Adaptive fuzzy PID flow field control method and device

Abstract

The application discloses a self-adaptive fuzzy PID flow field control method and device, and the method comprises the following steps: generating initial PID gain parameters in real time through a fuzzy reasoning module, introducing an RBF neural network module to dynamically optimize and compensate the initial parameters, and generating corrected gains. The final PID parameters are dynamically synthesized by using a weighted fusion formula, and the weight coefficient is adjustable to balance the contribution of fuzzy rules and neural network learning. By using the nonlinear approximation ability of the RBF network, the network parameters are updated in real time through the gradient descent method, the Jacobian matrix of the controlled system is output, and the sensitivity of the control quantity to the input change is accurately identified. In combination with the environmental sensor data, the flow feedback and the historical deviation, the parameter update rate is dynamically adjusted through the momentum term and the exponential decay function, and a closed-loop control loop of monitoring-fuzzy reasoning-RBF optimization-PID output-feedback is formed. The application improves the accuracy of flow field control.

Description

Technical Field

[0001] This application relates to flow field control technology, and in particular to an adaptive fuzzy PID flow field control method and apparatus. Background Technology

[0002] Traditional fuzzy PID (proportional, integral, derivative) controllers face significant limitations in complex industrial scenarios such as port water quality control. They rely on a pre-set fuzzy rule base (such as a seven-level classification system including NB, NM, NS, ZO, PS, PM, and PB) to output PID gain parameters. When sudden environmental disturbances cause control deviations to exceed the fuzzy domain, the rule base cannot provide effective parameters. Furthermore, the introduction of nonlinear characteristics from actuators such as control valves further contributes to steady-state errors, while fixed-parameter PID controllers lack dynamic adaptability. Improvements to traditional PID control typically involve adjusting the controller's three parameters to enhance the system's dynamic and static performance. However, in practical applications, changes in external environmental disturbances and internal performance parameters of control valves can lead to decreased control accuracy. Additionally, PID controllers suffer from weak anti-interference capabilities and limited applicability in nonlinear systems.

[0003] In the field of control valves, although traditional PID control algorithms are widely used, they exhibit three problems in the regulation of large-diameter valves: firstly, insufficient stability, leading to overshoot and system oscillations when flow changes suddenly occur; secondly, sluggish dynamic response, making it difficult for fixed parameters to match real-time load changes; and thirdly, deterioration of steady-state accuracy, with long-term operation causing accumulated flow deviations and reduced accuracy. While existing improvement schemes can locally optimize performance, they still suffer from strong reliance on human experience, delayed parameter adjustment response, and a lack of real-time self-learning mechanisms. Summary of the Invention

[0004] This application provides an adaptive fuzzy PID flow field control method and apparatus to at least solve the above-mentioned technical problems existing in the prior art.

[0005] According to a first aspect of this application, an adaptive fuzzy PID flow field control method is provided, comprising:

[0006] Collect water quality index data in the target water area, and determine the deviation and the rate of change of the deviation between the water quality index data and the target value; the water quality index data includes at least inorganic nitrogen concentration and reactive phosphate concentration; input the deviation and the rate of change of the deviation into the fuzzy inference module, perform fuzzification processing through a seven-level fuzzy classification system, and generate initial PID gain parameters based on a preset fuzzy rule base;

[0007] The initial PID gain parameters and real-time feedback values ​​are input into the radial basis function network (RBF). The Jacobian matrix is ​​calculated using the radial basis function, and the network parameters are updated online using the gradient descent method.

[0008] The initial PID gain parameters output by the fuzzy inference module and the correction parameters output by the RBF are weighted and fused through an adaptive parameter fusion mechanism to obtain the final PID parameters.

[0009] The control signal is generated based on the final PID parameters, which drives the actuator to adjust the port flow field to form a closed-loop control circuit.

[0010] In some optional embodiments, the method further includes:

[0011] The rules in the fuzzy rule base are generated in the form of conditional statements, and an intelligent environmental perception system is established to form the fuzzy rule base;

[0012] The correction amounts for the proportional, integral, and differential coefficients are calculated in real time based on the rules in the fuzzy rule base. , , The correction coefficient is dynamically adjusted using an exponential decay function, the mathematical expression of which is: ,in, , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. This represents the elapsed time since the initial moment.

