Distributed Algorithms For Energy Management In Smart Grid

On top of the physical electricity network,a smart grid also includes a communication net-work as well as a variety of renewable energy resources and flexible loads,advanced metering infrastructures,and control techniques.Similar to traditional power systems,energy management is also of vital importance to smart grids,in the sense that optimal energy management not only brings remarkable economic benefits,improving the economic efficiency of smart grids,but al-so enhances the operational stability.In a smart grid scenario,traditional centralized algorithms may be infeasible or unscalable due to limited communication and computation capabilities.We therefore need to design distributed algorithms such that through local computation and commu-nicating with neighbors,all the nodes can cooperatively achieve a global goal.This thesis mainly studies the distributed algorithms for energy management in smart grid,the contributions of which include:1.Some preliminaries and the background of smart grid,distributed algorithms,and demand-side management are presented.The state-of-the-art research progresses in both international and domestic academia are also introduced.2.We have studied the static economic dispatch problem(EDP)which aims at minimizing the total generation costs at one single instant,and proposed the distributed bisection method to solve the EDP in a distributed fashion.The proposed algorithm is based on the "consensus-like" algorithm and adopts the idea of bisection to asymptotically find the optimal Lagrange multiplier and the optimal solution.The proposed algorithm has mild requirement on com-munication network in the sense that it is applicable to connected undirected graphs as well as strongly connected digraphs.3.We have further considered the dynamic EDP with energy storage which aims at the mini-mization of aggregated generation costs over multiple periods,where energy storage can be used for energy arbitrage as well as providing spinning reserves.To solve this problem,we propose a distributed algorithm based on alternating direction method of multipliers(ADM-M).Adopting the idea of ADMM,we decompose the original problem into quadratic pro-gram(QP)subproblems which are solved locally by each node as well as an unconstrained QP for the aggregator.4.We have investigated the optimal charging and discharging scheduling problem of electric vehicles(EVs)which aims at flattening the aggregated demand curve as much as possible on the premise that the energy requirement of each EV is satisfied.We assume that EVs can release energy back to the grid using inverters such that EVs can be used for not only "valley filling" but also for "peak shaving".Since we consider energy losses in both directions of energy conversion,the optimal charging and discharging scheduling problem is inevitably formulated as a mixed integer programming(MIP)problem.To solve the MIP problem,we first reasonably assume the characteristic of the night base load,and then approximate it into a QP problem for which a distributed algorithm is proposed.5.At last,we have studied the demand-side management through thermostatically controlled loads(TCLs)and proposed a distributed control algorithm,through which the controlled TCLs jointly follow a demand signal while fair dispatch in the sense of thermal comfort is achieved.