LAUSR

Coordinated operation of integrated electricity and heat system can improve operation flexibility and reduce cost. However, multiple uncertainties challenge its optimal operation. This paper aims at developing a risk-averse and computationally efficient policy for real-time stochastic dispatch of integrated electricity and heat system, which improves the economy as well as avoiding the risk of high costs in critical scenarios. First, real-time dispatch of integrated electricity and heat system is formulated as a multistage risk-averse stochastic sequential optimization problem with dynamic risk measure, where combined heat and power unit, energy storage, flexible electricity and heat load are jointly utilized to minimize the risk-adjusted total costs. Next, a risk-averse dynamic programming formulation of the original problem is presented, upon which a data-driven risk-averse approximate dynamic programming is employed to address computational challenge, and develop almost optimal and computationally efficient policy. By exploiting information from training samples in off-line learning, the proposed algorithm can efficiently responses to the stochastic exogenous information. Comparative simulations with different risk-aversion preferences and different methods verify the effectiveness of the proposed algorithm.