[0013] proportionality coefficient The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. The differential coefficients from the previous moment Compared with the current adjustment amount Determined jointly; among them, , , They represent in The initial settings of each parameter at any given time. , , It is a parameter adjustment amount determined based on a fuzzy rule base.

[0014] In some optional embodiments, the step of calculating the Jacobian matrix using radial basis functions and updating the network parameters online using gradient descent includes:

[0015] The Gaussian function is chosen as the specific implementation of the radial basis function, and its mathematical expression is shown below:

[0016]

[0017] in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m; Calculate the neural network output based on the weight coefficients of the RBF neural network. and network recognition performance index function ,as follows:

[0018]

[0019]

[0020] in, This represents the measured values ​​of water quality indicators in the target water area. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron;

[0021] Iteratively update the node center vector Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

[0022] In some optional embodiments, the method further includes:

[0023] During the network parameter update process, the following operations are performed sequentially at each time step t: based on the network learning rate Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change ;

[0024] Update weight parameters By using an adder Weighting coefficients at different times and Add them together and include them in the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference in time with respect to the weighting coefficient Update;

[0025] Based on network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector calculated by the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change ;

[0026] Update the base width vector At that time, The basis width vector at time step and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector using a multiplier exist Time and The difference at each time step is used as the updated weight coefficient.

[0027] In some optional embodiments, the method further includes:

[0028] Update the computation center vector Each component Based on network learning efficiency Error difference Updated weight coefficient values Input vector With the center vector components The difference, and the magnitude of the updated base width vector The change in the central vector components is calculated using dividers and multipliers. ;

[0029] Update center vector components At that time, the adder will be used to... The center vector component at time t and Add them together, and include the momentum term. Multiply by the central vector component exist Time and The difference in time;

[0030] Accumulate all neurons using an adder. and , central vector components The control change output from the PID controller to the regulating valve The difference is the same The product of the ratios yields the Jacobian matrix. .

[0031] In some optional embodiments, generating the control signal based on the final PID parameters includes:

[0032] PID parameter increments output by RBF neural network , , Determined by the following formula:

[0033]

[0034] in, , , These represent the PID gain and learning rate, respectively.

[0035] The process of generating control signals based on the final PID parameters includes:

[0036] The control signal is determined by the following formula:

[0037]

[0038] Among them, control signals The calculation involves the following parameters: time point controller output value Time point Deviation change and time points The error of the second-order backward difference The second derivative used for approximation error, time point Deviation change ;

[0039] Calculating the output value of the PID controller During the process, the time points in the historical deviation data Deviation change and time point Deviation change Used for predicting trends in water quality indicators.

[0040] According to a second aspect of this application, an adaptive fuzzy PID flow field control device is provided, comprising:

[0041] The calculation unit is used to collect water quality index data in the target water area and determine the deviation and the rate of change of the deviation between the water quality index data and the target value; wherein, the water quality index data includes at least inorganic nitrogen concentration and reactive phosphate concentration; the fuzzy processing unit is used to input the deviation and the rate of change of the deviation into the fuzzy inference module, perform fuzzification processing through a seven-level fuzzy classification system, and generate initial PID gain parameters based on a preset fuzzy rule base.

[0042] The update unit is used to input the initial PID gain parameters and real-time feedback values ​​into the radial basis function network RBF, calculate the Jacobian matrix through the radial basis function, and update the network parameters online using the gradient descent method.

[0043] The fusion unit is used to weight and fuse the initial PID gain parameters output by the fuzzy inference module with the correction parameters output by the RBF through an adaptive parameter fusion mechanism to obtain the final PID parameters.

[0044] The drive unit is used to generate control signals based on the final PID parameters, and drive the actuator to adjust the port flow field to form a closed-loop control circuit.

[0045] In some alternative embodiments, the apparatus further includes:

[0046] The generation unit generates rules in the fuzzy rule base in the form of conditional statements and establishes an intelligent environmental perception system, thus forming the fuzzy rule base;

[0047] The adjustment unit is used to calculate the correction amounts of the proportional, integral, and derivative coefficients in real time based on the rules in the fuzzy rule base. , , The correction coefficient is dynamically adjusted using an exponential decay function, the mathematical expression of which is: ,in, , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. The elapsed time since the initial moment; the scaling factor. The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. The differential coefficients from the previous moment Compared with the current adjustment amount Determined jointly; among them, , , They represent in The initial settings of each parameter at any given time. , , It is a parameter adjustment amount determined based on a fuzzy rule base.

