Arbitraging Variable Efficiency Energy Storage Using Analytical Stochastic Dynamic Programming

Vehicle-to-Grid Fleet Service Provision considering Nonlinear Battery Behaviors

“…We take a two-step approach to solving the EVCS control problem (1). First, we formulate the charging session of a single EV as a price arbitrage problem, which satisfies all constraints in (1) except the EVCS power limit constraints (1f) and (1g), and solve it using an analytical SDP method [35]. Second, we aggregate the control policies from the EVs in an active session at the current time step and prioritize the distribution of control signals according to an LLF approach [14] to ensure compliance with the EVCS power limit.…”

“…The EVCS operator can access the system's day-ahead price (DAP) information and a DAP-based real-time price (RTP) prediction tool. For this work, RTP is modeled as a 1storder Markov process with 12 nodes per time step trained with historical RTP-DAP bias price data as in setting DB-Dep in [35]. The proposed solution algorithm uses the resulting prediction to provide a control decision that accounts for future price uncertainty.…”

“…The core of the proposed control method is a single energy storage price arbitrage problem formulated as an SDP [35].…”

“…We extend the solution approach from [35] by incorporating the SoC-dependency of battery behavior parameters into the first-order optimality condition expression as…”

“…We restate the control policy from [35]. After the value function computation step, control can be executed by responding to realized market prices and looking for the closest price node 𝜋 𝑡 ,𝑖 such that 𝜋 𝑡 ,𝑖 ≤ 𝜆 𝑡 < 𝜋 𝑡 ,𝑖 , then the storage control decision is updated as…”