Abstract: Aiming at the problems of load-side users’ electricity bills, new energy consumption rate, and electricity peak-valley difference, an improved adaptive decomposition-based multi-objective evolutionary algorithm is proposed to optimize the controllable load dispatch of contracted households in the building microgrid. By analyzing the power consumption characteristics of the load, the load is divided into five categories, which establish mathematical models, optimize objective functions and constraints. A new adaptive weight vector is generated by combining generalized decomposition with uniform distribution, which makes the non dominated solution closer to the real Pareto front; the idea of historical experience is adopted to realize adaptive selection strategy by counting the contribution rate of SBX and DE crossover operators to external archive and using roulette. The subgeneration points are modified by the characteristic constraint mapping, which indirectly expands the search space of the algorithm and improves the diversity of the population. The convergence and superiority of the improved AWS-MOEA/D algorithm are verified by the test function; the simulation experiment results of the dispatching of the residents in a residential area show that the improved algorithm can save more electricity bills and effectively increase new energy consumption rate after dispatching.