[0048] In some optional embodiments, the updating unit is further configured to:

[0049] The Gaussian function is chosen as the specific implementation of the radial basis function, and its mathematical expression is shown below:

[0050]

[0051] in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m; Calculate the neural network output based on the weight coefficients of the RBF neural network. and network recognition performance index function ,as follows:

[0052]

[0053]

[0054] in, This represents the measured values ​​of water quality indicators in the target water area. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron;

[0055] Iteratively update the node center vector Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

[0056] In some optional embodiments, the updating unit is further configured to:

[0057] During the network parameter update process, the following operations are performed sequentially at each time step t: based on the network learning rate Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change ;

[0058] Update weight parameters By using an adder Weighting coefficients at different times and Add them together and include them in the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference in time with respect to the weighting coefficient Update;

[0059] Based on network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector calculated by the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change ;

[0060] Update the base width vector At that time, The basis width vector at time t and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector using a multiplier exist Time and The difference at each time step is used as the updated weight coefficient.

[0061] According to a third aspect of this application, an electronic device is provided, comprising:

[0062] At least one processor; and

[0063] A memory communicatively connected to the at least one processor; wherein,

[0064] The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the steps of the adaptive fuzzy PID flow field control method described in this application.

[0065] According to a fourth aspect of this application, a non-transitory computer-readable storage medium is provided storing computer instructions for causing the computer to perform the steps of the adaptive fuzzy PID flow field control method described in this application.

[0066] The technical solution of this application improves the accuracy of the PID gain parameters and the control precision of the control signal by generating initial PID gain parameters in real time and introducing an RBF neural network to dynamically optimize and compensate the initial parameters. The weight coefficients corresponding to the final PID parameters are adjustable, thus balancing the contributions of fuzzy rules and neural network learning. Utilizing the nonlinear approximation capability of the RBF network, the network parameters are updated in real time using gradient descent, achieving accurate identification of the sensitivity of the control quantity to input changes. The parameter update rate is dynamically adjusted through momentum terms and exponential decay functions to form a closed-loop control circuit, ensuring that the flow field control always conforms to the dynamic PID parameters.

[0067] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0068] The above and other objects, features, and advantages of exemplary embodiments of this application will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of this application are illustrated in the drawings by way of example and not limitation, wherein:

[0069] In the accompanying drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

[0070] Figure 1 A flowchart illustrating the adaptive fuzzy PID flow field control method according to an embodiment of this application is shown;

[0071] Figure 2 This illustration shows a schematic diagram of dynamically adjusting the correction coefficient using an exponential decay function, according to an embodiment of this application.

[0072] Figure 3 A schematic diagram illustrating the online updating of network parameters using the gradient descent method in an embodiment of this application is shown;

[0073] Figure 4 This illustration shows a schematic diagram of the deviation and rate of change of water quality index data between real-time determination and target values ​​according to an embodiment of this application;

[0074] Figure 5 A schematic diagram of closed-loop control based on PID parameters according to an embodiment of this application is shown;

[0075] Figure 6 A schematic diagram of the composition structure of the adaptive fuzzy PID flow field control device according to an embodiment of this application is shown;

[0076] Figure 7 A schematic diagram of the composition structure of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0077] To make the objectives, features, and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0078] Figure 1 A flowchart illustrating the adaptive fuzzy PID flow field control method according to an embodiment of this application is shown, as follows: Figure 1 As shown, the adaptive fuzzy PID flow field control method of this application includes the following processing steps:

[0079] Step 101: Collect water quality index data in the target water area and determine the deviation and rate of change of the water quality index data from the target value.

[0080] In this embodiment, it is necessary to collect key water quality indicators such as inorganic nitrogen and reactive phosphate in the target port waters in real time. By calculating the deviation between the measured values ​​and the target values ​​and their rate of change, the processed data is input into a fuzzy inference module built on an intelligent environmental perception system. Step 102: The deviation and the rate of change of deviation are input into the fuzzy inference module, and fuzzification processing is performed through a seven-level fuzzy classification system to generate initial PID gain parameters based on a preset fuzzy rule base.

