Intelligent control method based on model prediction

Abstract

The invention discloses an intelligent control method based on model prediction, and belongs to the technical field of intelligent control, and the method comprises the steps: collecting the real-time operation data of a controlled system; wherein the real-time operation data comprises a state variable, an input variable and an output variable; constructing a dynamic model of the controlled system based on the real-time operation data; based on the constructed dynamic model, predicting the output of the controlled system in a preset time period to obtain a prediction result sequence; obtaining a target optimization function according to the prediction result and a set control target, and solving an optimal control sequence through an optimization algorithm; and acquiring a first control quantity based on the optimal control sequence, acting the first control quantity on the controlled system, and repeating the steps in the next acquisition period to realize closed-loop control.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and particularly relates to an intelligent control method based on model prediction. Background Art

[0002] In the current field of intelligent control, traditional control methods often struggle to cope with complex and ever-changing system environments and dynamic characteristics. Traditional control methods, such as proportional-integral-differential (PID) control, although to a certain extent can meet the control requirements of some simple systems, but for these complex systems, their control effects are often not satisfactory; Furthermore, there are still some problems and challenges in the existing model-predictive control methods in practical applications. For example, how to accurately collect and process the real-time operation data of the controlled system to improve the accuracy of the dynamic model; how to select appropriate prediction models and optimization algorithms to meet the control requirements in different application scenarios; and how to improve the computational efficiency and real-time performance of the algorithm while ensuring the control effect.

[0003] Therefore, further research on an intelligent control method based on model prediction is of great significance for promoting the development and application of intelligent control technology. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an intelligent control method based on model prediction to solve the problems raised in the above background art.

[0005] The purpose of the present invention can be achieved by the following technical solutions: An intelligent control method based on model prediction, comprising: Step 1, collect the real-time operation data of the controlled system; wherein, the real-time operation data includes state variables, input variables, and output variables; Connect to the controlled system through a sensor network, obtain the collected data according to the set sampling period, and filter and preprocess the collected data to remove noise and outliers to obtain the real-time operation data; Step 2, construct a dynamic model of the controlled system based on the real-time operation data; Step 3, predict the output of the controlled system within a preset time period based on the constructed dynamic model to obtain a sequence of prediction results; Step 4, obtain an objective optimization function according to the prediction results and the set control target, and solve the optimal control sequence through an optimization algorithm; Step 5, obtain the first control quantity based on the optimal control sequence, apply it to the controlled system, and repeat the above steps in the next acquisition cycle to achieve closed-loop control.

[0006] Preferably, in step 1, the acquired data is filtered and preprocessed, including: The acquired data is filtered using a low-pass filter to remove high-frequency interference; Based on the threshold judgment method, the outliers in the acquired data are removed; according to the actual situation and historical data of the controlled system, a threshold range is set for each variable; if the data at a certain sampling moment exceeds the corresponding threshold range, it is identified as an outlier, and the median within the neighborhood around the current data point is used to replace the outlier.

[0007] Preferably, in step 2, the dynamic model has the following expression: ; In the formula, x(k) is the state variable vector of the controlled system at the k-th sampling moment; u(k) is the input variable vector of the controlled system at the k-th sampling moment; y(k) is the output variable vector of the controlled system at the k-th sampling moment; A, B, C, and D are the state transition matrix, input matrix, output matrix, and direct transmission matrix respectively; Among them, the state variable vector is denoted as , the input variable vector is denoted as , and the output variable vector is denoted as ; T represents transpose; n, m, and p are the dimensions of the state variable, input variable, and output variable respectively.

[0008] Preferably, in step 3, the specific steps for prediction based on the constructed dynamic model are as follows: S301. Set the prediction horizon and control horizon; according to the characteristics and control requirements of the controlled system, determine the lengths of the prediction horizon and control horizon; S302. Obtain the initial state and input sequence collected at the current moment of the controlled system, and input them into the constructed dynamic model, and gradually calculate the predicted state and output prediction value of the controlled system at each sampling moment within the prediction horizon; S303. Organize the output prediction values of the controlled system at each sampling moment within the prediction horizon to obtain a prediction result sequence; S304. By comparing the output variable sequence collected through actual measurement with the prediction result sequence, obtain a prediction error sequence; calculate the prediction accuracy of the model based on the prediction error sequence to evaluate whether the model prediction performance reaches the preset standard.

