Objective Trade-off in MPC Based Energy Management for Microgrids

Dominik Mildt, Marco Cupelli, Antonello Monti
2019 2019 IEEE PES GTD Grand International Conference and Exposition Asia (GTD Asia)  
This work extends on the findings in [9] , where the maximization of the Potential Islanding Time (PIT) was added for the first time to the pool of EMS objectives, i.e. maximizing the time that the MG could disconnect from the main grid, if an islanding command was received. The EMS presented in [9]-[10] is extended, specifically to include a much more accurate approximation of the cost of battery energy storage system (BESS) wear [11] , as compared to the simplified linear and quadratic
more » ... ns of power which are employed by the majority of works. Parameter variation is performed for a weighted linear combination of four objectives for a 48h simulation of a generic MG test system. Analysis shows the trade-off between specific objectives. Chapter II briefly introduces the developed EMS. Chapter III explains the compared objective functions. Chapter IV outlines the simulation setup. Chapter V provides simulation results and chapter IV finally gives a short conclusion. EMS DEFINITION The proposed EMS employs a Model Predictive Control (MPC) scheme regarding a control horizon of discrete time steps ∈ 1, ... , of equal length ∆t. It is developed for a generic low voltage (LV) MG consisting of busses ∈ and lines , ∈ ℰ between busses and , where bus 1 is the PCC. A radial feeder structure is assumed, which is typically the case in LV distribution grids. This allows the integration of an exact relaxation of the power flow constraints in their DistFlow formulation [12] . The overall problem is then posed as a Mixed Integer Second Order Cone Problem (MISOCP) [9][10]. The radial grid topology can be mapped to a connected, tree-structured graph where the nodes equal to the busses and edges equal to the lines. For each line , ∈ ℰ in the given grid, let , denote the squared current magnitude. The complex line impedance is given as , , , , separated into resistance and reactance. Analogously, , , , is the complex power flow from bus to , given as active and reactive power values. Similarly, for each system bus ∈ , denotes its squared voltage magnitude and its complex net power injection. The convex relaxation of the power flow for node with parent node is then stated by: Abstract-Microgrids (MGs) are considered one of the key enabling factors of future electric power systems. Their limited size allows optimal control, as the solution space of optimization problems increases exponentially with the number of assets, if significant model simplifications are not undertaken. Additionally, the capability to disconnect from the main grid and autonomously supply loads can serve to increase the resilience to failures in the main grid and be provided as a service to the Distribution System Operator (DSO). An energy management system (EMS) is responsible to facilitate the optimal use of resources; however, optimality can refer to a multitude of target objectives. This paper builds up on previous research, implementing a Model Predictive Control (MPC) based EMS that is used to show the trade-off between four different objectives in MG operation. The different objective functions are the minimization of operational cost, minimization of the energy exchanged with the main grid, minimization of resistive losses and maximization of potential islanding time (PIT) for a possible future islanding event.
doi:10.1109/gtdasia.2019.8715952 fatcat:27lyy3gprrd7xcsuqtanruy27y