Network Representative Model for MV Distribution Grid by a Combination of Network Reduction and Time Aggregation

Planning of large-scale electric power systems presents significant computational challenges due to the extensive number of network components and time intervals. The primary objective of network reduction is to maintain fidelity to the original network's electrical behavior while minimizing computational complexity and simulation time. However, conventional network reduction methods may not fully capture the temporal dynamics and interactions present in the original network. Furthermore, their effectiveness for long-term planning horizons may be questionable. To bridge these gaps, this paper proposes a novel spatial and temporal system reduction approach using three-step time series and temporal aggregation. The first step performs time series aggregation to obtain a single representative time series from all data sources. In the second step, a diffusion-based clustering model is proposed for the first time, which adaptively adjusts the diffusion process based on temporal smoothness to enhance time series embeddings. In the third step, time series disaggregation is carried out using the average participation factors of all involved time series for representative time intervals. Performance assessments are conducted using a real-world 1759-bus system, where the accuracy of the proposed spatial and temporal system reduction method is evaluated under varying operating conditions.