Multi-sensor data fusion-based intelligent regulation and control method and system for crossing sluice gate

Abstract

The invention provides a crossing sluice intelligent regulation and control method and system based on multi-sensor data fusion, and the method comprises the steps: obtaining multi-source synchronous data collected by a water level sensor, a flow sensor and a structural stress sensor; evaluating whether a gate overload risk exists in the current working condition according to the multi-source synchronous data; when the overload risk exists, the dynamic model is used for obtaining the target gate opening degree; the gate opening degree is regulated and controlled according to the target gate opening degree; and the regulated gate opening degree is obtained to optimize the dynamic model. Through multi-source sensor data fusion and dynamic model prediction control, real-time early warning and self-adaptive regulation and control of the gate overload risk are realized, a closed-loop optimization system is formed, and compared with a traditional method, the control precision is remarkably improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of industrial control systems, and in particular to a method and system for intelligently controlling a road culvert gate based on multi-sensor data fusion, and the classification numbers thereof include G05B23 / 02 and G05B13 / 04. Background Art

[0002] With the rapid development of water conservancy projects, culverts, as important hydraulic structures, play a vital role in water resource scheduling, flood control and drainage, and irrigation and water supply. Intelligent control of culverts is a key technology in modern water conservancy projects, directly related to the rational use of water resources and the safe operation of projects.

[0003] Currently, traditional culvert control systems primarily utilize relay-contactor control, which relies on manual water level monitoring and gate opening adjustment to control water flow. This control method suffers from numerous drawbacks, including heavy wiring workload and low system reliability, making it difficult to meet the precise control requirements of modern water conservancy projects.

[0004] Therefore, it is urgent to develop an intelligent control method for road culvert gates to improve the accuracy of gate control. Summary of the Invention

[0005] The technical problem to be solved by the present disclosure is to provide an intelligent control method for a road culvert gate based on multi-sensor data fusion in response to the problem of low control accuracy of the gate control system in the prior art.

[0006] The technical solution adopted by the present disclosure to solve its technical problems is:

[0007] A method for intelligent control of a road culvert gate based on multi-sensor data fusion is provided, comprising: acquiring multi-source synchronous data collected by a sensing component, the sensing component comprising a water level sensor, a flow sensor and a structural stress sensor, the multi-source synchronous data comprising water level, flow and structural stress; evaluating a current working condition based on the multi-source synchronous data, wherein an evaluation result of the current working condition is one of: the presence of a gate overload risk and the absence of a gate overload risk; if the evaluation result is that there is a gate overload risk, obtaining a target gate opening using a dynamic model; regulating the gate opening according to the target gate opening; and obtaining the gate opening after regulation to optimize the dynamic model.

[0008] Through multi-source sensor data fusion and dynamic model predictive control, real-time early warning and adaptive regulation of gate overload risks are achieved, forming a closed-loop optimization system, which significantly improves control accuracy compared to traditional methods.

[0009] Preferably, the acquisition of multi-source synchronous data collected by the sensing component includes: collecting raw data through pre-deployed water level sensors, flow sensors and structural stress sensors; using a sliding window interpolation method to align the timestamps of the raw data to obtain the multi-source synchronous data, wherein the window size is 2 to 5 sampling cycles.

[0010] The sliding window interpolation method is used to achieve high-precision time synchronization of multi-sensor data with an error of <3ms, eliminating the misjudgment of working conditions caused by sampling delays in traditional methods.

[0011] Furthermore, the evaluating the current working condition based on the multi-source synchronous data includes: preprocessing the multi-source synchronous data, including data cleaning and missing value filling; inputting the preprocessed data into a working condition evaluation model, and outputting an evaluation result.

[0012] The data integrity and reliability of the input working condition assessment model are ensured through the preprocessing process of data cleaning and missing value filling.

[0013] Furthermore, the preprocessing of the multi-source synchronous data includes: extracting three types of time series features, namely water level deviation rate, flow mutation rate, and stress change rate, from the multi-source synchronous data; normalizing the water level deviation rate, flow mutation rate, and stress change rate to generate standardized data in the [0,1] interval.

[0014] Three types of standardized characteristic parameters, namely water level deviation rate, flow mutation rate and stress change rate, are constructed to convert the original data into interpretable engineering indicators and eliminate the interference of dimensional differences on the evaluation model.

[0015] Preferably, the current working condition is evaluated based on the structured data, including: performing a fusion analysis of the water level deviation rate, flow mutation rate, and stress change rate through the DS evidence theory fusion algorithm to obtain the confidence level of the gate overload risk; if the confidence level of the gate overload risk exceeds a preset value, it is determined that the current working condition is subject to a gate overload risk.

