Search

Using a simulation and optimization environment, this paper presents advances toward near-optimal building thermal mass control derived from full factorial analyses of the important parameters influencing the passive thermal storage process for a range of buildings and climate/utility rate structure combinations. Guidelines for the application of, and expected savings from, building thermal mass control strategies that can be easily implemented and result in a significant reduction in building operating costs and peak electrical demand are sought. In response to the actual utility rates imposed in the investigated cities, fundamental insights and control simplifications are derived from those buildings deemed suitable candidates. The near-optimal strategies are derived from the optimal control trajectory, consisting of four variables, and then tested for effectiveness and validated with respect to uncertainty regarding building parameters and climate variations. Due to the overriding impact of the utility rate structure on both savings and control strategy, combined with the overwhelming diversity of utility rates offered to commercial building customers, this study cannot offer universally valid control guidelines. Nevertheless, a significant number of cases, i.e., combinations of buildings, weather, and utility rate structure, have been investigated, which offer both insights and recommendations for simplified control strategies. These guidelines represent a good starting point for experimentation with building thermal mass control for a substantial range of building types, equipments, climates, and utility rates. [DOI: 10.1115/1.4001466]

In order to avoid high utility demand charges from cooling during the summer and to level a building's electrical demand profile, precooling of the building's massive structure can be applied to shift cooling-related thermal loads in response to utility pricing signals. Several previous simulation and experimental studies have shown that proper precooling can attain considerable reduction of operating cost in buildings. This paper systematically evaluates the merits of the passive building thermal capacitance to minimize energy cost for a design day using optimal control. The evaluation is conducted by means of a sensitivity analysis utilizing a dynamic building energy simulation program coupled to a popular technical computing environment. The optimal controller predicts the required extent of precooling (zone temperature set-point depression), depending on the utility rate structure, occupancy and on-peak period duration and onset, internal gains, building mass, occupancy period temperature set-point range, and weather as characterized by diurnal temperature and relative humidity swings. In addition to quantifying the building response, energy consumption, and utility cost, this paper extracts the dominant features of the optimal precooling strategies for each of the investigated cases so that guidelines for near-optimal building thermal mass savings may be developed in the future. [DOI: 10.1115/1.2770755]

This paper describes an investigation of machine learning for supervisory control of active and passive thermal storage capacity in buildings. Previous studies show that the utilization of active or passive thermal storage, or both, can yield significant peak cooling load reduction and associated electrical demand and operational cost savings. In this study, a model-free learning control is investigated for the operation of electrically driven chilled water systems in heavy-mass commercial buildings. The reinforcement learning controller learns to operate the building and cooling plant based on the reinforcement feedback (monetary cost of each action, in this study) it receives for past control actions. The learning agent interacts with its environment by commanding the global zone temperature setpoints and thermal energy storage charging/discharging rate. The controller extracts information about the environment based solely on the reinforcement signal; the controller does not contain a predictive or system model. 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 present analysis shows that learning control is a feasible methodology to find a near-optimal control strategy for exploiting the active and passive building thermal storage capacity, and also shows that the learning performance is affected by the dimensionality of the action and state space, the learning rate and several other factors. It is found that it takes a long time to learn control strategies for tasks associated with large state and action spaces. [DOI: 10.1115/1.271049|]

In contrast to building energy conversion equipment, less improvement has been achieved in thermal energy distribution, storage and control systems in terms of energy efficiency and peak load reduction potential. Cooling of commercial buildings contributes significantly to the peak demand placed on an electrical utility grid and time-of-use electricity rates are designed to encourage shifting of electrical loads to off-peak periods at night and on weekends. Buildings can respond to these pricing signals by shifting cooling-related thermal loads either by precooling the building's massive structure (passive storage) or by using active thermal energy storage systems such as ice storage. Recent theoretical and experimental work showed that the simultaneous utilization of active and passive building thermal storage inventory can save significant amounts of utility costs to the building operator, yet increased electrical energy consumption may result. The article investigates the relationship between cost savings and energy consumption associated with conventional control, minimal cost and minimal energy control, while accounting for variations in fan power consumption, chiller capacity, chiller coefficient-of-performance, and part-load performance. The model-based predictive building controller is employed to either minimize electricity cost including a target demand charge or electrical energy consumption. This work shows that buildings can be operated in a demand-responsive fashion to substantially reduce utility costs with marginal increases in overall energy consumption. In the case of energy optimal control, the reference control was replicated, i.e., if only energy consumption is of concern, neither active nor passive building thermal storage should be utilized. On the other hand, cost optimal control suggests strongly utilizing both thermal storage inventories. [DOI: 10.1115/1.1877513]