[0081] The seven-level fuzzy classification system in this application includes NB, NM, NS, ZO, PS, PM, PB, etc. It quantifies and evaluates the deviation between the measured value and the target value, and dynamically outputs the initial gain parameters of the PID controller according to the preset rule base. , , .

[0082] Step 103: Input the initial PID gain parameters and real-time feedback values ​​into the radial basis function network RBF, calculate the Jacobian matrix through the radial basis function, and update the network parameters online using the gradient descent method.

[0083] This application embodiment will use the initial control quantity The real-time feedback value is input into the RBF neural network. With the help of the nonlinear approximation capability and autonomous learning characteristics of the RBF neural network, the Jacobian matrix, which is the sensitivity information of the controlled system's control quantity to changes in the controller input, is output.

[0084] Step 104: Through the adaptive parameter fusion mechanism, the initial PID gain parameters output by the fuzzy inference module and the correction parameters output by the RBF are weighted and fused to obtain the final PID parameters.

[0085] This application embodiment optimizes and compensates PID parameters through a self-learning mechanism to generate a corrected gain. , , This enables dynamic adjustment of control parameters and continuous optimization of system performance.

[0086] Step 105: Generate a control signal based on the final PID parameters to drive the actuator to adjust the port flow field, thereby forming a closed-loop control circuit.

[0087] The following specific examples further illustrate the essence of the technical solutions in the embodiments of this application.

[0088] Figure 2 This illustration shows a schematic diagram of dynamically adjusting the correction coefficient using an exponential decay function, as described in an embodiment of this application. Figure 2 As shown, the PID flow field control in this embodiment uses an exponential decay function to dynamically adjust the correction coefficient, the mathematical expression of which is: ,in , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. This represents the elapsed time since the initial moment. Real-time data collection of key water quality indicators such as inorganic nitrogen and reactive phosphate in the target port area is performed. Water quality index set values Generate deviation value and deviation change value The fuzzy rule base used in this application embodiment is based on an intelligent environmental perception system, and its rules are in the form of conditional statements: If is…and is…, then is…, is…, The system calculates corrections for the proportional, integral, and derivative coefficients in real time based on these rules in the fuzzy rule base. , , And dynamically update the PID control parameters.

[0089] During parameter update, the scaling factor The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. Similarly, the differential coefficients from the previous moment... Compared with the current adjustment amount Determined jointly. Among them, , , These represent the controller in The initial settings of each parameter at any given time. , , This is the parameter adjustment amount calculated through a fuzzy rule base.

[0090] This application embodiment employs a three-layer feedforward RBF neural network, with the input layer of the RBF neural network... The hidden layer consists of signal source nodes that directly receive external input parameters. The hidden layer uses a nonlinear radial basis function with network weight coefficients. The input parameters undergo spatial transformation, and the output layer uses network weight coefficients. For the output of the hidden layer A linear weighted summation is performed, and finally, efficient function approximation is achieved by combining the nonlinear mapping of the hidden layer with the linear combination of the output layer.

[0091] The RBF neural network in this embodiment first performs nonlinear feature transformation on the input data. Its core mechanism is to achieve nonlinear mapping of the input data through radial basis functions. In the RBF neural network structure, each hidden layer node calculates the distance between the input vector and the center of the node, and uses this distance value as the node's output. During the training of the RBF network, three key parameters need to be solved: the center position of the radial basis function, the width parameter of the basis function, and the connection weights from the hidden layer to the output layer. This embodiment uses a Gaussian function as the specific implementation of the radial basis function, and its mathematical expression is as follows:

[0092] (1.1)

[0093] in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m. Based on the weight coefficients of the RBF neural network, the neural network output can be calculated. and network recognition performance index function ,as follows:

[0094] (1.2)

[0095] (1.3)

[0096] in, This represents the measured values ​​of water quality indicators. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron. To achieve optimal control, this embodiment uses gradient descent to update the parameters. Specifically, the performance index seeks the optimal solution along the negative gradient direction, and the node center vector is updated iteratively. Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

[0097] Figure 3 This illustration shows a schematic diagram of an embodiment of the present application using the gradient descent method to update network parameters online, as shown below. Figure 3 As shown, during the parameter update process, the following operations are performed sequentially at each time step t: First, based on the network learning rate... Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change Then update the weight parameters. By using an adder Weighting coefficients at different times and Add them together and introduce the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference in time, thereby enabling the adjustment of the weighting coefficients. Update.