[0009] Preferably, in S304, set the error range, count the number of output prediction values exceeding the error range, and calculate the prediction accuracy; compare the prediction accuracy with the preset standard. If the prediction accuracy ≥ the preset standard, it means that the model prediction performance meets the standard; otherwise, correct the elements in the A, B, C, and D matrices of the model to improve the prediction accuracy until it reaches the preset standard.

[0010] Preferably, the method for solving the optimal control sequence is as follows: Set the control objective to minimize the tracking error between the output set value and the output expected value; wherein, the output set value is set according to the output variable index of the controlled system; The calculation formula for the said tracking error is: ; In the formula, E is the tracking error; Yd is the output set value; Y is the output predicted value.

[0011] Preferably, according to the importance of the tracking error at different times, introduce a weight matrix, and calculate the objective optimization function that minimizes the sum of squares of the tracking error within the entire prediction time domain; The calculation formula is as follows: ; In the formula, J is the value of the objective optimization function; T is the transpose; Q is the weight matrix.

[0012] Preferably, find the input variable sequence that minimizes the objective optimization function through an optimization algorithm to obtain the optimal control sequence; wherein, the control sequence refers to the input control variable at each sampling moment within the prediction time domain.

[0013] Compared with the existing solutions, the beneficial effects achieved by the present invention: By collecting real-time operation data such as the state variables, input variables, and output variables of the controlled system, the present invention comprehensively grasps the system operation state, provides rich and accurate basic information for subsequent model construction and control strategy formulation, enables control decisions to be based on the actual situation of the system, and avoids control deviations caused by information loss; By constructing a dynamic model of the controlled system based on real-time operation data, the present invention can accurately reflect the dynamic characteristics of the system. This model fully considers the real-time changes of the system and can better conform to the operation law of the actual system compared with the traditional fixed model, providing strong model support for precise control; By using the constructed dynamic model to predict the output of the controlled system within a preset time period to obtain a prediction result sequence, this forward-looking analysis of the future behavior of the system enables the control strategy to consider the change trend of the system in advance, enhancing the adaptability and stability of the system; By obtaining the objective optimization function according to the prediction result and the set control objective, and solving the optimal control sequence through an optimization algorithm, this optimization-based control strategy can maximize the realization of the control objective on the premise of meeting the system constraint conditions, improve the control accuracy and efficiency, and can more effectively optimize the system performance compared with the traditional control method; By implementing closed-loop control, the present invention can adjust the control strategy according to the feedback of the system, ensuring that the system always operates towards the set goal, improving the robustness and reliability of the system. Even when the system parameters change or are subject to external interference, it can quickly adjust the control quantity to maintain the stable operation of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The present invention will be further described below with reference to the accompanying drawings.

[0015] Figure 1 It is a flow block diagram of an intelligent control method based on model prediction proposed by the present invention. SPECIFIC EMBODIMENTS

