Bagging of Gaussian Process for Large Generator Eddy Current Prediction

This paper proposes a large generator eddy current prediction method based on Bagging of Gaussian Process(GPR). Bagging ensemble learning model based on Gaussian Process Regression is proposed to predict eddy current loss. The slot wedge conductivity, slot wedge of the relative permeability, rotor outer diameter and stator outer diameter are as the input of the prediction model, and the eddy current loss is as the output. The experiments on the datasets calculated by Finite Element Model (FEM) show that the proposed approach has good predictive performance for Large generator rotor performance prediction and can be applied to practical projects.

I. Introduction

Steam Turbine generator is a generator driven by steam turbine. The superheated steam produced by boiler expands into steam turbine to make work, which makes the blades rotate and drives the generator to generate electricity. As an important equipment of power system, the safe and stable operation of large steam turbine generator is very important with the development of the generators, electromagnetic and thermal load increase. This could cause local overheating, which could lead to insulation deterioration, core slackness and sets vibration in generators. All of these could cause abnormal downtime, or threat to-power system security. This brings new opportunities and challenges for the design and operation of turbo-generators [1]–[2].