[0098] Subsequently, based on the network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector obtained through the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change Update the base width vector. At that time, The basis width vector at time t and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector through a multiplier exist Time and The difference in time, thereby enabling the adjustment of the weighting coefficients. Update.

[0099] This application embodiment also updates the computational center vector. Each component Based on online learning efficiency Error difference Updated weight coefficient values Input vector With the center vector components The difference, and the magnitude of the updated base width vector The change in the central vector components is calculated using dividers and multipliers. In updating the center vector components At that time, the adder will be used to... The center vector component at time t and Add them together, and include the momentum term. Multiply by the central vector component through a multiplier exist Time and The time difference. Finally, the sum of all neurons' values ​​is accumulated using an adder. and , central vector components The control change output from the PID controller to the regulating valve The difference is the same The product of the ratios yields the Jacobian matrix. The entire calculation process is mediated by the momentum term. By continuously incorporating historical parameter change trends, the convergence characteristics of the parameters are optimized.

[0100] Figure 4 This illustration shows a schematic diagram illustrating the real-time determination of the deviation and rate of change of water quality index data from target values ​​according to an embodiment of this application. Figure 4 As shown, in this embodiment of the application, the deviation at the current moment is calculated based on the difference between the water quality indicators monitored in real time by the environmental data monitoring module and the set target water quality indicators. The deviation is input into the fuzzy control system, and after fuzzification processing and fuzzy inference, the initial stage fuzzy controller gain is output. , , Subsequently, the control quantity generated by the initial PID controller gain is... and actual feedback values A common input RBF neural network. Utilizing the RBF's ability to approximate nonlinear systems and its self-learning characteristics, this neural network can learn the dynamic mapping relationship between the system control quantity and the controller input, that is, the Jacobian matrix of the sensitivity of the controlled system's output control quantity to changes in the controller input.

[0101] Finally, the PID parameters of the input controller are optimized and fitted using a self-learning module. The gradient descent algorithm is used to adjust the neural network parameters, thereby obtaining the PID parameter increments output by the RBF neural network. , , As shown in Equation 1.4 below, where , , , representing the PID gain learning rate and , respectively. Therefore, by combining the weights of the fuzzy controller and the RBF neural network, the adjustment increment of the PID control parameters can be calculated as follows: , , .in, , and The weighting coefficients reflect the contribution weights of fuzzy control and neural networks to the parameter increments. In this way, the system can adaptively adjust the PID parameter gain in real time, effectively suppressing interference from environmental disturbances and forming a closed-loop control circuit. This strategy significantly improves the system's control accuracy and dynamic response performance. The system's control signal is given by Equation 1.5.

[0102] (1.4)

[0103] (1.5)

[0104] As shown in Equation 1.5 above, in the control system, the output value of the PID controller is ,in Represents a discrete point in time. Control signal. The calculation process involves the following parameters: time point controller output value Time point Deviation change and time points The error of the second-order backward difference The second derivative used for approximation error, time point Deviation change The calculation of the PID controller output value During the process, the time points in the historical deviation data Deviation change and time point Deviation change It is used to predict the changing trends of water quality indicators, and the rule weights are dynamically optimized through a sliding window algorithm.

[0105] Figure 5 This paper illustrates an overall schematic diagram of closed-loop control based on PID parameters according to an embodiment of this application, as shown below. Figure 5 As shown, this embodiment first collects environmental parameters of the port waters using environmental data sensors. These parameters, along with real-time flow and velocity data obtained from flow and velocity sensors, are then input to the core control component—a neural network-based adaptive fuzzy control PID controller. This intelligent controller dynamically generates precise control signals and outputs them to the regulating valve by integrating the self-learning capability of neural networks and the adaptability of fuzzy control. Closed-loop control of the flow is achieved by changing the valve opening. The flow and velocity sensors continuously provide the latest data under the influence of the controlled flow, forming a two-way data interaction with the controller, thus constituting a complete closed-loop control system with environmental parameter perception, intelligent algorithm decision-making, and real-time feedback adjustment.

