Design Requirements and Technical Architecture of a Fault- and Restoration-Based Load Forecasting Mechanism for Tanzania Secondary Distribution Electric Power Grid

Several Load Forecasting (LF) models have been proposed in the existing literature. The proposed models are limited to forecasting in a fault-based and developing countries’ context whose load distribution is influenced by both long- and short-term factors. Thus, this paper identifies the design requirements and proposes an architecture for forecasting the load demand profile based on Tanzania’s fault and service maintenance information, using short- and long-term electricity consumption determinants. The requirements of the proposed architecture are obtained through focus group discussion and a thorough literature review. Moreover, the core forecasting part of the architecture is achieved using an extended Multivariate Non-Linear regression (MNLR) model. The designed architecture is then programmed in an R Studio programming environment. The results show the ability of the proposed mechanism to generate load profiles based on fault and service maintenance information.