This paper discusses the design and simulation of an integrated load forecasting and state estimation tool for distribution system operations. A predictive database is created and applied to forecast futurenetwork states in order to allow short-term (e.g., hours/days ahead) planning to be carried out. The predictive database is based on adaptive nonlinear auto-regressive exogenous (NARX) load estimation and forecasting models, which are continuously updated using feedback from the state estimator. This creates a closed-loop information flow designed to continuously monitor and improve the system state estimation performance by updating and retraining models where appropriate.
The aim of this methodology is to improve situational awareness and help to provide network operators with early warning of potential issues, in medium voltage (MV) networks where the number of on-line measurements is limited, and state estimation relies heavily on estimates of power injections. The applicability of the approach is demonstrated through simulation using supervisory control and data acquisition (SCADA) and smart meter measurements recorded from an actual MV distribution network.