[0106] Figure 6 A schematic diagram of the composition structure of the adaptive fuzzy PID flow field control device according to an embodiment of this application is shown, as follows: Figure 6 As shown, the adaptive fuzzy PID flow field control device according to an embodiment of this application includes:

[0107] The calculation unit 60 is used to collect water quality index data in the target water area and determine the deviation and the rate of change of the deviation between the water quality index data and the target value; wherein, the water quality index data includes at least inorganic nitrogen concentration and reactive phosphate concentration; the fuzzy processing unit 61 is used to input the deviation and the rate of change of the deviation into the fuzzy inference module, perform fuzzification processing through a seven-level fuzzy classification system, and generate initial PID gain parameters based on a preset fuzzy rule base.

[0108] The update unit 62 is used to input the initial PID gain parameters and real-time feedback values ​​into the radial basis function network RBF, calculate the Jacobian matrix through the radial basis function, and update the network parameters online using the gradient descent method.

[0109] The fusion unit 63 is used to weight and fuse the initial PID gain parameters output by the fuzzy inference module with the correction parameters output by the RBF through an adaptive parameter fusion mechanism to obtain the final PID parameters.

[0110] The drive unit 64 is used to generate control signals based on the final PID parameters and drive the actuator to adjust the port flow field to form a closed-loop control loop.

[0111] middle Figure 6 Based on the adaptive fuzzy PID flow field control device shown, the adaptive fuzzy PID flow field control device of this application embodiment further includes:

[0112] Generating unit ( Figure 6 (Not shown in the image) The rules in the fuzzy rule base are generated in the form of conditional statements, and an intelligent environmental perception system is established to form the fuzzy rule base;

[0113] Adjustment unit ( Figure 6 (Not shown in the image) is used to calculate the correction amounts of the proportional, integral, and differential coefficients in real time based on the rules in the fuzzy rule base. , , The correction coefficient is dynamically adjusted using an exponential decay function, the mathematical expression of which is: ,in, , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. The elapsed time since the initial moment; the scaling factor. The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. The differential coefficients from the previous moment Compared with the current adjustment amount Determined jointly; among them, , , They represent in The initial settings of each parameter at any given time. , , It is a parameter adjustment amount determined based on a fuzzy rule base.

[0114] In some optional embodiments, the updating unit 62 is further configured to:

[0115] The Gaussian function is chosen as the specific implementation of the radial basis function, and its mathematical expression is shown below:

[0116]

[0117] in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m; Calculate the neural network output based on the weight coefficients of the RBF neural network. and network recognition performance index function ,as follows:

[0118]

[0119]

[0120] in, This represents the measured values ​​of water quality indicators in the target water area. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron;

[0121] Iteratively update the node center vector Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

[0122] In some optional embodiments, the updating unit 62 is further configured to:

[0123] During the network parameter update process, the following operations are performed sequentially at each time step t: based on the network learning rate Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change ;

[0124] Update weight parameters By using an adder Weighting coefficients at different times and Add them together and include them in the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference in time with respect to the weighting coefficient Update;

[0125] Based on network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector calculated by the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change ;

[0126] Update the base width vector At that time, The basis width vector at time t and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector using a multiplier exist Time and The difference at each time step is used as the updated weight coefficient.

[0127] Update the computation center vector Each component Based on network learning efficiency Error difference Updated weight coefficient values Input vector With the center vector components The difference, and the magnitude of the updated base width vector The change in the central vector components is calculated using dividers and multipliers. ;

[0128] Update center vector components At that time, the adder will be used to... The center vector component at time t and Add them together, and include the momentum term. Multiply by the central vector component exist Time and The difference in time;

[0129] Accumulate all neurons using an adder. and , central vector components The control change output from the PID controller to the regulating valve The difference is the same The product of the ratios yields the Jacobian matrix. .

[0130] PID parameter increments output by RBF neural network , , Determined by the following formula:

[0131]

[0132] in, , , These represent the PID gain and learning rate, respectively.

[0133] The process of generating control signals based on the final PID parameters includes:

[0134] The control signal is determined by the following formula:

[0135]

[0136] Among them, control signals The calculation involves the following parameters: time point controller output value Time point Deviation change and time points The error of the second-order backward difference The second derivative used for approximation error, time point Deviation change ;

[0137] Calculating the output value of the PID controller During the process, the time points in the historical deviation data Deviation change and time point Deviation change Used for predicting trends in water quality indicators.

[0138] In an exemplary embodiment, each processing unit in the adaptive fuzzy PID flow field control device of this application embodiment may be implemented by one or more central processing units (CPUs), graphics processing units (GPUs), application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components.

