Autoencoder-Based Fault Diagnosis For Hydropower Plants

Industrial 4.0 is a technological revolution that encompasses innovative concepts including: the Internet of Things (IoT), cloud computing, data analytics, Artificial intelligence (AI), and machine learning algorithms. The benefits of industry 4.0 provide innovative approaches to increase the efficiency and performance of industrial equipment. One of the most crucial aspects of the production process is maintenance to prevent breakdowns. The significance of maintenance has increased in modern industry, and companies are aware of that. When it is done properly, maintenance plays a key role in achieving organizational objectives. Therefore, the implementation of predictive maintenance for fault detection and diagnosis has made significant progress in various sectors of the industry. In this paper, we concentrate on hydropower plants as one of the most valuable renewable energy resources and by using autoencoders networks, obtain a smart model for fault detection that would be so convenient for avoiding any potential damage or huge breakdowns. Eventually, To validate the performance of this model, we utilize one of the famous multi-variable control charts - T2 Hotelling chart - as a criterion.