A Neural Network Construction Method for Auto-Tuning of Feedforward Signal for Boiler Control of Thermal Power Plants

Abstract

This paper proposes a construction method of a neural network for auto-tuning of feedforward signal for the boiler control of thermal power plant. The neural network costructed by the proposed method produces an appropriate feedforward signal by tuning weights of the network. The weights are adapted effectively by using PI control output of teaching signal. The proposed neural network was evaluated not only by use of the simple model which expressed crucial features of the thermal plant, but also by the detail simulator which faith fully represented the features of the actual power system, and proved to have appropriate control per-formances.