46results about How to "Overcoming the time-consuming and labor-intensive adjustment problems" patented technology

The invention discloses a parameter self-tuning method of a MISO tight format model-free controller based on system errors. The method includes: a system error set is adopted as input of a BP neural network, the BP neural network performs forward calculation and outputs to-be-tuned parameters of the MISO tight format model-free controller, such as a penalty factor and a step factor, through an output layer, a control input vector of a controlled object is obtained through calculation by employing a control algorithm of the MISO tight format model-free controller, the minimization of a value ofa system error function is regarded as the target, reverse propagation calculation of the system errors is performed aiming at a gradient information set of each to-be-tuned parameter with the combination of control input by employing a gradient descent method, weight coefficients of a hidden layer and the output layer of the BP neural network are updated in real time in an online manner, and parameter self-tuning of the controller based on the system errors is realized. According to the parameter self-tuning method of the MISO tight format model-free controller based on the system errors, the online tuning problem of the parameters of the controller can be effectively overcome, and a good control effect of the MISO system is achieved.

The invention discloses a parameter self-tuning method of an SISO full-format model free controller based on partial derivative information. The partial derivative information is used as an input of aBP neural network, the BP neural network performs forward calculation and outputs a penalty factor, a step length factor and other parameters to be set of a controller through an output layer, a control algorithm of the controller is used to calculate and obtain control input for a controlled object, the gradient information of the control input for each parameter to be set is calculated, with aminimum value of a system error function as a target, a gradient descent method is used, combined with gradient information, the system error backpropagation calculation is carried out, a hidden layerweight coefficient and an output layer weight coefficient of the BP neural network are updated in real time in an online mode, and the gradient information is stored as the partial derivative information to be an input of the BP neural network of a next time. The invention provides the parameter self-tuning method of an SISO full-format model free controller based on partial derivative information, the tuning problem of the controller parameters can be effectively overcome, and a good control effect is achieved.

The invention discloses a partial deviation information based parameter self-setting method of an MIMO tight-format model-free controller. Partial deviation information serves as input of a BP neuralnetwork, the BP neural network carries out forward calculation and outputs to-be-set parameters, including an output-layer output punishing factor and a step factor, of the MIMO tight-format model-free controller, a control algorithm of the MIMO tight-format model-free controller is used to calculate a control input vector aimed at a controlled object, reverse spreading of system errors is calculated by taking minimizing a value of a system error function as the target, employing a gradient decrease method and controlling input aimed at a gradient information set of parameters to be set, a hidden-layer weight coefficient and an output-layer weight coefficient of the BP neural network are updated online in real time, and parameters of the controller are self-set on the basis of the partialdeviation information. The method can be used to overcome difficulty in online parameter setting of the controller, and has a good control effect for the MIMO system.

The invention discloses a parameter self-tuning MISO (Multiple Input and Single Output) different-factor tight-format model-free control method. In view of the limitation of the existing MISO tight-format model-free control method adopting the same factor structure, that is, the limitation of only adopting penalty factors with the same value and step-size factors with the same value for differentcontrol inputs in a control input vector at a k moment, an MISO tight-format model-free control method by adopting a different factor structure is provided, penalty factors with different values and / or step-size factors with different values can be adopted for different control inputs in the control input vector at the k moment, and the control problem of different characteristics of control channels existing in a complex object such as a strong nonlinear MISO system can be solved. A parameter self-tuning method is also provided to effectively overcome the problem that penalty factors and step-size factors need to consume time and labor for tuning. Compared with the existing control method, the method of the invention is higher in control accuracy, better in stability and wider in applicability.

The invention discloses a parameter self-setting method of a MISO full format model-free controller based on deviation information. The parameter self-setting method includes the steps of using deviation information as the input of a BP neural network, the BP neural network performing forward calculation and outputting a penalty factor, a step factor and other to-be-set parameters of an MISO fullformat no-model controller through an output layer, calculating to obtain a control input vector for a controlled object by adopting a control algorithm of the MISO full format no-model controller, conducting system error reverse propagation calculation for gradient information sets of all the to-be-set parameters by using a gradient descent method in combination with control output with minimization of the value of a system error function as a target, on-line updating the hidden layer weight coefficient of the BP neural network in real time, outputting the layer weight coefficient, and realizing the self-setting of parameters based on the deviation information. The parameter self-setting method can effectively overcome the online setting difficulty of the parameters of a controller, and agood control effect on an MISO system is achieved.

The invention discloses a parameter self-tuning method of a SISO compact format model-free controller based on integrated learning. The integrated learning algorithm includes three individual algorithms of PSO algorithm, BP neural network and cyclic neural network. Taking the system error as the input of the integrated learning algorithm, three individual algorithms are firstly used to tune the parameters of the SISO compact model-free controller online and output three sets of temporary tuning parameters, and input the results into the controller to calculate the control of the controlled object. Input, calculate three groups of temporary system errors and use the softmax function to calculate the weight ratio of individual algorithms, weight the weight ratio and the temporary tuning parameters to be summed as the final SISO compact model-free controller parameters to be tuned, and realize parameter self-tuning. The SISO compact format model-free controller proposed by the present invention is based on the parameter self-tuning method of integrated learning, combines the advantages of different individual algorithms, enhances the generalization of the algorithm, overcomes the problem of online tuning of controller parameters, and has a good control effect on the SISO system.

The invention discloses a parameter self-setting method of a MISO tight format model-free controller based on deviation information. The parameter self-setting method includes the steps of using deviation information as the input of a BP neural network, the BP neural network performing forward calculation and outputting a penalty factor, a step factor and other to-be-set parameters of an MISO tight format no-model controller through an output layer, calculating to obtain a control input vector for a controlled object by adopting a control algorithm of the MISO tight format no-model controller,conducting system error reverse propagation calculation for gradient information sets of all the to-be-set parameters by using a gradient descent method in combination with control output with minimizing the value of the system error function as a target, on-line updating the hidden layer weight coefficient of the BP neural network in real time, outputting the layer weight coefficient, and realizing the self-setting of parameters based on the deviation information. The parameter self-setting method can effectively overcome the online setting difficulty of the parameters of a controller, and agood control effect on an MISO system is achieved.