[0016] The DS evidence theory is applied to integrate multi-dimensional feature parameters to achieve confidence quantitative assessment of overload risk, which significantly reduces the misjudgment rate.

[0017] Furthermore, the method of obtaining the target gate opening by using the dynamic model includes: using the dynamic model to solve the optimal opening sequence with the goal of minimizing water level deviation and energy consumption; and obtaining the target gate opening according to the optimal opening sequence.

[0018] The multi-objective optimization algorithm based on the state space model can reduce energy consumption while controlling water level deviation, achieving dual optimization of safety and economy.

[0019] Furthermore, regulating the gate opening according to the target gate opening includes: generating a gate regulating instruction according to the target gate opening; and converting the gate regulating instruction into a PLC executable signal to control the gate opening.

[0020] The precise execution of control instructions is achieved through the PLC signal conversion mechanism, with a positioning accuracy of ±0.5° and a transmission delay of <50ms, ensuring strict synchronization between control instructions and physical actions.

[0021] Preferably, the obtaining of the gate opening after regulation to optimize the preset dynamic model includes: collecting data through a gate opening sensor to verify whether the gate opening is consistent with the target gate opening; and adjusting the parameter weights of the dynamic model according to the verification results to improve the regulation accuracy.

[0022] Establish a control effect verification and model parameter online update mechanism to enable the dynamic model to continuously evolve with running time, exponentially improve the control accuracy, and shorten the update cycle.

[0023] A multi-sensor data fusion-based intelligent control system for road culvert gates is also provided, the system comprising:

[0024] A data acquisition module is used to acquire multi-source synchronous data collected by a sensing component, wherein the sensing component includes a water level sensor, a flow sensor, and a structural stress sensor, and the multi-source synchronous data includes water level, flow, and structural stress;

[0025] a working condition evaluation module, configured to evaluate a current working condition based on the multi-source synchronous data, wherein an evaluation result of the current working condition is one of: the presence of a gate overload risk and the absence of a gate overload risk;

[0026] A target opening module is used to obtain a target gate opening using a dynamic model if the assessment result indicates that there is a gate overload risk;

[0027] An opening control module, configured to control the gate opening according to the target gate opening;

[0028] The model optimization module is used to obtain the gate opening after regulation to optimize the preset dynamic model.

[0029] The beneficial effects of the present disclosure are:

[0030] Through multi-source sensor data fusion and dynamic model predictive control, real-time early warning and adaptive regulation of gate overload risks are achieved, forming a closed-loop optimization system, which significantly improves control accuracy compared to traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1A schematic diagram of a flow chart of an intelligent control method for a road culvert gate based on multi-sensor data fusion according to an embodiment of the present disclosure is shown;

[0032] Figure 2 A flow chart showing step S100 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0033] Figure 3 A flow chart showing step S200 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0034] Figure 4 A schematic flow chart of step S210 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0035] Figure 5 A schematic flow chart of step S220 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0036] Figure 6 A flow chart showing step S300 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0037] Figure 7 A flow chart showing step S400 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0038] Figure 8 A schematic flow chart showing step S500 of the intelligent control method for a road culvert gate according to an embodiment of the present disclosure is shown;

[0039] Figure 9 A structural schematic diagram of an intelligent control system for a road culvert gate according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0040] The technical solutions of the present disclosure will be described clearly and completely below with reference to the accompanying drawings and embodiments. Obviously, the embodiments described are only a portion of the embodiments of the present disclosure, and not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present disclosure without inventive effort are also within the scope of protection of the present disclosure.

[0041] Example 1

[0042] Figure 1 A flow chart of a method for intelligently controlling a road culvert gate based on multi-sensor data fusion according to an embodiment of the present disclosure is shown. Figure 1 As shown, the method may include the following steps:

[0043] Step S100: Acquire multi-source synchronous data collected by a sensing component, wherein the sensing component may include a water level sensor, a flow sensor, and a structural stress sensor, and correspondingly, the multi-source synchronous data may include water level, flow, and structural stress data.

[0044] In this embodiment, the water level sensor can be an ultrasonic water level meter, installed at key locations upstream and downstream of the culvert. The flow sensor can be an electromagnetic flow meter, installed at the inlet and outlet of the culvert. The structural stress sensor can be a strain gauge stress sensor, installed at key structural locations of the gate. These sensors can collect data at a frequency of once per second, generating a raw data stream.

