Real-Time Energy Management in Microgrids With Reduced Battery Capacity Requirements

Energy storage units hold promise to transform the electric power industry, since they can supply power to end customers during peak demand times, and operate as customers upon a power surplus. This paper studies online energy management with renewable energy resources and energy storage units. For the problem at hand, the popular approaches rely on stochastic dual (sub)gradient (SDG) iterations for a chosen stepsize <inline-formula> <tex-math notation="LaTeX">${\mu }$ </tex-math></inline-formula>, which generally require battery capacity <inline-formula> <tex-math notation="LaTeX">${\mathcal{ O}}{(1/\mu)}$ </tex-math></inline-formula> to guarantee an <inline-formula> <tex-math notation="LaTeX">${\mathcal{ O}} {(\mu)}$ </tex-math></inline-formula>-optimal solution. With the goal of achieving optimal energy cost with considerably reduced battery capacity requirements, an online learning-aided management (OLAM) scheme is introduced for energy management, which incorporates statistical learning advances into real-time energy management. To facilitate real-time implementation of the proposed scheme, the alternating direction method of multipliers method is also leveraged to solve the involved subproblems in a distributed fashion. It is analytically established that OLAM incurs an <inline-formula> <tex-math notation="LaTeX">${\mathcal{ O}}{(\mu)}$ </tex-math></inline-formula> optimality gap, while only requiring battery capacity <inline-formula> <tex-math notation="LaTeX">${\mathcal{ O}}{(\log ^{2}(\mu)/\sqrt {\mu })}$ </tex-math></inline-formula>. Simulations on the IEEE power grid benchmark corroborate that OLAM incurs similar average cost relative to that of SDG, while requiring markedly lower battery capacity.