Search

Two simulation methods were developed to model the lighting energy use for a real office building and the results of those simulations were compared to submetered lighting energy data from that building. Three deterministic schedules and six stochastic first-order Markov-models were used to assess annual lighting energy density and peak demand density at both the office-level and the building-level. The stochastic simulations, which apply only to the private offices, resulted in much lower annual energy density and peak demand density predictions due to interactions of users with the lighting and shade controls, and widely different daily use profiles. For the whole-building simulations, comparing all three deterministic schedules resulted in an inverse relationship between peak demand density and annual energy density; the closest correlation with the submetered data was found through the most building-specific but least empirically-vetted schedule-based simulation. Keywords--Energy modeling, annual energy use, peak demand, whole-building simulation., 1 INTRODUCTION AND BACKGROUND According to the results of the 2003 Commercial Buildings Energy Consumption Survey [2008], the most recent survey for which statistically reliable data is available, lighting was [...]

Reducing energy consumption and greenhouse gas emissions in the built environment is a critical step in achieving emission goals to mitigate climate change impacts. Local, federal, and international jurisdictions are deploying several methods to reduce energy and emissions such as voluntary and mandatory benchmarking and building performance standards, requiring building owners to reach energy and emission targets.

Large commercial buildings may display demand flexibility, which reduces electric energy expenses for the building owner and carbon emissions from grid operations, provides distributed energy resources, and increases the penetration of renewable energy sources. Demand-controlled ventilation (DCV) and building thermal mass control can individually and jointly provide such flexibility. The performance and financial payback of these technology options can be dramatically improved if based on hourly electric prices and carbon emissions rates. In this study, a modeled but actual large office building, simulated using New York City hourly electric prices, hourly CO2e emissions rates, and weather data for the summer 2019 cooling season is based on these dynamic driving parameters. A joint optimization of a building’s thermal mass and indoor CO2 content is presented. Superior energy savings and carbon emissions reductions are found for the joint optimization scenario when compared to both the baseline operation and individual optimization of building thermal mass and indoor CO2 content. These findings motivate the development of a real-time joint control system that utilizes closed-loop model predictive control (MPC) to optimally harness both sources of demand flexibility, a system that would require the future development of forecasting algorithms for external and control-oriented system models.

The incorporation of emerging technologies, including solar photovoltaics, electric vehicles, battery energy storage, smart devices, Internet-of-Things devices, and sensors in buildings, desirable control objectives are becoming increasingly complex, calling for advanced controls approaches. Reinforcement learning (RL) is a powerful method for this. RL can adapt and learn from environmental interaction, but it can take a long time to learn and can be unstable initially due to limited environmental knowledge. In our research, we propose an online RL approach for buildings that uses data-driven surrogate models to guide the RL agent during its early training. This helps the controller learn faster and more stably than the traditional direct plug-and-learn online learning approach. In this research, we propose an online approach in buildings with RL where, with the help of data-driven surrogate models, the RL agent is guided during its early exploratory training stage, aiding the controller to learn a near-optimal policy faster and exhibiting more stable training progress than a traditional direct plug-and-learn online learning RL approach. The agents are assisted in their learning and action with information gained from the surrogate models generating multiple artificial trajectories starting from the current state. The research presented an exploration of various surrogate model-assisted training methods and revealed that models focusing on artificial trajectories around rule-based controls yielded the most stable performance. In contrast, models employing random exploration with a one-step look-ahead approach demonstrated superior overall performance.

Clustering is an unsupervised learning technique that is useful when working with a large volume of unlabeled data. Complex dynamical systems in real life often entail data streaming from a large number of sources. Although it is desirable to use all source variables to form accurate state estimates, it is often impractical due to large computational power requirements, and sufficiently robust algorithms to handle these cases are not common. We propose a hierarchical time series clustering technique based on symbolic dynamic filtering and Granger causality, which serves as a dimensionality reduction and noise-rejection tool. Our process forms a hierarchy of variables in the multivariate time series with clustering of relevant variables at each level, thus separating out noise and less relevant variables. A new distance metric based on Granger causality is proposed and used for the time series clustering, as well as validated on empirical data sets. Experimental results from occupancy detection and building temperature estimation tasks shows fidelity to the empirical data sets while maintaining state-prediction accuracy with substantially reduced data dimensionality.

Reinforcement learning (RL) has shown significant success in sequential decision making in fields like autonomous vehicles, robotics, marketing and gaming industries. This success has attracted the attention to the RL control approach for building energy systems which are becoming complicated due to the need to optimize for multiple, potentially conflicting, goals like occupant comfort, energy use and grid interactivity. However, for real world applications, RL has several drawbacks like requiring large training data and time, and unstable control behavior during the early exploration process making it infeasible for an application directly to building control tasks. To address these issues, an imitation learning approach is utilized herein where the RL agents starts with a policy transferred from accepted rule based policies and heuristic policies. This approach is successful in reducing the training time, preventing the unstable early exploration behavior and improving upon an accepted rule-based policy — all of these make RL a more practical control approach for real world applications in the domain of building controls.

For buildings designed to meet aggressive energy goals, there is a need for tools to assist in the monitoring and maintenance of performance once the building is in operation. In particular, dashboard visualizations that show real-time and historic end use energy consumption alongside expected performance are powerful tools for both occupant engagement and the identification of operational issues. This article focuses on two related approaches to calculating upper and lower control limits for acceptable ranges of end use, which use a combination of modeled and measured usage data to generate realistic energy-conservative control limits. The first approach centers on the analysis of frequency distributions for end use consumption as functions of a main effect variable, while the second approach uses multivariate quantile regression based on principal components to generate control limits from all available measured variables.

