Study On Echo State Network Predictive Control Based On Firefly Optimization Algorithm

Predictive control, which uses the known, past and present information to get control of the desired output behavior of future, is consisted mainly of four parts of reference trajectory, predictive model, roll optimization and feedback correction. Currently, the theoretical research of linear predictive control has been more mature, and its practical application is also very extensive, but the complicated nonlinear systems’predictive control has not been well solved, and it still universally exists the problems of difficult or low accuracy of nonlinear modeling, defective strategy of rolling optimization and so on. Aiming at these problems, this paper studies a kind of algorithm of predictive control based on nonlinear system for the controlled object, which can establish a high precision prediction model and fast and effective implement the online rolling optimization.The echo state network, which compared with traditional networks, has the unique advantages of simple calculation process, fast learning and training speed, high prediction accuracy, and ensure the stability of the network and the global optimality and so on. On the basis of deeply analyzing the theory of echo state network, this paper simulates the nonlinear object predicted by using the echo state network and BP network for comparison, and the experimental results prove that the echo state network has better prediction performance.The firefly algorithm, which compared with other optimization algorithms, has the advantages of high calculation efficiency, less parameters, better optimizing effect and so on. Through deeply analyzing the theory of firefly algorithm, this paper utilizes the firefly individual’s space location and light intensity to structure a stochastic inertia weight, which is applied to the equation of position renewal, and obtains the new adaptive firefly optimization algorithm. After the test consisting of five nonlinear functions’ optimal value and operation time and the Matlab simulated experiment of PⅡ) parameters tuning by choosing three kinds of intelligent optimization algorithm for comparison, The results further show that the adaptive fireflies optimization algorithm is suitable for nonlinear objective functions’ optimization, and the optimal speed is faster, the effect of solve problems are better. On the basis of analyzing the predictive performance of the echo state network and the optimization effect of the improved firefly algorithm, this paper uses the echo state network for establishing model, uses the adaptive fireflies optimization algorithm as the tool of roll optimization, so that a kind of algorithm of the echo state network predictive control based on firefly optimization algorithm can be constituted. After a study of the new algorithm’s feasibility, structure and process, the results of simulating the nonlinear objects verified the effectiveness of the new algorithm. In the end, it is applied to the simulation test of predictive control of the double-capacity water tank liquid level, the results show that the new algorithm is effective in the tracking control.