A Data Mining Model and a Real-Time Predictive Software Prototype for the Spatial Design and Planning of High Energy Performance Solar Community Microgrids

open_access
Open Access
Author:
Rahimian, Mina
Graduate Program:
Architecture
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
May 05, 2021
Committee Members:
Jose Pinto Duarte, Chair & Dissertation Advisor
Seth Blumsack, Outside Unit Member
Lisa Iulo, Major Field Member & Dissertation Advisor
Guido Cervone, Outside Field Member
Ute Poerschke, Program Head/Chair
Keywords:
Data Mining
Machine Learning
Community Microgrid
Energy Consumption
Urban Scale
Energy Performance
Predictive
Software
Abstract:
With severe natural disasters occurring around the globe, cities are experiencing the consequences of climate change more than before. Frequent power outages attributed to aging equipment of power distribution systems and coupled with natural disasters such as hurricanes or wildfires are threatening everyday lives and businesses of urban dwellers. Communities, especially urban communities, which have experienced frequent blackouts are taking a closer look at adopting microgrid technologies to operate independently from the main power grid during emergencies. Microgrids are local, decentralized power distribution systems involving the use of power sources such as solar panels and diesel engines and storage devices like batteries to provide electricity for a cluster of buildings. Providing resiliency and reliability under unexpected power interruptions, microgrids have typically been used for backup supporting critical loads such as military bases and hospitals. However, with the increasing environmental concerns associated with fossil fuels and the frequency of natural disasters, a growing interest in adopting microgrid technologies is rising in towns and communities in the interest of transitioning into energy-independent urban settlements. Known as community microgrids, these energy-independent urban settlements are generally comprised of various mixes of residential, commercial, agricultural, and industrial loads followed by the local renewable and/or nonrenewable sources of power. As with any other energy system, the efficiency of a community microgrid’s energy performance is evaluated by comparing the energy inputted to the system from the on- and off-site sources of energy, to the energy that is outputted from the system, mostly in the form of useful energy for buildings operation. Current research on improving energy performance in community microgrids has been exclusively advocating technological advances enhancing the limited supplies of local energy and addressing the constantly growing demands of the loads. However, researchers argue that focusing on technological innovations alone wouldn’t solve the current energy issues in the built environment. In the case of community microgrids this statement is especially accurate since they are contextualized in cities and urban areas; citing research from the 1960’s onwards, considerable attention has been directed towards the impact that spatial structure of urban form has on the energy required for space heating, cooling and lighting as well as the feasibility of adopting on-site renewable energy generators such as Photovoltaic (PV) panels and wind turbines. Literature today emphasizes the importance of obtaining an energy-conscious point of view when taking actions toward urban design and planning. Architects and urban planners are expected to consider the tradeoffs between the living qualities of an urban context and its potential for high-performance energy systems design and engineering. Despite the evident need for involving architects and urban planners in the development process of urban energy systems as community microgrids, in practice this engagement is generally neglected. This is possibly due to the complexity of understanding urban form and its impact on energy performance in community microgrids and the unavailability of custom tools for the spatial design and assessment of these energy systems. The intention of this research is to engage architects and urban planners in the process of developing and constructing community microgrids. Benefiting from artificial neural networks, this research adds a spatial dimension to the existing technical discourse of developing high energy performance community microgrids and by surrogate modeling, delivers a real-time energy simulation software prototype that aids architects and urban planners in designing and assessing urban scale energy systems. This dissertation focuses on solar community microgrids with San Diego county serving as a case study for data production. In this research artificial neural networks are used to uncover the complex relationship between San Diego’s urban form and its impact on community scale energy consumption and PV energy production as the two main pillars of evaluating energy performance. Results of the training procedure proves the existence of a strong statistical relationship between urban form and energy performance in communities. By architecturally and spatially interpreting the results of the outputted statistical model, a comprehensive set of design principles were extracted from this analysis that guides architects and urban designers towards spatially designing high energy performance community microgrids in San Diego. For instance, to spatially design a high energy performance community microgrid in the coastal region of San Diego, the distance between the community buildings needs to be maximized to prevent overshading adjacent buildings and minimize the community’s net energy demand for heating. A spatially aware design and assessment of solar community microgrids as studied herein brings the need for a new generation of computational modeling, simulation, and evaluation tools for the field. In this regard, an urban scale energy simulation software prototype is developed that predicts the energy performance of any given solar community microgrid design scenario - in real-time - by the virtue of its urban spatial configuration. The real-time prediction feature of this software (which is due to benefiting from the trained machine learning models at its backend rather than hard coding physical laws of energy) is a major contribution to the field and sets an example for future simulation software development, since current existing urban scale energy simulation tools are extremely time and resource consuming.