Model predictive control for thermal energy storage and thermal comfort optimization of building demand response in smart grids
Introduction
The rapid growth of power demand and the greater integration of renewable energy generations, which depend heavily on weather conditions, impose enormous stress on the balance of power grids [1]. Any power imbalance will cause severe consequences in the reliability and quality of power supply (e.g., voltage fluctuations and even power outrages). Facing the challenges for power balance, smart grid is considered as a promising solution to incorporate advanced technologies to offer better flexibility, reliability and security in grid operation [2]. The efforts conducted from the demand side to satisfy the grid requests (e.g., dynamic price and reliability information) is known as demand response (DR) [3]. DR programs cannot only benefit the operation of power grids but also offer economic benefits to end-users. Among the demand-side users of power grids, buildings consuming over 73.6% of overall electricity in the United States [4] and over 90% in Hong Kong [5] play an important role in DR programs. Moreover, with the help of advanced technologies such as building automation systems and smart meters, DR control strategies could be implemented conveniently in buildings to realize a bidirectional operation mode between buildings and power grids [6], [7].
When pricing changes or grid requests are informed day ahead or hours ahead, demand shifting by rescheduling the system operation is a preferable alternative for building demand response. Demand shifting is the process of shifting on-peak loads to off-peak hours so as to take advantage of electricity rate difference in different periods. Since air-conditioning systems in commercial buildings are the largest energy consumer [8], particularly in cooling dominant regions, the demand shifting control of air-conditioning systems is preferably adopted for optimizing the building power demand. Building thermal mass (i.e., passive storage) and thermal storage system (i.e., active storage) are two typical candidates to be used for building demand shifting during DR events and many control strategies for optimizing their cooling charging/discharging processes have been developed as different requirements considered for buildings or power grids [9], [10], [11], [12], [13]. Global indoor temperature adjustment plus precooling and cooling system adjustment are commonly adopted for the demand shifting and management in buildings. Xu and Haves [14] developed a simple demand-limiting strategy by resetting the indoor air temperature to utilize the building thermal mass for peak demand reduction in an office building in California. Yin et al. [15] proposed a control strategy “pre-cooling with exponential temperature set-up” to optimize the cooling charging/discharging processes of building thermal mass for DR control.
By contrast, when facing urgent requests and incentives from the smart girds, an immediate power reduction cannot be achieved within a very short time interval (i.e., minutes) by rescheduling the system operation (e.g., resetting the indoor air temperature), resulting from the inherent and significant delay of charging/discharging control processes [16], [17]. In such a case, shutting down part of operating chillers directly in a central air-conditioning system is considered as an effective proactive demand response strategy to achieve immediate power reduction within a very short time. Due to the effectiveness of this fast demand response strategy for the urgent requests of smart grids, many studies have been conducted. The authors of this paper [18] pointed out that imbalanced chilled water distribution in a central air-conditioning system occurred after simply shutting down some of operating chillers. A cooling distributor based on adaptive utility function was developed to solve this problem. They [19] also proposed a control concept (i.e., supply-based feedback control strategy) for such fast DR events, instead of conventional control strategy commonly used for central air-conditioning systems, to effectively avoid the serious operation problems (e.g., imbalanced cooling distribution) and ensure the expected immediate power reduction after shutting down part of operating chillers.
Compared with building thermal mass (i.e., passive storage), thermal storage system (i.e., active storage) has a larger capacity and better controllability in peak demand reduction responding to the requests of smart grids while has a less negative impact on building indoor environment during DR events. It reduces building peak demand contributed by the cooling system through the production and storage of cold energy during off-peak periods and the usage of the stored energy for cooling during peak periods. The cold energy storage in the central air-conditioning system is usually stored in the form of ice, chilled water, phase change materials (PCMs) or eutectic solution [20], [21]. Compared with the studies conducted for the optimal control of cold thermal storage during DR events (i.e., day ahead or hours ahead), the studies for the fast DR events just started in recent years. Xue et al. [22] developed a building thermal model to predict the discharged cooling from the building thermal mass after shutting down part of operating chillers in the fast DR event. Cui et al. [23] developed a design method to optimize the capacity of cold storage during a fast DR event by a quantitative analysis on its life-cycle cost-saving potential concerning the operational cost, initial investment and space cost. Cui et al. [24] also proposed a control strategy to optimize the cooling discharging rate of cold storage during a DR event to achieve an immediately stepped power demand reduction after shutting down part of operating chillers.
