Loading...
Scalable distributed model predictive control technique to achieve a system-level objective by shifting the demand of buildings and electric vehicles, A
Wald, Dylan J.
Wald, Dylan J.
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2025
Date Submitted
Collections
Files
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Aggressive clean energy goals and incentives have prompted an increase in the electrification of end use devices, such as buildings and electric vehicles (EVs). This increase has led to a larger electrical demand, stressing the capabilities of the current electric grid. It can be extremely costly and time consuming to develop grid infrastructure, increasing the uncertainty of its future reliability. However, as these end use devices are electrified, so is their flexibility and controllability. Hence, this thesis proposes an alternative approach to improve grid reliability through the use of advanced control techniques, such as model predictive control (MPC). Through MPC, the increased device flexibility can be exploited, intelligently shifting their demand to achieve a system-level objective, such as improving the reliability of the grid. This thesis is split into two themes: proof of concept and scalability and system-level impact. First, the network Lasso - Alternating Direction Method of Multipliers - limited Communication distributed Model Predictive Control (NALD) algorithm is introduced. NALD is used as a proof of concept, cooperatively controlling two different devices (a commercial building and an EV charging station), each with different objectives and optimization algorithms, to achieve both their local objectives and a system-level objective. NALD is then updated to handle a more generalized use-case without a loss in performance. Next, the forecast-aided prediction control (FAPC) framework is introduced. In FAPC, the powerful load shifting capabilities of the NALD algorithm are demonstrated in the event of realistic, intermittent generation. With the concept proved, the focus is shifted to system scalability and its implications in regards to the system-level impact. First, the adaptive neural parameter-varying model predictive control (ANPV-MPC) algorithm is introduced. The goal of ANPV-MPC is to improve the effectiveness of a controller without increasing the computational complexity of the optimal control problem. ANPV-MPC is used to control the heating, ventilation, and air conditioning (HVAC) system in a residential building, showing that combining machine learning and parameter-varying control can successfully improve the operation of a building without increasing complexity of the control problem. ANPV-MPC is then scaled to a larger system, where a distributed version, ANPV-DMPC is introduced. ANPV-DMPC controls each HVAC system across a large number of residential buildings, in parallel, to analyze the system-level impact on a test distribution system. It is shown that ANPV-DMPC can successfully shift the demand at a large scale to regulate the node voltages of the test distribution feeder.
Associated Publications
Rights
Copyright of the original work is retained by the author.