Expert System For Fault Diagnosis Of Molten Salt Receiver In The Solar Tower Power Plant

Solar Tower Power (STP) Plant is one of the utilized patterns of Concentrate Solar Power (CSP). The larger the plant is, the higher efficiency it will obtain, which is an outstanding advantage comparing with PV or other forms of solar thermal power. This promises STPP a bright prospect for commercialization. The use of molten salt as heat transfer fluid makes it possible that heat can be transmitted with high temperature and normal pressure in the plant so that the equipments face no high pressure problems. On one hand, molten salt helps the receiver to reach a higher temperature and can boost the plant's performance as a whole. On the other hand, molten salt will be used as heat storage media as well, which reduces heat transfer equipments and as a consequence reduces heat loss, to at last improve the plant's efficiency. However, a higher temperature calls for a more rigorous receiver protection. Of all the faults overheat happens most frequently. The local excessive temperature will lead to receiver structure damage and molten salt decomposition, or even a reciver shut down. This affects the solar power plant's long running and leaves great safety and economic risks.In this paper an expert system was designed for the fault diagnosis of the receiver, especially against the overheating fault. In order to find out the main factors influencing overheat, this paper used a CFD software to analyze temperature sensibilities of some easily obtained parameters. Numerical experiments about the receiver and molten salt fluid were done and data was analyzed. Results show that four parameters affect local maximum temperature. They are the maximum heat flux density qmax, the thickness of the tubeδ, the fluid velocity u and the fluid average temperature tf. Their temperature influence trends are also given in the paper with explanations. To employ this conclusion, a number of numerical simulations were done, with four parameters given many different values within separately allowable ranges as initial boundary conditions. The specific information they effect temperature was extracted from the simulation responses. An Artificial Neural Network (ANN) was then applied to identify the functions between these parameters and local maximum temperature. The fitting result is represented as a matrix and is written into the expert system as expert knowledge. When operating, real-time collection of the parameters'values will be put into the matrix, and the forecasting temperatures will be exported to be compared with allowed maximum temperatures to estimate whether there is an overheating fault occurring.