Research On Decentralized Model Predictive Control And Its Application

From the classic model predictive control algorithm to the accurate explicit model predictive control algorithm,to explicit model predictive control approximation algorithm,the trend of research on model predictive control becomes clear.Explicit model predictive control algorithms need to be partitioned into multiple partitions when calculating offline,to spend more time.The great complexity is taken to offline and online computing.For this shortcoming,the idea of decentralized control,starting from the model itself will be the overall control tasks into several small tasks,and then using the explicit model predictive control methods to control.It is used to solve the linear,time-invariant,finite-time optimal problem.Comparison of control effects with centralized explicit model control algorithm,it shows that the decentralized model prediction algorithm can reduce the space and time complexity of the whole system and simplify the superiority of the overall control task.In this paper,the method of decentralized model prediction is applied to the example of building central air conditioners for simulation research.Due to the packets drop during decentralized control,the algorithm is optimized to ensure system stability.Meanwhile,Summarizes the methods and research results of explicit model predictive controller design,summarizes the basic ideas and related achievements of decentralized control,and provides research direction for the research of this paper.This paper makes deep researchs on the following aspects:1 Introduce the basic principle and the steps of EMPC controller design.Moreover,the computational complexity of approximate solution is further analyzed for its optimal problem with multiple constraints.The result of the analysis is to reduce the computational complexity of explicit model predictive control by using the idea of decentralized control.2 For the explicit model predictive controller needs to divide multiple partitions in the calculation,the calculation time is long,which makes the offline and online have the disadvantage of relatively large complexity.By using the decentralized control idea,by reducing the size of the original large model indirectly reduce complexity.The results verify the feasibility of the algorithm and the real-time performance in the control process.3 The decentralized model predictive controller is used in the building temperature control system to model the building temperature control system.Contrasting the trajectory of each control state and the control input sequence of the building temperature control system under the centralized model predictive algorithm and the decentralized model prediction algorithm.Comparing the control results of the two algorithms,it is shown that the control effect of the decentralized control algorithm is similar to that of the centralized algorithm.But in terms of computation complexity and time consumption,the decentralized control algorithm is better than the centralized control algorithm and effective for the improvement of computational complexity.4 In the decentralized control process,The large model is separated into several sub-models,packets often drops in the state information measurement control network sub-models.For this situation,the improved decentralized model predictive control algorithm is adopted.The results show that the improved algorithm is better than the unimproved algorithm when packets drop in the system.