[0139] Regarding the apparatus in the above embodiments, the specific manner in which each module and unit performs its operations has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0140] According to embodiments of this application, this application also describes an electronic device and a readable storage medium.

[0141] Figure 7 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of this application is shown. Electronic device 800 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic device 800 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0142] like Figure 7As shown, the electronic device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 into a random access memory (RAM) 803. The RAM 803 may also store various programs and data required for the operation of the electronic device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0143] Multiple components in electronic device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of displays, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows electronic device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0144] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the adaptive fuzzy PID flow field control method. For example, in some embodiments, the adaptive fuzzy PID flow field control method can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the adaptive fuzzy PID flow field control method described above can be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform an adaptive fuzzy PID flow field control method by any other suitable means (e.g., by means of firmware).

[0145] Various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0146] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0147] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or electronic device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or electronic devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0148] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0149] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0150] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0151] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0152] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "a plurality of" means two or more, unless otherwise explicitly specified.

[0153] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive fuzzy PID flow field control method, characterized in that, The method includes: Collect water quality index data in the target water area, and determine the deviation and the rate of change of the deviation between the water quality index data and the target value; the water quality index data includes at least inorganic nitrogen concentration and reactive phosphate concentration; input the deviation and the rate of change of the deviation into the fuzzy inference module, perform fuzzification processing through a seven-level fuzzy classification system, and generate initial PID gain parameters based on a preset fuzzy rule base; The initial PID gain parameters and real-time feedback values ​​are input into the radial basis function network (RBF). The Jacobian matrix is ​​calculated using the radial basis function, and the network parameters are updated online using the gradient descent method. The initial PID gain parameters output by the fuzzy inference module and the correction parameters output by the RBF are weighted and fused through an adaptive parameter fusion mechanism to obtain the final PID parameters. The control signal is generated based on the final PID parameters to drive the actuator to adjust the port flow field, thereby forming a closed-loop control circuit. The step of calculating the Jacobian matrix using radial basis functions and updating network parameters online using gradient descent includes: The Gaussian function is chosen as the specific implementation of the radial basis function, and its mathematical expression is shown below: in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m; This is the input layer of the RBF neural network. ; Calculate the neural network output based on the weight coefficients of the RBF neural network. and network recognition performance index function ,as follows: in, This represents the measured values ​​of water quality indicators in the target water area. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron; Iteratively update the node center vector Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

2. The control method according to claim 1, characterized in that, The method further includes: The rules in the fuzzy rule base are generated in the form of conditional statements, and an intelligent environmental perception system is established to form the fuzzy rule base; The correction amounts for the proportional, integral, and differential coefficients are calculated in real time based on the rules in the fuzzy rule base. , , The correction coefficient is dynamically adjusted using an exponential decay function, the mathematical expression of which is: ,in, , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. This represents the elapsed time since the initial moment. proportionality coefficient The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. The differential coefficients from the previous moment Compared with the current adjustment amount Determined jointly; among them, , , They represent in The initial settings of each parameter at any given time. , , It is a parameter adjustment amount determined based on a fuzzy rule base.

3. The control method according to claim 2, characterized in that, The method further includes: During the network parameter update process, the following operations are performed sequentially at each time step t: based on the network learning rate Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change ; Update weight parameters By using an adder Weighting coefficients at different times and Add them together and include them in the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference at time points affects the weighting coefficients Update; Based on network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector calculated by the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change ; Update the base width vector At that time, The basis width vector at time step and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector using a multiplier exist Time and The difference at each time step is used as the updated weight coefficient.

4. The control method according to claim 3, characterized in that, The method further includes: Update the computation center vector Each component Based on network learning efficiency Error difference Updated weight coefficient values Input vector With the center vector components The difference, and the magnitude of the updated base width vector The change in the central vector components is calculated using dividers and multipliers. ; Update center vector components At that time, the adder will be used to... The center vector component at time t. and Add them together, and include the momentum term. Multiply by the central vector component exist Time and The difference in time; Accumulate all neurons using an adder. and , central vector components The control change output from the PID controller to the regulating valve The difference is the same The product of the ratios yields the Jacobian matrix. .

