Abstract

Concentrated Solar Power (CSP) plants generate extensive multivariate time series data due to their operational complexity. In these plants, heat transfer fluids in parabolic trough systems are crucial in collecting and transferring solar heat to the power generation system. However, detecting anomalies such as vacuum heat losses in parabolic trough systems is challenging. This is further exacerbated by the daily fluctuations in solar radiation, which make accurate measurements difficult. To address these challenges, we propose a new automated anomaly detection algorithm using least linear squares approximation. The application of our approach to time series data from a commercial CSP plant in Spain provides promising preliminary results. Future applications of this methodology can improve the reliability and efficiency of parabolic trough systems in CSP plants, which represents a key aspect of our future work.