[0045] Figure 2 FIG1 shows a flow chart of step S100 of an embodiment of the present disclosure. Figure 2 As shown, the acquisition of multi-source synchronous data collected by the sensing component may include:

[0046] Step S110: Collect raw data using pre-deployed water level sensors, flow sensors, and structural stress sensors. The water level sensors can be configured to collect upstream and downstream water level data with an accuracy of ±0.5 cm. The flow sensors can be configured to collect inlet and outlet flow data with an accuracy of ±0.2 m³ / s. The structural stress sensors can be configured to collect gate body structural stress data with an accuracy of ±0.01 MPa.

[0047] Step S120: Use a sliding window interpolation method to align the timestamps of the raw data to obtain multi-source synchronized data. The window size can be 2 to 5 sampling periods. In this embodiment, the sliding window size is set to 3 sampling periods, and the interpolation accuracy is 2ms. The specific operation is: for each sensor data sequence, a time window containing 3 sampling points is established. Within the window, a cubic spline interpolation algorithm is used to align the data from different sensors to a unified timestamp, thereby generating time-synchronized multi-source data.

[0048] Step S200: Evaluate the current working condition based on the multi-source synchronized data, wherein the evaluation result of the current working condition may be that there is a gate overload risk or there is no gate overload risk.

[0049] Figure 3 FIG2 is a flow chart showing step S200 of an embodiment of the present disclosure. Figure 3 As shown, evaluating the current working condition based on the multi-source synchronous data may include:

[0050] Step S210: Preprocess the multi-source synchronized data, including data cleaning and missing value filling. Data cleaning involves removing outliers and noise, and using a median filter algorithm to process sudden changes with a filter window size of five data points. Missing values are filled using linear interpolation. If there are more than five consecutive missing data points, historical data from the same period is used as a replacement.

[0051] Figure 4 FIG2 shows a flow chart of step S210 of an embodiment of the present disclosure. Figure 4 As shown, the preprocessing of the multi-source synchronous data may include:

[0052] Step S211: Extract three types of time series features from the multi-source synchronized data: water level deviation rate, flow mutation rate, and stress change rate. The water level deviation rate is calculated as (measured water level - design water level) / design water level × 100%; the flow mutation rate is calculated as (current flow - previous moment flow) / previous moment flow × 100%; and the stress change rate is calculated as (current stress - previous moment stress) / previous moment stress × 100%.

[0053] Step S212: Normalize the water level deviation rate, flow mutation rate, and stress change rate to generate standardized data in the [0, 1] interval. Normalization uses the Min-Max normalization method, calculated as follows: X_norm = (X - X_min) / (X_max - X_min), where X is the original eigenvalue, and X_min and X_max are the historical minimum and maximum values of the feature, respectively.

[0054] Step S220: Input the pre-processed data into the working condition assessment model and output an assessment result, which is one of: the presence of gate overload risk and the absence of gate overload risk.

[0055] Figure 5 FIG2 shows a flow chart of step S220 of an embodiment of the present disclosure. Figure 5 As shown, the inputting of the pre-processed data into the working condition evaluation model and the outputting of the evaluation results may include:

[0056] Step S221: Perform fusion analysis on the water level deviation rate, flow mutation rate, and stress change rate using the DS evidence theory fusion algorithm to obtain the confidence level of the gate overload risk.

[0057] The fusion process of DS evidence theory is as follows:

[0058] First, a basic probability assignment function (BPA) is established for each feature, and the identification framework Θ = {overload risk, no overload risk} is defined.

[0059] For the water level deviation rate (WD), its BPA function is:

[0060] m1(overload risk) = WD

[0061] m1 (no overload risk) = 1-WD

[0062] m1(Θ) = 0

[0063] For the traffic mutation rate (FR), the BPA function is:

[0064] m2(overload risk) = FR

[0065] m2 (no overload risk) = 1-FR

[0066] m2(Θ) = 0

[0067] For the stress change rate (SC), the BPA function is:

[0068] m3(overload risk) = SC

[0069] m3 (no overload risk) = 1-SC

[0070] m3(Θ) = 0

[0071] Then, the DS combination rule is used to perform evidence fusion:

[0072] K = Σ m1(A)m2(B)m3(C), where A∩B∩C=∅

[0073] m(overload risk) = (1 / (1-K))×Σ m1(A)m2(B)m3(C), where A∩B∩C=overload risk

[0074] Step S222: If the confidence level of the gate overload risk exceeds a preset value, the current operating condition is determined to be at risk of gate overload. The preset value can be any value between 0.75 and 0.85. Preferably, the preset value is 0.8.

[0075] Step S300: If the evaluation result shows that there is a gate overload risk, the target gate opening is obtained using a dynamic model.

