Interpretable controlled object model of furnace temperature for MSWI process based on a novle linear regression decision tree

The furnace temperature is one of the most critical controlled variables in the municipal solid waste incineration (MSWI) process. The primary challenge of intelligent optimal control is to construct a high-precision and interpretable controlled object model in terms of furnace temperature. To address this problem, this article proposes a novel modeling method, i.e., linear regression decision tree (LRDT), to construct furnace temperature with airflow and grate speed as the inputs. LRDT model consists of (T/2-1) intermediate nodes and T leaves. The intermediate nodes are specified by the mean square error for developing the tree-based model structure. These leaves provide the predicted output by operating the Tikhonov least square method, which can boost the prediction performance. Moreover, LRDT uses leaf prediction under a unique path as the final output, which improves the interpretation of the LRDT-based furnace temperature model. Finally, the proposed method is verified by using actual MSWI process data, and the interpretability of the model is analyzed in detail.

I. Introduction

The municipal solid waste incineration (MSWI) process is a typical complex process industry [1],[2], which can realize the transformation from waste-to-energy (WTE) via fermentation, combustion, heat exchange, and purification [3]. Considering the potential value of MSWI in the economy and environmental protection, MSWI technology has been widely applied in developed countries [4]. However, the automatic combustion control (ACC) [5] of the MSWI process is difficult to replicate in developing countries because of the waste management and regional differences between developed and developing countries [6],[7]. Hence, to study the ACC technology with regional characteristics is necessary, in which intelligent optimal control (IOO) of furnace temperature is one of the essences of ACC. Building a controlled object model for furnace temperature is the basis of IOO. Therefore, furnace temperature modeling has become one of the challenging problems that must be solved [8].