AbstractA pharmaceutical facility located in Southern Germany is experiencing a trend of growing cooling loads to be met by the chilled water plant composed of ten chillers of greatly varying cost effectiveness. With a capacity shortfall inevitable, the question arises whether to install an additional chiller or improve the utilization of the existing chillers, in particular those with low operating costs per unit cooling, through the addition of a chilled water thermal energy storage (TES) system. To provide decision support in this matter, an optimization environment was developed and validated that adopts mixed integer programming as the approach to optimizing the chiller dispatch for any load condition, while an overarching dynamic programming based optimization approach optimizes the charge/discharge strategy of the TES system. In this fashion, the chilled water plant optimization is decoupled but embedded in the TES control optimization. The approach was selected to allow for arbitrary constraints and optimization horizons, while ensuring a global optimum to the problem. The results show that a relatively small TES tank provides significant economic and operational benefits. Yet, in order to facilitate long-term supply security, a larger TES tank capacity was decided on and the TES system was constructed in 2008.

This paper describes a simulation-based investigation of machine-learning control for the supervisory control of building energy systems. Model-free reinforcement learning control is investigated for the operation of electrically driven cool thermal energy storage systems in commercial buildings. The reinforcement learning controller learns to charge and discharge a thermal storage tank based on the feedback it receives from past control actions. The learning agent interacts with its environment by commanding the thermal energy storage system and extracts cues about the environment solely based on the reinforcement feedback it receives, which in this study is the monetary cost of each control action. No prediction or system model is required. Over time and by exploring the environment, the reinforcement learning controller establishes a statistical summary of plant operation, which is continuously updated as operation continues. The controller learns to account for the time-dependent cost of electricity (both time-of-use and real-time pricing), the availability of thermal storage, part-load performance of the central chilled water plant, and weather conditions. Though reinforcement learning control proved sensitive to the selection of state variables, level of discretization, and learning rate, it effectively learns a difficult task of controlling thermal energy storage and displays good performance. The cost savings compare favorably with conventional cool storage control strategies but do not reach the level of predictive optimal control.

This paper describes the development of a simulation environment for the evaluation of the performance of various controls of a set of common ice storage systems. The cooling plant used for this evaluation consisted of one chiller for icemaking and chilled-water modes, an ice storage, an air-handling unit, a condenser, and all required fans and pumps. The plant model allows the selection of one of three ice storage systems as well as one of three compressor types. Components are modeled using polynomial fits to manufacturers' data. Using dimensionless parameters, components can be scaled to arrive at a proper cooling plant design. The performance of a selected cooling plant configuration can be assessed for three conventional control strategies (chiller-priority, constant-proportion, and storage-priority control) and compared to optimal control that achieves the theoretical maximum of operating cost savings. Optimal control is found using a dynamic programming based global search in the domain of on-peak and off-peak demands. Mixed integer programming is used for validation using a linear plant model. A set of artificial rate structures with marked characteristics is applied to generate results that allow the discussion of fundamental features of cool storage control. The effect of non-cooling electrical loads on optimal control is illustrated.

This paper describes the development and simulation of a predictive optimal controller for thermal energy storage systems. The “optimal” strategy minimizes the cost of operating the cooling plant over the simulation horizon. The particular case of a popular ice storage system (ice-on-coil with internal melt) was investigated in a simulation environment. Various predictor models were analyzed with respect to their performance in forecasting cooling load data and information on ambient conditions (dry-bulb and wet-bulb temperatures). The predictor model provides load and weather information to the optimal controller in discrete time steps. An optimal storage charging and discharging strategy was planned at every time step over a fixed look-ahead time window utilizing newly available information. The first action of the optimal sequence of actions was executed over the next time step and the planning process was repeated at every following time step. The effect of the length of the planning horizon was investigated. It was found that short horizons of one day were only marginally suboptimal relative to one that is optimal over the entire simulation horizon of one week. Various utility rate structures were analyzed to cover a range of potential real-time pricing scenarios. The predictive optimal controller was then compared to conventional control heuristics. In summary, it was found that in the presence of complex rate structures the proposed controller had a significant performance benefit over the conventional controls, while requiring only a simple predictor.

Recent efforts to reduce energy consumption and greenhouse gas emissions have resulted in the development of sustainable, smart districts with highly energy efficient buildings, renewable distributed energy resources (DERs), and support for alternative modes of transportation. However, there is typically little if any coordination between the district developers and the local utility. Most attention is paid to the district's annual net load and generation without considering their instantaneous imbalance or the connecting network's state. This presents an opportunity to learn lessons from the design of distribution feeders for districts characterized by low loads and high penetrations of DERs that can be applied to the distribution grid at large. The aim of this overview is to summarize current practices in sustainable district planning as well as advances in modeling and design tools for incorporating the power distribution system into the district planning process. Recent developments in the modeling and optimization of district power systems, including their coordination with multi‐energy systems and the impact of high penetration levels of renewable energy, are introduced. Sustainable districts in England and Japan are reviewed as case studies to illustrate the extent to which distribution system planning has been considered in practice. Finally, newly developed building‐to‐grid modeling tools that can facilitate coordinated district and power system design with utility involvement are introduced, along with suggestions for future research directions. This article is categorized under:Wind Power > Systems and InfrastructureEnergy Policy and Planning > Systems and InfrastructureEnergy Efficiency > Systems and Infrastructure Engaging the local utility and including electrical distribution system design in district master planning can reduce costs and improve performance of sustainable urban districts with high penetrations of renewable energy, but new research tools are needed to facilitate an integrated design. Source: fujisawasst.com/EN/.