Distributed storage operation in distribution network with stochastic renewable generation

The integration of distributed renewable energy generations and storages in the distribution system imposes new challenges on the paradigm of the operation. The resulting control problems are intrinsically hard because of nonlinearity in power flow equations, stochasticity in renewable generations, and inter-temporal coupling of decision problems due to energy storage. This paper provides an initial attempt to solve these nonlinear stochastic control problems numerically, within the framework of Markov decision processes (MDPs). Detailed MDPs are first designed for two-bus sub-networks of the distribution system. Optimal control policies for each sub-network are then combined to guide the design of a control policy to operate the entire storage system on a distribution network. The performance of this approach is demonstrated numerically by comparing with other typical operation policies.