[0076] Figure 6 FIG3 shows a flow chart of step S300 of an embodiment of the present disclosure. Figure 6 As shown, the method of obtaining the target gate opening by using the dynamic model may include:

[0077] Step S310: Using the dynamic model, the optimal opening sequence is solved with the goal of minimizing water level deviation and energy consumption. The dynamic model can be a model predictive control model based on the state space equation, and its state space equation is as follows:

[0078] x(k+1) = Ax(k) + Bu(k) + Ew(k)

[0079] y(k) = Cx(k) + Du(k) + v(k)

[0080] Among them, x(k) is the system state vector, including state variables such as water level and flow; u(k) is the control input vector, that is, the gate opening; w(k) is the system disturbance; y(k) is the system output; v(k) is the measurement noise; A, B, C, D, and E are system matrices.

[0081] The objective function of model predictive control can be:

[0082] J = Σ[Q(y(k+i|k)-r(k+i))² + Ru(k+i-1)²]

[0083] Among them, Q is the output error weight matrix, R is the control input weight matrix, and r(k+i) is the reference trajectory.

[0084] By solving the above optimization problem, the optimal opening sequence {u(k|k), u(k+1|k),..., u(k+N-1|k)} for the next N moments is obtained.

[0085] Step S320: Obtain the target gate opening according to the optimal opening sequence. In this embodiment, the first element u(k|k) of the optimal opening sequence is taken as the target gate opening at the current moment.

[0086] Step S400: regulating the gate opening according to the target gate opening.

[0087] Figure 7 FIG4 shows a flow chart of step S400 of an embodiment of the present disclosure. Figure 7 As shown, regulating the gate opening according to the target gate opening may include:

[0088] Step S410: Generate a gate control instruction based on the target gate opening. The gate control instruction includes parameters such as the target opening value, the control rate, and the execution time. The control rate is dynamically calculated based on the difference between the current opening and the target opening. The larger the difference, the smaller the control rate, ensuring smooth gate operation.

[0089] Step S420: Convert the gate control command into a PLC executable signal to control the gate opening. After receiving the control command, the PLC control system converts it into an analog output signal (4-20mA) or a digital output signal, driving the gate actuator to adjust the gate opening at a specified rate.

[0090] Step S500: Obtain the gate opening after adjustment to optimize the dynamic model.

[0091] Figure 8 FIG. 5 is a flow chart showing step S500 of an embodiment of the present disclosure. Figure 8 As shown, the acquisition of the gate opening after regulation to optimize the preset dynamic model may include:

[0092] Step S510: Collect data from the gate opening sensor to verify whether the gate opening is consistent with the target gate opening. The gate opening sensor can use an angle encoder with an accuracy of 0.1 degrees to monitor the actual gate opening in real time. The actual opening is compared with the target opening. Any deviation exceeding ±1% is recorded as a control deviation.

[0093] Step S520: Adjust the parameter weights of the dynamic model based on the verification results to improve the control accuracy. Specifically, when there is a deviation between the actual opening and the target opening, the Q and R weight matrices in the model predictive control can be adjusted through an adaptive algorithm. The adjustment formula can be:

[0094] Q(k+1) = Q(k) + α·e(k)·e(k)ᵀ

[0095] R(k+1) = R(k) + β·Δu(k)·Δu(k)ᵀ

[0096] Where e(k) is the output error, Δu(k) is the control increment, and α and β are learning rates, which are set to 0.01 and 0.005 respectively.

[0097] Through the above steps, this embodiment realizes the intelligent control of the road culvert gate, can realize real-time early warning and adaptive control of the gate overload risk, form a closed-loop optimization system, and significantly improve the control accuracy compared with traditional methods.

[0098] Example 2

[0099] The embodiments of the present disclosure provide a method for intelligently controlling a road culvert gate based on multi-sensor data fusion, which may include the following steps:

[0100] Data acquisition step: Acquire multi-source synchronous data collected by the sensing component.

[0101] The sensing components include water level sensors, flow sensors, and structural stress sensors. The multi-source synchronized data includes water level, flow, and structural stress data. In this embodiment, the water level sensors are radar level gauges installed at key locations upstream and downstream of the culvert; the flow sensors are ultrasonic flow meters installed at the inlet and outlet of the culvert; and the structural stress sensors are fiber Bragg grating (FBG) stress sensors installed at key locations within the gate structure. These sensors collect data twice per second, generating a raw data stream.

[0102] Raw data is collected through pre-deployed water level sensors, flow sensors, and structural stress sensors. The water level sensors collect upstream and downstream water level data with an accuracy of ±0.3 cm; the flow sensors collect inlet and outlet flow data with an accuracy of ±0.15 m³ / s; and the structural stress sensors collect gate structural stress data with an accuracy of ±0.008 MPa.

