Modeling of furnace operation with a new adaptive data echo state network method integrating block recursive partial least squares

Abstract

Thermal efficiency is an important index related to industrial furnace operation. Modeling of reheating furnace operation based on optimal thermal efficiency can greatly improve the actual furnace operation performance. However, some key variables in the complex chemical processes are very difficult to measure due to the nonlinearity, the disturbances, and the technological limitations, and the multicollinearity among several correlation variables will make the established model inaccurate. In this paper, a new adaptive data echo state network modeling method integrating block recursive partial least squares (BRPLS-ESN) is proposed. First, the Echo state network (ESN) is used to process the historical data, and the calculation of the weights are handled by partial least squares (PLS) instead of the least-squares, which will overcome the multicollinearity among certain variables. The block recursive method is also used during the weight calculation, which will help to reduce the computation time and occupied the memory of the proposed PLS based algorithm. The prediction results will be obtained lastly after multiple iterations. In order to verify the effectiveness of the proposed BRPLS-ESN method, an industrial reheating furnace is introduced to build an operational model using the proposed algorithm. The traditional ESN, the back-propagation network, and the support vector regression are also tested to compare with the proposed method, and the results show that the proposed BRPLS-ESN can obtain a better modeling effect.