in real time and perform preprocessing; and uses an LSTM model to predict the downstream ecological water demand curve; the decision module is used to combine the ecological water demand prediction curve and the preprocessed ecological hydraulic operation data, and use a fuzzy PD+I control algorithm to perform rule reasoning to obtain a preliminary set of gate control parameters; the constraint module is used to compare the gate's safe operation constraints using numerical constraint analysis methods, perform boundary mapping and smoothing adjustments, and obtain a set of constraint-compliant gate control parameters; the simulation module is used to perform rapid simulation verification based on the constraint-compliant gate control parameter set and the ecological water demand prediction curve using a digital twin model to obtain a verified gate control parameter set; the optimization module is used to fuse gate control commands and generate gate execution strategies using a multi-objective dynamic optimization method to obtain the optimal intelligent gate control scheme.

[0109] This embodiment also provides a computer device applicable to the control method of the intelligent gate for ecological outflow of hydropower station, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the control method of the intelligent gate for ecological outflow of hydropower station as proposed in the above embodiment.

[0110] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0111] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements the control method for the intelligent gate for ecological outflow of hydropower station as proposed in the above embodiment. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0112] In summary, this invention employs an LSTM model with multivariate time series input and a dynamic sliding window mechanism to capture the nonlinear evolution of ecological hydraulic data, generating a high-precision water demand prediction curve to provide a reliable target benchmark for the control algorithm. The fuzzy PD+I control algorithm, based on the fusion of the prediction curve and real-time data, enables the gate control command to accurately track changes in ecological water demand and suppress flow fluctuations, while also adaptively adjusting the control intensity to avoid mechanical overshoot. This provides a highly stable and low-disturbance control foundation for subsequent constraint verification and multi-objective optimization, achieving a highly efficient balance between ecological protection goals and gate operational safety.

[0113] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A control method of an intelligent gate for ecological low discharge of a hydropower station, characterized in that: include, Real-time collection of ecological hydraulic operation data, preprocessing, and the use of an LSTM model to predict the downstream ecological water demand curve; By combining the ecological water demand prediction curve with the pre-processed ecological hydraulic operation data, the preliminary control parameter set of the gate is obtained through rule reasoning using the fuzzy PD+I control algorithm. By comparing the constraints on the safe operation of the gate using numerical constraint analysis, boundary mapping and smoothing adjustments are performed to obtain the set of control parameters for the constraint-compliant gate. The specific steps are as follows. Based on the preliminary set of control parameters, the numerical constraint analysis method is used to compare each parameter with the gate's safe operation constraints to obtain the difference between the control parameters and the safety boundary. By using the boundary mapping method, the initial control parameter set of the gate that exceeds the limit is compressed and stretched to the range allowed by the gate's safe operation constraints, thus generating the initial boundary adjustment control parameter set; By using a smoothing adjustment algorithm to adjust the rate of change of the initial boundary control parameter set, continuous optimization is performed to eliminate abrupt changes, thereby obtaining the control parameter set for the constraint-compliant gate. Based on the constrained and compliant gate control parameter set and the ecological water demand prediction curve, a rapid simulation verification was performed using a digital twin model to obtain a verified gate control parameter set. The optimal intelligent gate control scheme is obtained by fusing gate control commands and generating gate execution strategies using a multi-objective dynamic optimization method. The specific steps are as follows. Based on the verification of the gate control parameter set and the ecological water demand prediction curve, the objective function is optimized by a multi-objective weight allocation method to obtain the multi-objective optimization parameter set; A set of gate control instructions with optimal solution candidates is obtained by searching the solution space and selecting the Pareto front through a dynamic multi-objective optimization algorithm. By using the strategy fusion method, under the condition of satisfying the multi-objective optimization parameter set, the optimal solution candidates are screened and normalized and synthesized to obtain a unified scheduling execution instruction set; The optimal intelligent gate control scheme is generated by converting the data to the standard Modbus RTU format through format mapping and communication packaging methods.

2. The control method of the intelligent gate for ecological low discharge of a hydropower station according to claim 1, characterized in that: The ecological hydraulic operation data includes water flow data, water level data, upstream and downstream pressure difference data, water temperature data, and dissolved oxygen data; The preprocessing includes filtering and denoising, time synchronization, missing value imputation, and normalization.

