Network-cognizant personalized optimization of distributed energy resources: Multi-period offline and online implementations

Considering the increasing penetration of distributed energy resources (DERs), distributed optimization has been widely adopted for DER management due to its unique advantages in scalability and privacy preservation. However, the heterogeneous preferences and characteristics of DERs hinder the convergence performance of distributed optimization and its tracking capability for fast time-varying fluctuations. To this end, this paper proposes a network-cognizant personalized optimization strategy of DERs based on a fast Newton-like alternating direction method of multipliers (FNL-ADMM). It utilizes the Hessian matrices, i.e., the second-order information of DER objective functions, to develop a personalized diagonal matrix for each DER by means of preconditioning methods. These personalized diagonal matrices are further leveraged to facilitate multi-period distributed DER management, subject to distribution network constraints, where the optimality and convergence properties of FNL-ADMM are also established. The FNL-ADMM accelerates the convergence rate of ADMM via a Newton-like manner and exhibits a fast convergence rate, indicating an excellent tracking capability for fast time-varying fluctuations. Its multi-period online implementation is developed to cope with the rapid environmental fluctuations by taking advantage of the real-time measurements and model predictive control. Test cases conducted for a modified IEEE 123-bus network demonstrate the effectiveness and superiority of our proposed strategy.