Scenario generation with clustering for optimal allocation of renewable DG

Active introduction of renewable distributed generation (DG) to the distribution system has become one of the challenging research in recent years. In the introduction plan of the renewable energy in Japan, long-term goal for high penetration of DG towards 2030 is shown. However, specific implementation plan such as the allocation, capacity, and timing has not been decided. In this paper, for decision-making of introducing DG, a novel scenario generation method with K-means is proposed. This method generates electric load, wind, and PV-production scenarios representing uncertainty, which are used as input for stochastic programming. Then, the problem is formulated as two-stage multiperiod mixed-integer linear programming (MILP) model. In this model, investment variables are decided in the first stage and operation and maintenance variables that depends on scenarios are solved in the second stage. This model minimize the total distributed system cost that is affected by demand growth. Constraints include power flow constraints, substation and feeders capacities, investment constraints, voltage, and current limits. In numerical simulation, several scenarios are generated using historical weather and demand data of Japan and the model is tested on 34-bus system.