Optimal energy management and control of microgrids in modern electrical power systems

Microgrids (MGs) are becoming more popular in modern electric power systems owing to their reliability, efficiency, and simplicity. The proportional-integral (PI) based droop control mechanism has been widely used in the MG control domain as the setpoint generator for the primary controller which has several drawbacks. In order to mitigate these issues, and to enhance the transient and steady-state operations in islanded MGs, advanced control and intelligent optimization methodologies are presented in this dissertation. First, to improve the existing PI-based droop relationship in DCMGs, a multi-objective optimization (MOO) based optimal droop coefficient computation method is proposed. Considering the system voltage regulation, system total loss minimization, and enhanced current sharing among the distributed generators (DGs), the Pareto optimal front is obtained using the Elitist non dominated sorting genetic algorithm (NSGA II). Then, a fuzzy membership function approach is introduced to extract the best compromise solution from the Pareto optimal front. The drawbacks of PI-based droop control cannot be entirely mitigated by tuning the droop gains. Hence, a droop free, approximate optimal feedback control strategy is proposed to optimally control DGs in islanded DCMGs. Further, to gain the fully optimal behavior, and to mitigate constant power load (CPL) instabilities, a decentralized optimal feedback control strategy is also introduced for the active loads (ALs) in the MG. In both algorithms, the approximate dynamic programming (ADP) method is employed to solve the constrained input infinite horizon optimal control problem by successive approximation of the value function via a linear in the parameter (LIP) neural network (NN). The NN weights are updated online by a concurrent reinforcement learning (RL) based tuning algorithm, and the convergence of the unknown weights to a neighborhood of the optimal weights is guaranteed without the persistence of excitation (PE). Finally, a local optimal control strategy is presented to path optimization of islanded ACMGs to enhance the transient operations while mitigating the voltage and frequency deviations caused by the traditional droop control. Optimal state and control transient trajectories in the d-q reference frame are obtained by Pontryagin's minimum principle which drives each DG from a given initial condition to their steady-state manifold. Both simulation and experimental results are presented to validate the concepts