[0016] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0017] Please refer to Figure 1 , the present invention is an intelligent control method based on model prediction, including the following steps: Step 1: Collect the real-time operation data of the controlled system; among them, the real-time operation data includes state variables, input variables, and output variables; In step 1, it is connected to the controlled system through a sensor network, acquires the collected data according to the set sampling period, and filters and preprocesses the collected data to remove noise and outliers to obtain the real-time operation data; It should be noted that the state variable is the variable representing the state of the controlled system at a certain moment; the input variable is the control quantity of various external inputs acting on the controlled system; the output variable is the result generated by the controlled system under the action of the input. Filtering and preprocessing the collected data includes: Using a low-pass filter to filter the collected data to remove high-frequency interference; Based on the threshold judgment method, removing the outliers in the collected data; according to the actual situation and historical data of the controlled system, setting a threshold range for each variable; if the data at a certain sampling moment exceeds the corresponding threshold range, it is determined as an outlier, and the median within the neighborhood around the current data point is used to replace the outlier; In this implementation step, by filtering and preprocessing the collected data, the quality of the collected data is improved, which is beneficial to improving the response speed of the controlled system; Step 2: Construct a dynamic model of the controlled system based on the real-time operation data; In step 2, the expression of the dynamic model is: ; Wherein, x(k) is the state variable vector of the controlled system at the k-th sampling moment; u(k) is the input variable vector of the controlled system at the k-th sampling moment; y(k) is the output variable vector of the controlled system at the k-th sampling moment; A, B, C, and D are the state transition matrix, input matrix, output matrix, and direct transmission matrix respectively; Among them, the state variable vector is denoted as , the input variable vector is denoted as , and the output variable vector is denoted as ; T represents transpose; n, m, and p are the dimensions of the state variable, input variable, and output variable respectively; the elements of matrices A, B, C, and D are determined through system parameter identification, theoretical derivation, or experimental data; In the implementation steps of the present invention, the dynamic model can reflect the relationship among the state, input, and output of the controlled system under different working conditions; Step 3: Based on the constructed dynamic model, predict the output of the controlled system within a preset time period to obtain a prediction result sequence. The specific steps are as follows: S301: Set the prediction horizon and control horizon; determine the lengths of the prediction horizon and control horizon according to the characteristics and control requirements of the controlled system; Among them, the prediction horizon refers to the prediction range of the output variables of the controlled system in the future for a period of time, and the control horizon refers to the time range for implementing the control action starting from the current moment; S302: Obtain the initial state and input sequence collected at the current moment of the controlled system, input them into the constructed dynamic model, and gradually calculate the predicted state and output prediction value of the controlled system at each sampling moment within the prediction horizon; Among them, within each sampling moment in the prediction horizon, the input variable sequences u(k), u(k + 1),..., u(k + N) are known control quantities; N is the length of the prediction horizon; S303: Organize the output prediction values of the controlled system at each sampling moment within the prediction horizon to obtain a prediction result sequence; S304: Compare the output variable sequence collected by actual measurement with the prediction result sequence to obtain a prediction error sequence; calculate the prediction accuracy rate of the model according to the prediction error sequence to evaluate whether the model prediction performance reaches the preset standard; Set an error range, count the number of output prediction values exceeding the error range, and calculate the prediction accuracy rate; compare the prediction accuracy rate with the preset standard. If the prediction accuracy rate ≥ preset standard, it means that the model prediction performance meets the standard; otherwise, correct the elements in matrices A, B, C, and D of the model to improve the prediction accuracy rate until it reaches the preset standard; Among them, the preset standard can be set to 85%; the error range is set according to the requirements for the output variables of the controlled system in the actual application scenario; Step 4: According to the prediction result and the set control objective, obtain the target optimization function, and solve the optimal control sequence through an optimization algorithm; The method for solving the optimal control sequence is as follows: Set the control objective to minimize the tracking error between the output set value and the output expected value; among them, the output set value is set according to the output variable index of the controlled system; The calculation formula for the said tracking error is: ; In the formula, E is the tracking error; Yd is the output set value; Y is the output predicted value; Among them, the output set value is ; The output predicted value is ; According to the importance of the tracking error at different times, introduce a weight matrix, and calculate the target optimization function that minimizes the sum of squares of the tracking error within the entire prediction time domain; The calculation formula is as follows: ; In the formula, J is the value of the target optimization function; T is the transpose; Q is the weight matrix; It should be noted that in actual applications, the selection of the weight matrix depends on the characteristics of the specific problem and the experience of the designer, and is determined by observing the changes in system performance through experimenting with different weight configurations; Find the input variable sequence that minimizes the target optimization function through an optimization algorithm to obtain the optimal control sequence; among them, the control sequence refers to the input control variables at each sampling moment within the prediction time domain; the optimization algorithms include but are not limited to particle swarm optimization algorithm, gradient descent algorithm, quadratic programming algorithm, genetic algorithm; Exemplarily, the said particle swarm optimization algorithm finds the optimal solution by simulating the cooperation and competition behaviors of particles in the search space; takes the input variable sequence as the position vector of the particle, and the target optimization function as the fitness function, and the particle updates its flight speed and position according to its own optimal position and the optimal position of the group, and finds the control sequence that minimizes the target optimization function after multiple iterations to obtain the optimal control sequence; the specific calculation process is not elaborated here; In the embodiment of the present invention, by introducing a weight matrix, the importance of the tracking error at different times in the objective function is adjusted according to different weight coefficients, so as to achieve more flexible and targeted optimal control.

[0018] Step 5: Obtain the first control quantity based on the optimal control sequence, apply it to the controlled system, and repeat the above steps in the next acquisition cycle to achieve closed-loop control; In the implementation steps of the present invention, a complete closed-loop control process is formed through the above steps. Each step is based on the result of the previous step, and through continuous iteration and optimization, precise control of the controlled system is achieved.