3. The control method of the intelligent gate for ecological low discharge of the hydropower station according to claim 2, characterized in that: The specific steps for using the LSTM model to predict the downstream ecological water demand curve are as follows. By sliding and stitching the preprocessed historical ecohydraulic operation data into an input sequence in chronological order, a multivariate time series input matrix is ​​generated. The optimal time step and prediction length are analyzed by using a dynamic sliding window mechanism to generate the structural configuration parameters of the LSTM model. The Adam optimization algorithm is then used to train the LSTM model, resulting in an LSTM prediction model that fits the historical ecological flow evolution pattern. Based on the preprocessed ecological hydraulic operation data, the downstream ecological water demand prediction curve is obtained by sequential inference through the LSTM prediction model.

4. The control method of the intelligent gate for ecological low discharge of the hydropower station according to claim 3, characterized in that: The process involves combining the ecological water demand prediction curve with preprocessed ecological hydraulic operation data, and using a fuzzy PD+I control algorithm to perform rule-based reasoning to obtain the initial control parameter set for the gate. The specific steps are as follows: Based on the ecological water demand prediction curve and the preprocessed ecological hydraulic operation data, ecological regulation data is obtained by fusing them through time alignment and feature normalization fusion methods, and fuzzification processing is performed using fuzzy membership rules to generate fuzzy ecological regulation data. The gate control rules are matched and reasoned through a fuzzy rule base to obtain fuzzy gate control quantities. The PD+I defuzzification method is then used for quantification calculation to generate a preliminary gate control parameter set including gate opening set value, gate opening speed set value, and gate pre-pressure differential set value.

5. The control method of the intelligent gate for ecological low discharge of the hydropower station according to claim 1, characterized in that: The control parameter set for the constrained and compliant gate, based on the ecological water demand prediction curve, is rapidly simulated and verified using a digital twin model to obtain a validated gate control parameter set. The specific steps are as follows. Based on the historical constraint compliance gate control parameter set and the historical ecological water demand prediction curve, the digital twin model is trained by supervised learning method to fit parameters. Based on the set of constrained and compliant gate control parameters and the ecological water demand prediction curve, the driving data is obtained by combining the parameters through feature alignment and time series fusion. The trained digital twin model is then used for rapid dynamic simulation to simulate the flow change process after the execution of the constrained and compliant gate control parameters, and the simulated flow response results are obtained. By comparing the deviation analysis method with the ecological water demand prediction curve, the flow satisfaction and flow-pressure difference stability index are calculated to obtain the performance evaluation results. A predefined threshold judgment strategy is adopted to analyze the positive and negative tolerance of the predicted demand and the stability of the pressure difference fluctuation, and to obtain the verification gate control parameter set.

6. A control system for an intelligent gate for ecological outflow discharge in a hydropower station, based on the control method for the intelligent gate for ecological outflow discharge in a hydropower station as described in any one of claims 1 to 5, characterized in that: It includes a prediction module, a decision-making module, a constraint module, a simulation module, and an optimization module; The prediction module is used to collect ecological hydraulic operation data in real time and perform preprocessing; it uses an LSTM model to predict the downstream ecological water demand curve. The decision module is used to combine the ecological water demand prediction curve with the pre-processed ecological hydraulic operation data, and obtain the preliminary control parameter set of the gate through rule reasoning using the fuzzy PD+I control algorithm. The constraint module is used to compare the constraints on the safe operation of the gate using numerical constraint analysis methods, perform boundary mapping and smooth adjustment, and obtain a set of control parameters for the gate that comply with constraints. The simulation module is used to perform rapid simulation verification based on the constrained and compliant gate control parameter set and the ecological water demand prediction curve through a digital twin model, thereby obtaining a verified gate control parameter set. The optimization module is used to fuse gate control commands and generate gate execution strategies through a multi-objective dynamic optimization method to obtain the optimal intelligent gate control scheme.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the control method for the intelligent gate for ecological outflow of hydropower station as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the control method for the intelligent gate for ecological outflow of hydropower station as described in any one of claims 1 to 5.

Images