(724d) A Supervisory Predictive Control System for Solar-Load Balancing: Application to Building Energy Management

Energy consumption is typically divided among four major sectors, including residential, commercial, industrial, and transportation sectors. Non-industrial energy usage comprises a significant part of the total energy budget in the United States. For example, in 2016 20% of total energy use was due to the residential sector, while another 18% of end use energy was by the commercial sector [1]. As a result, even small improvements in residential and commercial building energy efficiency, if widely adopted, hold the potential for significant impact. It is now largely recognized that addressing energy use in buildings can reduce total fossil fuel consumption and associated greenhouse gas emissions. Benefits such as reduced building operational energy costs have prompted growing interest among policy makers, the technical community, and the general public in addressing building energy issues and investigating solutions for decreasing building energy consumption (e.g., [2]-[9]). Reducing existing building energy consumption consists of two synergistic approaches. One is to reduce the need for energy through implementation of energy efficiency measures, and the other is to offset the remaining building energy needs through use of on-site renewable energy systems. The latter approach also holds the promise of reducing transmission and distribution losses associated with the electricity generated by power plants.

In practice, the most easily accessible and logistically feasible renewable energy source for buildings is rooftop solar. Consequently, local solar deployment at both residential and commercial sites has grown in recent years as local users seek to save on their electric utility bills while contributing less to non-renewable energy sources. This increased renewable energy penetration into the existing grid infrastructure is necessary for sustainability reasons, although with it new problems have emerged. For the budding years of residential and commercial on-site solar generation the connection with the grid was assumed to give the user access to an infinite entity where any amount of electricity could be sold to or bought from at any time. While on small scales this is accurate, as any excess or shortage a single user would reasonably generate could easily be handled by the grid, when this paradigm is expanded to all grid users this treatment of the grid connection is no longer a viable option.

The current usage of this technology is that when on-site solar generation outpaces local demand (or when it is financially advantageous to sell power due to time of use), the user could sell electricity back to the grid, and when the reverse is true electricity could be bought. While this proves an attractive option - as it provides additional source of income for users installing on-site generation capabilities - it is not deployable at larger scales. One can see how if a substantial amount of users are trying to sell electricity at similar times then this is problematic from a supply demand standpoint.

As more solar generation from these users enters a geographic area, this typically leads to a reduction in demand from the generation side during daylight hours. However, due to the nature of solar generation, the availability is not consistent. Most buildings with on-site generation do not take into account the varying solar load when implementing building controls. For commercial users this is especially problematic due to a demand charge being implemented for higher use electric bills, where a fee is paid for each kilowatt-hour used in the peak fifteen minute usage period over the past billing cycle. It can be imagined that if a cloud passes over the solar array of a building this could sharply increase the electricity needed from the grid, potentially resulting in the monthly peak demand. The result of this oversight is that the demand peaks for such buildings can be unnecessarily high. This is non-ideal for both the commercial grid user and the grid itself, where the grid user pays a higher bill and the grid has to supply high loads for short periods of time which is cumbersome from an operational standpoint.

Motivated by these considerations, we focus in this work on the scalability problem from a local consumer user standpoint. The reasoning being that it would be advantageous for each building on the grid to act in such a way to minimize its own grid impact independent of other grid users. This would allow a larger number of local users to integrate local cost saving renewable energy, while at the same time reducing the collective demand on the grid, all without the need for any inter-user communication or coordination. To this end, an optimization-based supervisory predictive control system is developed to leverage flexible building loads throughout the day while taking into account solar forecasting to reduce the cost of electricity to the user, as well as ease the load to the grid of peaks resulting from the on-site solar generation. From the user standpoint, this will result in a monthly reduction in demand charges as well as a reduction in electricity cost with solar installation, and in turn the grid supplies a more consistent load to the user resulting in longevity of equipment and higher operational efficiency resulting in a win-win scenario.

The developed supervisory model predictive control (MPC) system is tested on a simulated three-story 150,000 square foot office building located in Merced, California, with dimmable lights, rooftop fixed-tilt solar, electric vehicle charging stations, HVAC system, and simulated miscellaneous loads. The predictive controller is implemented on the building with various objective functions in an attempt to find a computationally agreeable function that reasonably minimizes the grid impact from a practical point of view. Three functions are selected for the test cases to minimize over the MPC horizon, including the net building load, the peak building load, and a weighted sum of both the peak and net building loads. Case studies are carried out with both perfect and imperfect forecasting of the available solar energy. For the perfect forecasting case, historical data of solar irradiance are used for both the building solar generation and the MPC solar availability forecast. In the imperfect forecasting case, the historical solar data are used for the building solar generation while a clear sky model is used for the MPC solar availability forecast. The clear sky model data are corrected in real simulation time via a moving-horizon estimation scheme that uses the historical data over a past horizon to estimate a solar ``fault" value to correct the clear sky model output over some future horizon [10]. The building energy and demand costs are compared between the cases with both perfect and imperfect forecasting, and with reference to a rules-based base case for context. Through deployment of a framework for building energy management taking into account the grid impact and cost from the local user's standpoint, this paves the way for a more substantial utilization of solar energy on a large scale.

References:

[1] “Energy information administration energy review 2016.” https://www.eia.gov/consumption/.

[2] F. Oldewurtel, A. Parisio, C. N. Jones, D. Gyalistras, M. Gwerder, V. Stauch, B. Lehmann, and M. Morari, “Use of model predictive control and weather forecasts for energy efficient building climate control,” Energy and Buildings, vol. 45, pp. 15 – 27, 2012.

[3] T. Salsbury, P. Mhaskar and S. Z. Qin, “Predictive control methods to improve energy efficiency and reduce demand in buildings,” Comp. & Chem. Eng., vol. 51, pp. 77-85, 2013.

[4] V. M. Zavala, “Real-time optimization strategies for building systems,” Industrial & Engineering Chemistry Research, vol. 52, no. 9, pp. 3137–3150, 2013.

[5] P. Zhao, S. Suryanarayanan, and M. G. Simoes, “An energy management system for building structures using a multi-agent decision-making control methodology,” IEEE Transactions on Industry Applications, vol. 49, pp. 322-330, 2013.

[6] Skogestad, S. and P. Mhaskar (Eds.), Energy Efficient Buildings, Special Issue of Journal of Process Control, vol. 24, 2014.

[7] P. H. Shaikh, N. B. M. Nor, P. Nallagownden, I. Elamvazuthi, and T. Ibrahim, “A review on optimized control systems for building energy and comfort management of smart sustainable buildings,” Renewable and Sustainable Energy Reviews, vol. 34, pp. 409 – 429, 2014.

[8] Y. M. Lee, R. Horesh, and L. Liberti, “Simulation and optimization of energy efficient operation of HVAC system as demand response with distributed energy resources,” in 2015 Winter Simulation Conference (WSC), pp. 991–999, Dec 2015.

[9] X. Wang, A. Palazoglu, and N. H. El-Farra, “Operational optimization and demand response of hybrid renewable energy systems,” Applied Energy, vol. 143, pp. 324-335, 2015.

[10] J. T. Allen and N. H. El-Farra, “A model-based framework for fault estimation and accommodation applied to distributed energy resources,” Renewable Energy, vol. 100, pp. 35-43, 2017.