[0103] A sliding window interpolation method is used to align the timestamps of the raw data to obtain multi-source synchronized data. In this embodiment, the sliding window size is set to 4 sampling periods, and the interpolation accuracy is 1ms. The specific operation is: for each sensor data sequence, a time window containing 4 sampling points is established. Within this window, the Lagrangian interpolation algorithm is used to align the data from different sensors to a unified timestamp, thereby generating time-synchronized multi-source data.

[0104] Working condition evaluation step: Evaluate the current working condition based on multi-source synchronized data.

[0105] Preprocess multi-source synchronized data, including data cleaning and missing value filling. Data cleaning involves removing outliers and noise. Noise points are processed using a wavelet transform denoising algorithm, using the db4 wavelet basis function and a decomposition layer of three. Missing values are filled using the Kalman filter prediction method. When there are more than three consecutive missing data points, historical data from the same period is used as a replacement.

[0106] Three types of time series features were extracted from multi-source synchronized data: water level deviation rate, flow mutation rate, and stress change rate. The water level deviation rate was calculated as (measured water level - design water level) / design water level × 100%; the flow mutation rate was calculated as (current flow - previous flow) / previous flow × 100%; and the stress change rate was calculated as (current stress - previous stress) / previous stress × 100%.

[0107] Normalize the water level deviation rate, flow mutation rate, and stress change rate to generate standardized data in the [0, 1] interval. This normalization uses the Z-score method followed by a Min-Max transformation. The calculation formula is: X_norm = (X -μ) / σ, where X is the original eigenvalue, μ is the mean of the feature, and σ is the standard deviation. The Z-score value is then mapped to the [0, 1] interval.

[0108] The DS evidence theory fusion algorithm is used to fuse and analyze the water level deviation rate, flow mutation rate, and stress change rate to obtain the confidence level of the gate overload risk. The DS evidence theory fusion process is as follows:

[0109] First, a basic probability assignment function (BPA) is established for each feature, and the identification framework Θ = {overload risk, no overload risk} is defined.

[0110] For the water level deviation rate (WD), its BPA function is:

[0111] m1(overload risk) = WD^2

[0112] m1(no overload risk) = (1-WD)^2

[0113] m1(Θ) = 1 - WD^2 - (1-WD)^2

[0114] For the traffic mutation rate (FR), the BPA function is:

[0115] m2(overload risk) = FR^2

[0116] m2(no overload risk) = (1-FR)^2

[0117] m2(Θ) = 1 - FR^2 - (1-FR)^2

[0118] For the stress change rate (SC), the BPA function is:

[0119] m3(overload risk) = SC^2

[0120] m3(no overload risk) = (1-SC)^2

[0121] m3(Θ) = 1 - SC^2 - (1-SC)^2

[0122] Then, the DS combination rule is used to perform evidence fusion:

[0123] K = Σ m1(A)m2(B)m3(C), where A∩B∩C=∅

[0124] m(overload risk) = (1 / (1-K))×Σ m1(A)m2(B)m3(C), where A∩B∩C=overload risk

[0125] If the confidence level of the gate overload risk exceeds the preset value of 0.8, the current working condition is determined to have a gate overload risk.

[0126] The pre-processed data is fed into the working condition assessment model, which outputs an assessment result, which can be either a gate overload risk or a gate overload risk.

[0127] Target opening step: If the assessment result shows that there is a gate overload risk, the target gate opening is obtained using the dynamic model.

[0128] The dynamic model is used to solve the optimal opening sequence with the goal of minimizing water level deviation and energy consumption. The dynamic model is a model predictive control model based on the state space equation, and its state space equation is as follows:

[0129] x(k+1) = Ax(k) + Bu(k) + Ew(k)

[0130] y(k) = Cx(k) + Du(k) + v(k)

[0131] Among them, x(k) is the system state vector, including state variables such as water level and flow; u(k) is the control input vector, that is, the gate opening; w(k) is the system disturbance; y(k) is the system output; v(k) is the measurement noise; A, B, C, D, and E are system matrices.

[0132] The objective function of model predictive control is:

[0133] J = Σ[Q(y(k+i|k)-r(k+i))² + Ru(k+i-1)² + SΔu(k+i-1)²]

[0134] Among them, Q is the output error weight matrix, R is the control input weight matrix, S is the control increment weight matrix, and r(k+i) is the reference trajectory.

[0135] By solving the above optimization problem, the optimal opening sequence {u(k|k), u(k+1|k),..., u(k+N-1|k)} for the next N moments is obtained.

