Model predictive control of voltage profiles in MV networks with distributed generation

Abstract The presence of distributed generators in Medium Voltage (MV) networks can produce local voltage increase, with inversion of power flows, and emergence of dangerous inverse currents. For this reason, the control of the voltage profile is becoming of paramount importance. However, the design of a dynamic controller is problematic due to the multivariable and large scale nature of the problem, and to the difficulty to derive a reliable model of the system. In this paper, we first identify a MIMO impulse response model of the system, through a suitable identification phase, where a detailed industrial reference simulator of the network is used. Then we propose and design an MPC-based algorithm for control of the network, used at the intermediate level of a three-layer hierarchical structure. At the upper level a static Optimal Power Flow (OPF) computes the required voltage profiles to be transmitted to the MPC level, while at the lower level local Automatic Voltage Regulators (AVR), one for each Distributed Generator (DG), track the reactive power reference values computed by MPC. The proposed method allows to cope with constraints on the voltage profiles and/or on the reactive power flows along the network. If these constraints cannot be satisfied by acting on the available DGs, the algorithm acts on the On-Load Tap Changing (OLTC) transformer. A radial rural network with two feeders, eight DGs, and thirty-one loads is used as case study. The model of the network is implemented in DIgSILENT PowerFactory®, while the control algorithm runs in MATLAB ® . A number of simulation results is reported to witness the main characteristics and limitations of the proposed approach.