Extreme-learning-machine-based architecture to assess the reliability of the integrated energy systems

The global energy crisis has contributed to the development of integrated energy systems (IES). IES enables the coupling, conversion, and utilisation of multiple energy, such as electricity, heat and cooling. As renewable energy penetration increases, the precariousness of renewable energy in IES causes risks to the reliability of energy supply. Traditional optimal energy flow (OEF) models become complex and computationally intensive. To solve this problem, a machine-learning-based method is proposed instead of the traditional OEF model. The method obtains the model optimal coding parameters by extreme learning machine (ELM) model, establishes the nonlinear mapping relationship between the system operation state and the optimal load reduction, and realizes the high accuracy and fast solution of the reliability indices. In this article, simulation experiments are conducted on the integrated community energy system (ICES) to test the advantage and practicality of the approach in this paper. Compared with the model-driven approach, the method in this paper greatly improves the efficiency of reliability assessment and has an accuracy rate of over 95%, which is better than random forest, back propagation neural network and autoencoder.