A Model-Predictive Control Method for Coordinating Virtual Power Plants and Packetized Resources, with Hardware-in-the-Loop Validation

TL;DR

This work tackles power-system balancing in high-DER contexts by integrating packetized energy management with model-predictive control in a bi-level framework and validating it on a hardware-in-the-loop cyber-physical platform. The upper-level scheduler provides economic references while the lower-level MPC dispatches generators and PEM-enabled VPPs under energy-state constraints, solving an open-loop problem over a horizon to minimize $J^*$ with dynamics and constraints captured in a linearized model. The results show that conventional AGC can lead to VPP capacity saturation, whereas energy-aware MPC maintains service by adjusting VPP SOC and dispatch in a receding-horizon manner, demonstrated on a realistic 161-bus system with PEM DERs. The findings highlight the practical impact of incorporating energy-state awareness into real-time dispatch, improving reliability and reducing the risk of saturation in flexible resources.

Abstract

In this paper, we employ a bi-level control system to react to disturbances and balance power mismatch by coordinating distributed energy resources (DERs) under packetized energy management. Packetized energy management (PEM) is a novel bottom-up asynchronous and randomizing coordination paradigm for DERs that guarantees quality of service, autonomy, and privacy to the end-user. A hardware-in-the-loop (HIL) simulation of a cyber-physical system consisting of PEM enabled DERs, flexible virtual power plants (VPPs) and transmission grid is developed in this work. A predictive, energy-constrained dispatch of aggregated PEM-enabled DERs is formulated, implemented, and validated on the HIL cyber-physical platform. The energy state of VPPs, composed of a fleet of diverse DERs distributed in the grid, depending upon the distinct real-time usage of these devices. The experimental results demonstrate that the existing control schemes, such as AGC, dispatch VPPs without regard to their energy state, which leads to unexpected capacity saturation. By accounting for, the energy states of VPPs, model-predictive control (MPC) can optimally dispatch conventional generators and VPPs to overcome disturbances while avoiding undesired capacity saturation. The results show the improvement in dynamics by using MPC over conventional AGC and droop for a system with energy-constrained resource

Figures (5)

• Figure 1: Cyber-physical platform overview: The transmission grid is simulated on the OP5600 and MPC-corrective dispatch is realized on a host PC and generates balancing signals. ESP8266 devices are connected to a python-based server via WiFi and transmit the VPPs' states through the analog interface. The packetized load is emulated on a high performance PC and requests energy packets from the VPP through WiFi communication.
• Figure 2: Overview of control scheme showing controller including OPF and MPC part and how each part is related to the power grid.
• Figure 3: Diagram showing a control schematic for the test system including all of the generation in the internal and external areas.
• Figure 4: (a) The HIL VPP's actual and reference power (MW) (b) Grid scale battery's actual power (MW), reference power (MW) and state of charge (SOC $\%$) during charge/discharge events (c) Generators' power output (MW) (d) Generators' mean frequency (Hz). The saturation of the VPP to a step decrease in load is shown in this figure. For the change in load, the HIL VPP and the battery charges at a continuous rate. The battery saturates at about t = $36$ mins, after which their output goes to zero and cannot support the requested flexibility.
• Figure 5: (a) The HIL VPP's actual power (MW) and reference power (MW) (b) Grid scale battery's actual power (MW), reference power (MW) and state of charge (SOC $\%$) during charge/discharge events (c) Generators' power output (MW) (d) Generators' mean frequency (Hz). MPC with capacity saturation takes into consideration the current state of charge of the VPP and initially ramps up to the requested $50$ MW. However, at t = $13$ mins, MPC lowers the setpoint in steps to avoid VPP saturation and provide support to the system for a longer time.