Cooling of commercial buildings contributes significantly to the peak demand placed on an electrical utility grid. Time-of-use electricity rates encourage shifting of electrical loads to off-peak periods at night and on weekends. Buildings can respond to these pricing signals by shifting cooling-related thermal loads either by precooling the building's massive structure or by using active thermal energy storage systems such as ice storage. While these two thermal batteries have been engaged separately in the past, this paper investigates the merits of harnessing both storage media concurrently in the context of optimal control for a range of selected parameters. A parametric' analysis was conducted utilizing an EnergyPlus-based simulation environment to assess the effects of building mass, electrical utility rates, season and location, economizer operation, central plant size, and thermal comfort. The findings reveal that the cooling-related on-peak electrical demand and utility cost of commercial buildings can be substantially reduced by harnessing both thermal storage inventories using optimal control for a wide range of conditions. [DOI: 10.1115/1.1824110]

It has been shown that a neural network with sufficient hidden units can approximate any continuous function defined on a closed and bounded set. This has inspired the use of neural networks as general nonlinear regression models. As with other nonlinear regression models, tools of conventional statistical analysis can be applied to neural networks to yield a test for the relevance or irrelevance of a free parameter. The test, a version of Wald's test, can be extended to yield a test for the relevance or irrelevance of an input variable. This test was applied to the building energy use data of the Energy Prediction Shootout II contest. Input variables were selected by initially constructing a neural network model which had many inputs, then cutting out the inputs which were deemed irrelevant on the basis of the Wald test. Time-lagged values were included for some input variables, with the time lag chosen by inspecting the autocovariance function of the candidate variable. The results of the contest entry are summarized, and the benefits of applying Wald's test to this problem are assessed.

This paper surveys past and current research of optimal control for central chilled water plants with ice thermal energy storage. The motivation for thermal energy storage in commercial buildings is provided and common operating strategies for ice storage including their implementation are presented. The concept of optimality serves as the basis for introducing the various approaches to optimal control of thermal energy storage. Optimal strategies minimizing either energy or demand costs, near-optimal rule based control minimizing total cost, comfort-based energy optimal control, and combined optimal sizing and energy cost optimal control are discussed and contrasted with mathematically non-optimal, but heuristically improved operating strategies. Fully optimal control based on perfect knowledge is introduced and subsequent developments of predictive optimal control subject to uncertain weather, load, and utility rate information illustrated. In addition, recent investigations of adaptive optimal reinforcement learning based control are presented. [DOI: 10.1115/1.1591801]

This paper investigates adaptive optimal control of a grid-independent photovoltaic system consisting of a collector, storage, and a load. The control algorithm is based on Q-Learning, a model-free reinforcement learning algorithm, which optimizes control performance through exploration. Q-Learning is used in a simulation study to find a policy which performs better than a conventional control strategy with respect to a cost function which places more weight on meeting a critical base load than on those non-critical loads exceeding the base load. [DOI: 10.1115/1.1532005]

A simplified ice storage system model was developed in which the icemaking mode is reflected by a higher power consumption per unit cooling than in chilled-water mode. The performance off our control strategies for ice storage systems is evaluated. The four control strategies investigated are chiller-priority and constant-proportion as conventional, instantaneous controls, while storage-priority and optimal control represent sophisticated controls employing load forecasting. Six parameters were investigated with respect to their influence on the ice storage system performance: Storage losses, utility rate structures, rate periods, penalty for icemaking, storage capacity, and the impact of load forecasting. Optimal control was determined to provide maximal operating cost savings. The storage-priority control yields operating costs only slightly higher than those of optimal control. Chiller-priority control realized savings that were typically on the order of 50% of what is theoretically possible (optimal control). Constant-proportion control proved to be a simple control strategy yielding higher savings than chiller-priority, yet lower than storage-priority control. [DOI: 10.1115/1.1530629]

