A vision-based fault diagnosis system for heliostats in a central receiver solar power plant

This paper presents an automatic heliostats fault detection and diagnosis system using machine vision techniques and common CCD devices for a solar power plant. The heliostats of a solar power plant reflect solar radiation onto a receiver placed at the top of a tower in order to provide a desired energy flux distribution correlated with the coolant flow through the receiver, usually in an open loop control configuration. Each heliostat maintains reflection of the moving sun onto the receiver. A long time running will make the mechanical components which control the heliostat to modify the azimuth angle and pitch angle break down, so the heliostats cannot reflect sunlight to the receiver or even stop working. In a large power plant, there may be hundreds to hundreds of thousands of heliostats which will increase the complexity of manually recognizing and detecting which heliostat is fault or broken-down. Each heliostat can be equipped with sensors or some other equipment to detect whether fault occurs, but it will greatly increase the cost. So a novel method for fault diagnosis, which is based on image processing and machine vision, is presented in this paper. Experiments have shown promising results.