[0136] The target gate opening is obtained according to the optimal opening sequence. In this embodiment, the weighted average of the first three elements of the optimal opening sequence is taken as the target gate opening at the current moment, with weights of 0.6, 0.3, and 0.1 respectively.

[0137] Opening control steps: Control the gate opening according to the target gate opening.

[0138] A gate control command is generated based on the target gate opening. This command includes parameters such as the target opening value, control rate, and execution time. The control rate utilizes an adaptive algorithm that dynamically calculates the optimal control rate based on the current water level difference, flow rate, and opening difference to ensure smooth gate operation.

[0139] The gate control instructions are converted into PLC executable signals to control the gate opening. After receiving the control instructions, the PLC control system transmits the instructions to the execution unit via the PROFIBUS fieldbus. The execution unit converts the instructions into PWM signals, driving the gate actuator to adjust the gate opening at the specified rate.

[0140] Model optimization step: obtain the gate opening after regulation to optimize the dynamic model.

[0141] The gate opening sensor collects data to verify whether the gate opening is consistent with the target gate opening. The gate opening sensor uses an absolute encoder with an accuracy of 0.05 degrees and monitors the actual gate opening in real time. The system compares the deviation between the actual opening and the target opening. Any deviation exceeding ±0.8% is recorded as a control deviation.

[0142] Based on the validation results, the dynamic model's parameter weights are adjusted to improve control accuracy. Specifically, when the actual opening deviates from the target opening, the system uses a particle swarm optimization algorithm to adjust the Q, R, and S weight matrices in the model predictive control to reduce the error between the system model and the actual system. During this adjustment process, the number of particles is set to 30, the number of iterations is set to 50, and the objective function is to minimize the historical control error.

[0143] Through the above steps, this embodiment realizes the intelligent control of the road culvert gate, can realize real-time early warning and adaptive control of the gate overload risk, form a closed-loop optimization system, and significantly improve the control accuracy compared with traditional methods.

[0144] Example 3

[0145] Figure 9 The schematic diagram of the structure of a road culvert gate intelligent control system based on multi-sensor data fusion according to an embodiment of the present disclosure is shown. The road culvert gate intelligent control system based on multi-sensor data fusion according to an embodiment of the present disclosure belongs to an industrial control system, see Figure 9 As shown, the system may include the following modules:

[0146] The data acquisition module 100 is used to acquire multi-source synchronized data collected by the sensing components. The sensing components include a water level sensor, a flow sensor, and a structural stress sensor. The multi-source synchronized data includes water level, flow, and structural stress data. In this embodiment, the data acquisition module includes a data acquisition unit and a data synchronization unit.

[0147] The data acquisition unit connects to various sensors via a fieldbus to collect raw data. The water level sensors are pressure gauges installed at key locations upstream and downstream of the culvert; the flow rate sensors are Doppler flowmeters installed at the inlet and outlet of the culvert; and the structural stress sensors are resistance strain gauges installed at key locations within the gate structure. These sensors collect data at a rate of 1.5 times per second, generating a raw data stream.

[0148] The data synchronization unit uses a sliding window interpolation method to align the timestamps of the raw data to generate multi-source synchronized data. In this embodiment, the sliding window size is set to 5 sampling periods, and the interpolation accuracy is 3ms. Specifically, for each sensor data sequence, a time window containing 5 sampling points is established. Within this window, the Newton interpolation algorithm is used to align the data from different sensors to a unified timestamp, thereby generating time-synchronized multi-source data.

[0149] The working condition evaluation module 200 is used to evaluate the current working condition based on multi-source synchronous data. The working condition evaluation module includes a data pre-processing unit and a risk evaluation unit.

[0150] The data preprocessing unit preprocesses multi-source synchronized data, including data cleaning and missing value filling. Data cleaning involves removing outliers and noise. Noise points are processed using the Savitzky-Golay filter algorithm, with a filter window size of 7 data points and a polynomial order of 3. Missing values are filled using the ARIMA model forecasting method. When there are more than 4 consecutive missing data points, historical data from the same period is used as a replacement.

[0151] The data preprocessing unit extracts three types of time series features from multi-source synchronized data: water level deviation rate, flow mutation rate, and stress change rate. The water level deviation rate is calculated as (measured water level - design water level) / design water level × 100%; the flow mutation rate is calculated as (current flow - previous flow) / previous flow × 100%; and the stress change rate is calculated as (current stress - previous stress) / previous stress × 100%.

