Soft-Sensor Model of Gas Concentration Based on WOA-ESN Network

摘要

Aiming at gas concentration with characteristics of nonlinearity and time-varying, we propose a soft-sensor model based on WOA-ESN network. The traditional echo state network (ESN) has good performance in the prediction of nonlinear chaotic system, but it's easy to produce pathological solution in the process of output weight. Whale optimization algorithm(WOA) has the characteristics of high search precision and strong search ability, so we adopt it to optimize the output weight of ESN to solve the problem of pathological solution. Thus, the model of prediction accuracy is improved and the dynamic characteristics of ESN are maintained. Through the simulation and verification of the real-measured mine historical gas concentration, comparing with traditional ESN and PSO-ESN models, our proposed soft-sensor model is more accurate in precision, and its feasibility and validity of the soft-sensor model are proved.