[0019] In several embodiments provided by the present invention, it should be understood that the disclosed system can be implemented in other ways. For example, the described invention embodiments are merely illustrative. For example, the division of modules is only a logical function division, and there can be other division methods in actual implementation.

[0020] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0021] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a combination of hardware and software functional modules.

[0022] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0023] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent control method based on model prediction, characterized in that: include: Step 1: Collect real-time operation data of the controlled system; wherein the real-time operation data includes state variables, input variables, and output variables; Connecting to the controlled system through the sensor network, acquiring data according to the set sampling period, filtering and preprocessing the collected data, removing noise and abnormal values, and obtaining real-time operation data; Step 2: Build a dynamic model of the controlled system based on real-time operation data; Step 3: Based on the constructed dynamic model, the output of the controlled system within a preset time period is predicted to obtain a prediction result sequence; Step 4: According to the prediction results and the set control objectives, the target optimization function is obtained, and the optimal control sequence is solved by the optimization algorithm; Step 5: Obtain the first control variable based on the optimal control sequence, apply it to the controlled system, and repeat the above steps in the next acquisition cycle to achieve closed-loop control.

2. The intelligent control method based on model prediction according to claim 1, characterized in that: In step 1, the collected data is filtered and preprocessed, including: Use a low-pass filter to filter the collected data to remove high-frequency interference; Based on the threshold judgment method, outliers in the collected data are removed; according to the actual situation and historical data of the controlled system, a threshold range is set for each variable; if the data at a certain sampling moment exceeds the corresponding threshold range, it is identified as an outlier, and the median of the neighborhood around the current data point is used to replace the outlier.

3. The intelligent control method based on model prediction according to claim 2, characterized in that: In step 2, the dynamic model is expressed as: ； In the formula, x(k) is the state variable vector of the controlled system at the kth sampling time; u(k) is the input variable vector of the controlled system at the kth sampling time; y(k) is the output variable vector of the controlled system at the kth sampling time; A, B, C, and D are the state transfer matrix, input matrix, output matrix, and direct transmission matrix, respectively; The state variable vector is recorded as , the input variable vector is recorded as , the output variable vector is recorded as ; T represents transpose; n, m, and p are the dimensions of state variables, input variables, and output variables, respectively.

4. The intelligent control method based on model prediction according to claim 3, characterized in that: In step 3, the specific steps for prediction based on the constructed dynamic model are as follows: S301, setting the prediction time domain and the control time domain; determining the length of the prediction time domain and the length of the control time domain according to the characteristics and control requirements of the controlled system; S302, obtaining the initial state and input sequence collected by the controlled system at the current moment, inputting them into the constructed dynamic model, and gradually calculating the predicted state and output predicted value of the controlled system at each sampling moment in the prediction time domain; S303, sorting the output prediction values ​​of the controlled system at each sampling time in the prediction time domain to obtain a prediction result sequence; S304. Obtain a prediction error sequence by comparing the output variable sequence actually measured and collected with the prediction result sequence; calculate the prediction accuracy of the model according to the prediction error sequence to evaluate whether the prediction performance of the model meets the preset standard.

5. The intelligent control method based on model prediction according to claim 4, characterized in that: In S304, an error range is set, the number of output prediction values ​​that exceed the error range is counted, and the prediction accuracy is calculated; the prediction accuracy is compared with the preset standard. If the prediction accuracy ≥ the preset standard, it means that the model prediction performance meets the standard; otherwise, the prediction accuracy is improved by correcting the elements in the model A, B, C and D matrices until it meets the preset standard.

6. The intelligent control method based on model prediction according to claim 5, characterized in that: The method to solve the optimal control sequence is as follows: The control objective is set to minimize the tracking error between the output set value and the output expected value; wherein the output set value is set according to the output variable index of the controlled system; The calculation formula of the tracking error is: ； Where E is the tracking error; Yd is the output set value; Y is the output predicted value.

7. The intelligent control method based on model prediction according to claim 6, characterized in that: According to the importance of tracking errors at different moments, a weight matrix is ​​introduced to calculate the target optimization function that minimizes the sum of squares of tracking errors in the entire prediction time domain. The calculation formula is as follows: ； Where J is the target optimization function value; T is the transpose; Q is the weight matrix.

8. The intelligent control method based on model prediction according to claim 7, characterized in that: The optimal control sequence is obtained by finding the input variable sequence that minimizes the target optimization function through the optimization algorithm; wherein the control sequence refers to the input control variable at each sampling moment in the prediction time domain.

Images