5. The control method according to claim 4, characterized in that, The process of generating control signals based on the final PID parameters includes: PID parameter increments output by RBF neural network , , Determined by the following formula: in, , , These represent the PID gain and learning rate, respectively. The process of generating control signals based on the final PID parameters includes: The control signal is determined by the following formula: Among them, control signals The calculation involves the following parameters: time point controller output value Time point Deviation change and time points The error of the second-order backward difference The second derivative used for approximation error, time point Deviation change ; Calculating the output value of the PID controller During the process, the time points in the historical deviation data Deviation change and time point Deviation change Used for predicting trends in water quality indicators.

6. An adaptive fuzzy PID flow field control device, characterized in that, The device includes: The calculation unit is used to collect water quality index data in the target water area and determine the deviation and the rate of change of the deviation between the water quality index data and the target value; wherein, the water quality index data includes at least inorganic nitrogen concentration and reactive phosphate concentration; the fuzzy processing unit is used to input the deviation and the rate of change of the deviation into the fuzzy inference module, perform fuzzification processing through a seven-level fuzzy classification system, and generate initial PID gain parameters based on a preset fuzzy rule base. The update unit is used to input the initial PID gain parameters and real-time feedback values ​​into the radial basis function network RBF, calculate the Jacobian matrix through the radial basis function, and update the network parameters online using the gradient descent method. The fusion unit is used to weight and fuse the initial PID gain parameters output by the fuzzy inference module with the correction parameters output by the RBF through an adaptive parameter fusion mechanism to obtain the final PID parameters. The drive unit is used to generate control signals based on the final PID parameters, and drive the actuator to adjust the port flow field to form a closed-loop control loop. The update unit is further configured to: The Gaussian function is chosen as the specific implementation of the radial basis function, and its mathematical expression is shown below: in, For the first The base width vector of each node. Indicates the first The center vector of each node =1,2,…m; This is the input layer of the RBF neural network. ; Calculate the neural network output based on the weight coefficients of the RBF neural network. and network recognition performance index function ,as follows: in, This represents the measured values ​​of water quality indicators in the target water area. These are the network weight coefficients from the hidden layer to the output layer. It is the first hidden layer The output of each neuron; Iteratively update the node center vector Base width vector Network weight coefficients And Jacobi matrix Identify sensitivity information, update network identification metrics, and obtain optimal values ​​for network identification metrics.

7. The control device according to claim 6, characterized in that, The device further includes: The generation unit generates rules in the fuzzy rule base in the form of conditional statements and establishes an intelligent environmental perception system, thus forming the fuzzy rule base; The adjustment unit is used to calculate the correction amounts of the proportional, integral, and derivative coefficients in real time based on the rules in the fuzzy rule base. , , The correction coefficient is dynamically adjusted using an exponential decay function, the mathematical expression of which is: ,in, , , These represent the correction coefficients for the proportional, integral, and differential terms, respectively. The value of the exponential decay function. These are the initial values ​​for each correction factor. Represents the decay rate constant. The base of the natural logarithm. The elapsed time since the initial moment; the scaling factor. The value is determined by the coefficient at the previous moment. Compared with the current adjustment amount The integral coefficients are determined jointly. The value is derived from the integral coefficient of the previous time step. Compared with the current adjustment amount The differential coefficient values ​​are determined jointly. The differential coefficients from the previous moment Compared with the current adjustment amount Determined jointly; among them, , , They represent in The initial settings of each parameter at any given time. , , It is a parameter adjustment amount determined based on a fuzzy rule base.

8. The control device according to claim 7, characterized in that, The update unit is further configured to: During the network parameter update process, the following operations are performed sequentially at each time step t: based on the network learning rate Actual system output With model output The difference between them, and the first Output value of each neuron Calculate the weight parameters Change ; Update weight parameters By using an adder Weighting coefficients at different times and Add them together and include them in the momentum term. Multiply by the weighting coefficient using a multiplier exist Time and The difference at time points affects the weighting coefficients Update; Based on network learning rate Error difference , No. Output value of each neuron Updated weighting coefficients The input vector calculated by the dot product With the center vector square Euclidean distance and the magnitude of the base width vector Calculate the base width vector Change ; Update the base width vector At that time, The basis width vector at time step and The addition is performed using an adder, and a momentum term is introduced. Multiply by the base width vector using a multiplier exist Time and The difference at each time step is used as the updated weight coefficient.

Images