[0152] The data preprocessing unit normalizes the water level deviation rate, flow mutation rate, and stress change rate to generate standardized data in the [0, 1] interval. This normalization uses the Sigmoid function, calculated as follows: X_norm = 1 / (1+e^(-α(X-β))), where X is the original eigenvalue and α and β are adjustment parameters, set to 0.5 and the mean, respectively.

[0153] The risk assessment unit uses the DS evidence theory fusion algorithm to perform a fusion analysis of the water level deviation rate, flow mutation rate, and stress change rate to obtain the confidence level of the gate overload risk. The DS evidence theory fusion process is as follows:

[0154] First, a basic probability assignment function (BPA) is established for each feature, and the identification framework Θ = {overload risk, no overload risk} is defined.

[0155] For the water level deviation rate (WD), its BPA function is:

[0156] m1(overload risk) = WD·w1

[0157] m1(no overload risk) = (1-WD)·w1

[0158] m1(Θ) = 1 - w1

[0159] For the traffic mutation rate (FR), the BPA function is:

[0160] m2(overload risk) = FR·w2

[0161] m2(no overload risk) = (1-FR)·w2

[0162] m2(Θ) = 1 - w2

[0163] For the stress change rate (SC), the BPA function is:

[0164] m3 (overload risk) = SC w3

[0165] m3 (no overload risk) = (1-SC) w3

[0166] m3(Θ) = 1 - w3

[0167] Among them, w1, w2, and w3 are the weights of each feature, which are set to 0.4, 0.3, and 0.3 respectively.

[0168] Then, the DS combination rule is used to perform evidence fusion:

[0169] K = Σ m1(A)m2(B)m3(C), where A∩B∩C=∅

[0170] m(overload risk) = (1 / (1-K))×Σ m1(A)m2(B)m3(C), where A∩B∩C=overload risk

[0171] If the confidence level of the gate overload risk exceeds the preset value of 0.7, the current working condition is determined to have a gate overload risk.

[0172] The result of the working condition assessment is either that there is a gate overload risk or that there is no gate overload risk.

[0173] The target opening module 300 is used to obtain the target gate opening using the dynamic model if the evaluation result shows that there is a gate overload risk. The target opening module includes a model calculation unit and an opening optimization unit.

[0174] The model calculation unit uses a dynamic model to solve the optimal opening sequence with the goal of minimizing water level deviation and energy consumption. The dynamic model is a model predictive control model based on the state space equation, and its state space equation is as follows:

[0175] x(k+1) = Ax(k) + Bu(k) + Ew(k)

[0176] y(k) = Cx(k) + Du(k) + v(k)

[0177] Among them, x(k) is the system state vector, including state variables such as water level and flow; u(k) is the control input vector, that is, the gate opening; w(k) is the system disturbance; y(k) is the system output; v(k) is the measurement noise; A, B, C, D, and E are system matrices.

[0178] The objective function of model predictive control is:

[0179] J = Σ[Q(y(k+i|k)-r(k+i))² + Ru(k+i-1)² + SΔu(k+i-1)² + T·max(0, g(y(k+i|k)))]

[0180] Where Q is the output error weight matrix, R is the control input weight matrix, S is the control increment weight matrix, T is the constraint violation penalty weight, g(y) is the system constraint function, and r(k+i) is the reference trajectory.

[0181] By solving the above optimization problem, the optimal opening sequence {u(k|k), u(k+1|k),..., u(k+N-1|k)} for the next N moments is obtained.

[0182] The opening optimization unit obtains the target gate opening according to the optimal opening sequence. In this embodiment, a rolling optimization strategy is adopted, where the first element u(k|k) of the optimal opening sequence is taken as the target gate opening at the current moment, and the optimization problem is solved again in the next control cycle.

[0183] The opening control module 400 is used to control the gate opening according to the target gate opening. The opening control module includes an instruction generation unit and an execution control unit.

[0184] The command generation unit generates gate control commands based on the target gate opening. These commands include parameters such as the target opening, control rate, and execution time. The control rate is calculated based on hydraulic conditions and gate characteristics, using a piecewise linear function to ensure smooth gate operation.

[0185] The execution control unit converts gate control commands into PLC executable signals to control the gate opening. After receiving the control commands, the PLC control system transmits them to the execution unit via the EtherCAT fieldbus. The execution unit converts the commands into analog output signals (0-10V), driving the gate actuator to adjust the gate opening at the specified rate.

[0186] The model optimization module 500 is used to obtain the gate opening after adjustment to optimize the preset dynamic model. The model optimization module includes a feedback verification unit and a parameter optimization unit.

[0187] The feedback verification unit collects data from the gate opening sensor to verify whether the gate opening is consistent with the target gate opening. The gate opening sensor uses a magnetostrictive displacement sensor with an accuracy of 0.2mm, which monitors the actual gate opening in real time. The system compares the deviation between the actual opening and the target opening. Any deviation exceeding ±1.2% is recorded as a control deviation.