In addition, although the primary objective of DR events is to satisfy the request of smart grids (i.e., power reduction), the indoor thermal comfort may be potentially sacrificed to unacceptable levels due to the power limiting control of central air-conditioning systems and hence would be taken into consideration. Zhang et al. [25] investigated 56 subjects’ thermal comfort during DR events and pointed out that subjects’ thermal comfort zone during DR events was wider than that predicted by Fanger’s PMV/PPD model. Tang et al. [26] developed optimal and near-optimal control strategies to achieve a pre-determined power demand reduction and meanwhile ensure the indoor environment within the acceptable range in the DR events after shutting down part of operating chillers. In fact, during fast DR events, using active cold storage with proper control will be effective to increase the immediate power reduction for the power grid and meanwhile ensure the acceptable indoor environment. However, no study can be found in the literature in addressing the online optimal control issue for central air-conditioning systems integrated with active cold storages during fast DR events, considering the requirements at both supply and demand sides, i.e., power demand reduction (for smart grids) and indoor environment control (for buildings), simultaneously.
Model predictive control (MPC) is a simple yet effective approach for constrained control, which is able to predict the future behaviors of the controlled systems and to determine proper control actions by optimizing an objective function depending on the predictions over a given horizon subject to some constraints [27]. It uses a receding horizon (i.e., at each iteration, only the first step of the control strategy is implemented and then the control signal is calculated again) to enhance its robustness and control accuracy. MPC is now popularly adopted in the areas of built environment control and building demand management considering its obvious advantages [28], [29]. Due to the sudden changes caused by shutting down part of operating chillers at the start of the DR event, the control stability will be challenged and the control states in the air-conditioning system will experience a rather serious fluctuation before reaching a new control balance. Considering the advantage of MPC on the control robustness, this method would be preferable to be used for the optimal control issues of central air-conditioning systems during such fast DR events in the smart grids. Hence, it is necessary to use the MPC approach to address the optimal control problems of building fast DR.
In this study, the MPC approach is therefore adopted to optimize the control of a central air-conditioning system with active cold storage considering the expected building power reduction and the acceptable indoor environment during a fast DR event. The chiller power demand and the cooling discharging rate of cold storage are optimized online using the proposed MPC approach. The main contributions of this work include: (1) The online optimal control issue for the air-conditioning system with active cold storage during the fast DR event is effectively addressed, considering the requirements of power grid and building simultaneously. The power demand reduction is maximized as the expected profile pattern and meanwhile the indoor air temperature is maintained within a pre-determined acceptable range; (2) A linear state-space model together with a simple parameter identification method is developed for online prediction and optimization during the fast DR event, allowing the proposed control strategy computationally efficient; (3) The first-order exponential average method is used to handle the prediction errors caused by the model simplification and inaccuracy. Moreover, the MPC with shrunk prediction horizon, instead of a fixed width commonly used, is proposed to improve the control performance considering the characteristics of optimal control issues for fast DR events.
