Physics and data‐driven approach for online joint state and parameter estimation of electricity and steam networks

Abstract The electricity and steam integrated energy systems, which can capture waste heat and improve the overall energy efficiency, have been widely utilised in industrial parks. However, intensive and frequent changes in demands would lead to model parameters with strong time‐varying characteristics. This paper proposes a hybrid physics and data‐driven framework for online joint state and parameter estimation of steam and electricity integrated energy system. Based on the physical non‐linear state space models for the electricity network (EN) and steam heating network (SHN), relevance vector machine is developed to learn parameters' dynamic characteristics with respect to model states, which is embedded with physical models. Then, the online joint state and parameter estimation based on unscented Kalman filter is proposed, which would be learnt recursively to capture the spatiotemporal transient characteristics between electricity and SHNs. The IEEE 39‐bus EN and the 29‐nodes SHN are employed to verify the effectiveness of the proposed method. The experimental results validate that the proposed method can provide a higher estimation accuracy than the state‐of‐the‐art approaches.