Improved SPSA-based control and its application to lithium bromide system

In the industrial process, we often encounter some complex systems with large changes in operating conditions, strong nonlinearity and time-lag problems. In this paper we propose an improved model-free control method based on SPSA. Firstly, the lag-time constant τ is added to the control rate of the control scheme. Then according to the error curve and neural network, the λ is adaptively optimized. In order to verify the effect of the improved algorithm, simulations are carried out on linear system as well as nonlinear system respectively. As a practical application, the lithium bromide unit model is used in our simulation. Under different working conditions, the original SPSA algorithm and the improved SPSA algorithm are tested. The results show that the improved SPSA algorithm has strong resistance to noise and can achieve ideal control effect.