摘要： The use of photovoltaic solar power generation is rising as worldwide energy demand increases. Therefore, reliability, safety, life cycle, and improved efficiency of photovoltaic plants have all become a major concern in research nowadays. In this context, monitoring systems are necessary to guarantee the required operating productivity and to avoid overpriced maintenance costs. This paper studies the non-ideal operating conditions for grid-connected photovoltaic plants and proposes an anomaly detection methodology that combines the advantages of the 2-sigma, short-window simple-moving average control charts with shading strength and irradiance transition parameters to detect early deviation in photovoltaic plant operational data. The key aspect of proposed methodology is that it requires neither historical data for model training procedure nor parameters from previous simulation. Only instantaneous meteorological and electrical parameters are required. The efficiency of the condition monitoring methodology has been validated through experimental results conducted in actual operating conditions. Results demonstrated that the proposed methodology is effective to identify non-ideal operating conditions for grid-connected photovoltaic plants, i.e., (i) normal operating condition, (ii) natural dynamic shading, (iii) artificial dynamic shading, and (iv) artificial static shading. Moreover, a low-cost and non-invasive internet-of-things-based embedded architecture is proposed to monitor photovoltaic plant operation in real-time.