Section snippets
Schematic of proactive building demand response strategy for central air-conditioning systems with active cold storages during fast DR events
Generally, the total power demand in a commercial building can be classified into two parts: sheddable power demand and controllable power demand. The sheddable power demand, such as lighting, electrical equipment and lift, can be conveniently obtained based on the operation schedule. While the controllable power demand, such as air-conditioning systems, can be altered by power demand controls. The chillers always account for the largest power demand in a central air-conditioning system and
Mechanism of model predictive control
The basic mechanism of model predictive control (MPC) is to use a system model to predict the future evolution of the system performance using the predicted operating conditions over the prediction horizon. At each sampling interval, beginning at the current state, an optimization problem is formulated and solved over a finite horizon to achieve the expected system operation. The optimization result is a trajectory of future control signals into a system that satisfies the system dynamics and
Test platform
In this study, a virtual test platform is built to test the proposed MPC for optimizing the operation of a central air-conditioning system integrated with an active cold storage during a fast DR event. The dynamic models are developed on TRNSYS [35]. The models used are validated by real data [36]. The TRNSYS multi-zone building model (type 56) is used to simulate the building thermal behavior. The detailed physical models, building envelop and major components (e.g. chillers [36], pumps [26],
Validation of the dynamic building thermal model
The dynamic building thermal model was validated based on the simulation data, and the comparison between the predicted and actual indoor air temperatures is shown in Fig. 5. In order to quantify the deviation of the predicted data from the actual data, three indices, i.e., mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE), were used to evaluate the prediction performance and the results are listed in Table 2. Undoubtedly, the accuracy of dynamic
Conclusions
Shutting down part of operating chillers in a commercial building is an effective means to achieve an immediate power demand reduction responding to urgent requests of smart grids. The active thermal storage is necessary to be used for a further power reduction and meanwhile for ensuring an acceptable indoor environment when this proactive demand response (DR) method is adopted. In this study, an optimal control strategy using model predictive control (MPC) is developed to optimize the
Acknowledgements
The research presented in this paper is financially supported by a grant (152152/15E) of the Research Grant Council (RGC) of the Hong Kong SAR and a research grant under strategic focus area (SFA) scheme of the research institute of sustainable urban development (RISUD) in The Hong Kong Polytechnic University.
References (40)
- et al.
A review of the development of Smart Grid technologies
Renew Sustain Energy Rev
(2016) - et al.
A summary of demand response in electricity markets
Electr Power Syst Res
(2008) - et al.
A roadmap towards intelligent net zero-and positive-energy buildings
Sol Energy
(2011) - et al.
Optimal control strategy of central air-conditioning systems of buildings at morning start period for enhanced energy efficiency and peak demand limiting
Energy.
(2018) - et al.
A review on buildings energy consumption information
Energy Build
(2008) - et al.
An exploratory analysis of California residential customer response to critical peak pricing of electricity
Energy
(2007) - et al.
Load shifting using the heating and cooling system of an office building: quantitative potential evaluation for different flexibility and storage options
Appl Energy
(2017) - et al.
Peak load reductions: electric load shifting with mechanical pre-cooling of residential buildings with low thermal mass
Energy
(2015) - et al.
Quantifying flexibility of commercial and residential loads for demand response using setpoint changes
Appl Energy
(2016) - et al.
Study on Auto-DR and pre-cooling of commercial buildings with thermal mass in California
Energy Build
(2010)
A direct load control strategy of centralized air-conditioning systems for building fast demand response to urgent requests of smart grids
Autom Constr
Supply-based feedback control strategy of air-conditioning systems for direct load control of buildings responding to urgent requests of smart grids
Appl Energy
A review on cool thermal storage technologies and operating strategies
Renew Sustain Energy Rev
Energy performance and optimal control of air-conditioned buildings with envelopes enhanced by phase change materials
Energy Convers Manage
An interactive building power demand management strategy for facilitating smart grid optimization
Appl Energy
Life-cycle cost benefit analysis and optimal design of small scale active storage system for building demand limiting
Energy
Effectiveness and life-cycle cost-benefit analysis of active cold storages for building demand management for smart grid applications
Appl Energy
Thermal comfort during temperature cycles induced by direct load control strategies of peak electricity demand management
Build Environ
Optimal and near-optimal indoor temperature and humidity controls for direct load control and proactive building demand response towards smart grids
Autom Constr
Ten questions concerning model predictive control for energy efficient buildings
Build Environ
Cited by (125)
Quantifying the impact of building load forecasts on optimizing energy storage systems
2024, Energy and BuildingsData-driven aggregation of thermal dynamics within building virtual power plants
2024, Applied EnergyGrey-box and ANN-based building models for multistep-ahead prediction of indoor temperature to implement model predictive control
2023, Engineering Applications of Artificial Intelligence