[0188] The parameter optimization unit adjusts the dynamic model's parameter weights based on the validation results to improve control accuracy. Specifically, when the actual opening deviates from the target opening, the system uses a genetic algorithm to optimize the model parameters, including the system matrices A, B, C, D, and E, and the control weight matrices Q, R, and S. The genetic algorithm uses a population size of 50, a crossover probability of 0.8, a mutation probability of 0.1, and a maximum number of iterations of 100, with the fitness function minimizing the historical prediction error.

[0189] Through the collaborative work of the above modules, the intelligent control system for road culvert gates based on multi-sensor data fusion in this embodiment can achieve real-time early warning and adaptive control of gate overload risks, forming a closed-loop optimization system, which significantly improves control accuracy compared with traditional methods.

[0190] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present disclosure.

Claims

1. A method for intelligent control of road culvert gates based on multi-sensor data fusion, characterized in that: The method comprises: Acquire multi-source synchronous data collected by a sensing component, wherein the sensing component includes a water level sensor, a flow sensor, and a structural stress sensor, and the multi-source synchronous data includes water level, flow, and structural stress; evaluating a current operating condition based on the multi-source synchronized data, wherein an evaluation result of the current operating condition is one of: a gate overload risk existing and a gate overload risk not existing; If the assessment result shows that there is a gate overload risk, the target gate opening is obtained using the dynamic model; regulating the gate opening according to the target gate opening; The gate opening after regulation is obtained to optimize the dynamic model.

2. The method according to claim 1, characterized in that The acquiring of multi-source synchronous data collected by the sensing component includes: Collect raw data through pre-deployed water level sensors, flow sensors, and structural stress sensors; The timestamps of the original data are aligned using a sliding window interpolation method to obtain the multi-source synchronous data, wherein the window size is 2 to 5 sampling periods.

3. The method according to claim 2, characterized in that The evaluating the current operating condition according to the multi-source synchronous data includes: Preprocessing the multi-source synchronized data, including data cleaning and missing value filling; The preprocessed data is input into the working condition evaluation model and the evaluation results are output.

4. The method according to claim 3, characterized in that The preprocessing of the multi-source synchronous data includes: Extracting three types of time series features, namely water level deviation rate, flow mutation rate and stress change rate, from the multi-source synchronous data; The water level deviation rate, flow mutation rate, and stress change rate are normalized to generate standardized data in the [0,1] interval.

5. The method according to claim 4, characterized in that The pre-processed data is input into the working condition evaluation model and the evaluation results are output, including: The confidence level of gate overload risk is obtained by fusion analysis of water level deviation rate, flow mutation rate, and stress change rate using the DS evidence theory fusion algorithm. If the confidence level of the gate overload risk exceeds a preset value, it is determined that the current working condition has a gate overload risk.

6. The method according to claim 5, characterized in that The method of obtaining the target gate opening by using the dynamic model includes: Using the dynamic model, the optimal opening sequence is solved with the goal of minimizing water level deviation and energy consumption; The target gate opening is obtained according to the optimal opening sequence.

7. The method according to claim 6, characterized in that The regulating the gate opening according to the target gate opening includes: generating a gate control instruction according to the target gate opening; The gate control instruction is converted into a PLC executable signal to control the opening of the gate.

8. The method according to claim 7, characterized in that The obtaining of the gate opening after regulation to optimize the preset dynamic model includes: Collect data through the gate opening sensor to verify whether the gate opening is consistent with the target gate opening; Adjust the parameter weights of the dynamic model based on the verification results to improve the control accuracy.

9. The method according to any one of claims 6 to 8, characterized in that: The dynamic model is a model predictive control model based on state space equations.

10. An intelligent control system for road culvert gates based on multi-sensor data fusion, characterized in that: The system comprises: A data acquisition module is used to acquire multi-source synchronous data collected by a sensing component, wherein the sensing component includes a water level sensor, a flow sensor, and a structural stress sensor, and the multi-source synchronous data includes water level, flow, and structural stress; a working condition evaluation module, configured to evaluate a current working condition based on the multi-source synchronous data, wherein an evaluation result of the current working condition is one of: the presence of a gate overload risk and the absence of a gate overload risk; A target opening module is used to obtain a target gate opening using a dynamic model if the assessment result indicates that there is a gate overload risk; An opening control module, configured to control the gate opening according to the target gate opening; The model optimization module is used to obtain the gate opening after regulation to optimize the preset dynamic model.

Images