Abstract Self-consumption is an increasingly economically attractive solution for electricity generated by distributed energy resources, which also reduces transmission losses, power line overload and grid instability. Yet, at single-building scale, production and demand do often not coincide. Microgrids are proposed as a means to trade surplus energy produced with local prosumers and consumers, to take advantage of complementary load profiles. Here we define a consumer as a single metered residential unit, a prosumer as a consumer with a photovoltaic system, which first uses the on-site energy generated prior to exporting the surplus, and the microgrid as a clearly defined number of both prosumers and consumers. In this article, we systematically analyze to what extent varying microgrid sizes and prosumer-consumer ratios affect local self-consumption and self-sufficiency rates. To that end, we developed a simulation model that uses real-world load profiles from 4190 residential buildings as input data. We find that the prosumer-to-consumer ratio is more important than the absolute microgrid size, for microgrids sizes greater than ten. The results also indicate that prosumer-to-consumer ratios in the range of 40%-60% have the best performance. Each simulation is also compared to the baseline scenario of a stand-alone prosumer, which shows significantly better self-consumption ratios and self-sufficiency ratios for microgrids due to aggregation effects. Finally, this work may also be used as a reference to design residential microgrid communities for various prosumer-consumer compositions and various production-to-demand ratios. [ABSTRACT FROM AUTHOR]

Energy modelling for the prediction of energy use in buildings, especially under novel energy management strategies, is of great importance. In buildings there are several flexible electrical loads which can be shifted in time such as thermostatically controllable loads. The main novelty of this paper is to apply an aggregation method to effectively characterize the electrical energy demand of air-conditioning (AC) systems in residential buildings under flexible operation during demand response and demand shaping programs. The method is based on clustering techniques to aggregate a large and diverse building stock of residential buildings to a smaller, representative ensemble of buildings. The methodology is tested against a detailed simulation model of building stocks in Houston, New York and Los Angeles. Results show good agreement between the energy demand predicted by the aggregated model and by the full model during normal operation (normalized mean absolute error, NMAE, below 10%), even with a small number of clusters (sample size of 1%). During flexible operation, the NMAE rises (around 20%) and a higher number of representative buildings become necessary (sample size at least 10%). Multiple cases for the input data series were considered, namely by varying the time resolution of the input data and the type of input data. These characteristics of the input time series data are shown to play a crucial role in the aggregation performance. The aggregated model showed lower NMAE compared to the original model when clustering is based on a hybrid signal resolved at 60-minute time intervals, which is a combination of the electricity demand profile and AC modulation level. [ABSTRACT FROM AUTHOR]

Recent advances in information and communications technology (ICT) have initiated development of a smart electrical grid and smart buildings. Buildings consume a large portion of the total electricity production worldwide, and to fully develop a smart grid they must be integrated with that grid. Buildings can now be ‘prosumers’ on the grid (both producers and consumers), and the continued growth of distributed renewable energy generation is raising new challenges in terms of grid stability over various time scales. Buildings can contribute to grid stability by managing their overall electrical demand in response to current conditions. Facility managers must balance demand response requests by grid operators with energy needed to maintain smooth building operations. For example, maintaining thermal comfort within an occupied building requires energy and, thus an optimized solution balancing energy use with indoor environmental quality (adequate thermal comfort, lighting, etc.) is needed. Successful integration of buildings and their systems with the grid also requires interoperable data exchange. However, the adoption and integration of newer control and communication technologies into buildings can be problematic with older legacy HVAC and building control systems. Public policy and economic structures have not kept up with the technical developments that have given rise to the budding smart grid, and further developments are needed in both technical and non-technical areas. [ABSTRACT FROM AUTHOR]