For the first time, and in response to the widespread effects of the COVID-19 pandemic, the 2020 Summer Study will be held virtually, which brings the added benefit also of widening access to new audiences.
The accepted papers for 2020 and their abstracts are:
Buildings as Batteries: Connecting AC and PV for Efficiency and Demand Optimization in Large Commercial Buildings
Craig Roussac and Hao Huang
It is becoming increasingly common for large commercial buildings and shopping malls to install photovoltaic (PV) arrays, sometimes with generation capacities that exceed 50% of the building’s peak electricity demand. During warm and sunny weather conditions, this causes a shift in demand from the middle of the day to late-afternoon when solar PV generation is at its minimum. Rechargeable battery technologies can provide firming capacity to reduce capacity requirements and smooth loads on electricity grids; however, significant costs are associated with their production, installation and disposal. Furthermore, battery production is struggling to keep up with total PV investment in the United States which is expected to more than double over the next five years. This paper reports on a research study conducted at a suburban shopping mall with a large PV array. Using machine learning techniques, we built forecasting models to predict both PV power generation and the buildings’ energy demand. These forecasts were embedded in automated daily messages to the operations manager (chief engineer) who used the information to anticipate electricity demand profiles and make air-conditioning (AC) control changes to shift peak load. We demonstrate that by utilizing thermal mass, the thermal capacity of chilled water systems, and an adaptive approach to human thermal comfort, the building’s peak demand was able to more closely align with PV output during hot weather conditions. Reductions in late afternoon peak demand of up to 15% were observed on extreme days. Smaller, yet significant, energy efficiency gains were also achieved.
It Takes A City: Harnessing the Demand Response Potential of Commercial Buildings
Cameron Roach and Craig Roussac
The potential to reduce peak electricity demand from commercial buildings by adopting demand response strategies is well documented. However, this potential is not realized because most building operators do not actively consider the impact of their buildings on the stability of the electricity supply system. This paper reports on a peak demand warning program undertaken by Buildings Alive that brought together 164 large office, retail and institutional buildings in major cities across Australia with the objective of coordinating their response to network capacity constraints. The program was run during the summer of 2019-20 – Australia’s second warmest on record. Adopting a forewarning methodology based on a ‘continuously-learning’ statistical model reported in the literature previously, a ‘peak demand warning’ email was sent to each building’s facility manager (or chief engineer) five days in advance of whenever demand was predicted to exceed a predefined threshold, or if the electricity network operator expected a capacity constraint. Each building brought its own demand reduction technology and strategies to the program. With the support of the forewarnings, information-sharing forums, and a summary message illustrating the differences between predicted and actual demand profiles issued on the morning following each peak event, building operators were able to experiment and refine strategies over the course of the summer. Normalized peak demand reductions in the range of 10-20% were recorded during extreme mid-afternoon peak conditions.
From local optima to overall optimum: a field study of model-based anomaly detection in commercial buildings
Hao Huang, Cameron Roach and Craig Roussac
Equipment faults and control issues cause energy waste, thermal discomfort, and drive up maintenance costs in commercial buildings. Automated fault detection and diagnostics (AFDD) tools help to systematically control these faults to achieve better occupant comfort and energy savings. However, AFDD tools are labor intensive to set up and often generate excessive numbers of false alarms causing alert fatigue and disengaging building operators. In this study, we present a novel model-based anomaly detection method developed with the objective of identifying the full extent of a building’s energy saving ‘potential’ and materially significant system anomalies, which, if addressed, can lead to optimum building performance. No engineering ‘rules’ were programmed into the system. Instead, the system acted as a search engine with machine learning algorithms finding patterns and correlations between multiple heterogeneous data sources, including weather data, interval metering data, and building management system (BMS) data. The outputs—estimates of energy performance improvement potential and specific recommendations to achieve it—were directly used by facility managers and technicians to inform their decisions. To evaluate the advantages and limitations of this ‘self-learning’ method, we field-tested in parallel with a rule-based fault detection platform across ten office buildings in Australia over a period of 18 months. Building engineers and facility managers were consulted throughout the study duration to evaluate the value of the information and their levels of engagement. The results show that, when compared to the rule-based method, the anomaly detection method identified more materially significant faults that impacted energy performance in the subject buildings, with fewer false alarms.