To address the energy shortage problem in rural areas, significant attention has been paid to off-grid solar power plants. However, ensuring the security of these plants, improving the utilization rate of energy and, finally, proposing a sustainable energy development scheme for rural areas are still challenges. Under this, this work proposes a novel regression model-based stand-alone power plant load management system. This not only shows great potential in increasing load prediction in the real-time process but also provides effective anomaly detection for improving energy efficiency. The proposed predictor is a hybrid model that can effectively reduce the influence of fitting problems. Meanwhile, the proposed detector exhibits an efficient pattern matching process. That is, for the first time, a support vector machine (SVM) and the fruit fly optimization algorithm (FOA) are combined and applied to the field of energy consumption anomaly detection. This method was applied to manage the load of an off-grid solar power plant in a rural area in Tanzania with more than 50 households. In this paper, both the prediction and detection of our method are proven to exhibit better results than those of some previous works, and a comprehensive discussion on the establishment of a real-time energy management system has also been proposed.
Turbomachinery Engineering and Optimization for 25 MW and 50 MW Compressed Air Energy Storage Systems
