動的 CO2 排出係数に基づいたモデル予測制御による熱源機器の低炭素制御

宮田 翔平, 桑原 康浩, 林 鍾衍, 赤司 泰義, 吉本 尚起
日本建築学会環境系論文集, 2020, 85 巻, 777 号, pp. 827-835
https://doi.org/10.3130/aije.85.827

Abstract
To reduce carbon dioxide (CO₂) emissions, interest in renewable energy, such as solar and wind power, is gradually growing. However, there is an imbalance between the supply and demand of such electric powers. Demand Response (DR) has been the subject of many research and demonstrations as a control measure aimed at power load leveling. In this study, we propose a carbon activated demand response (CADR) that considers a dynamic CO₂ emission factor that is estimated hourly. We investigate the control methods used to minimize carbon emissions from the central cooling plant which has two centrifugal chillers. Based on the estimated dynamic CO₂ emission factor that fluctuates every hour, the operation of the chillers with water heat storage tanks as an energy buffer was optimized. The novelty of this study lies in the proposition of the dynamic CO₂ emission factor and the quantification of the low-carbon effect when the cooling plant with water heat storage tank is optimally controlled based on the dynamic CO₂ emission factor.
The CO₂ emission factor is defined as the ratio of the amount of CO₂ emitted from power generation by the amount of power generation [kg- CO₂ / kWh]. Since data on hourly power generation was available, we used the hourly emission factor for estimating the dynamic CO₂ emission factor in our study.
The cooling plant of a large office building, with a water heat storage tank (total 6,704 m³) and three refrigerators (total 2,470 Rt), was the subject in this study. The control behavior of the cooling plant was simulated to investigate low carbon control. Conventional control starts heat storage at 22:00 or 8:00 and releases heat according to the load. In contrast, the proposed CADR uses Model Predictive Control (MPC) that optimizes the operation of the chillers every two hours. Based on the dynamic CO₂ emission factor and the power consumption predicted by the simulation, the combination of the start and stop of the chillers that minimizes the CO₂ emissions for the next 48 hours is calculated. Since load prediction is required for power consumption prediction, cases where load prediction has errors, were also included. In this study, we compared the performance of CADR with the conventional method for one week, from 0:00 of June 4, 2017, to 23:59 of June 10.
The results show that the CADR was able to reduce CO₂ emissions by 46.77% and 12.89% compared to the conventional method that starts heat storage at 22:00 and 8:00, respectively. In conventional control, the heat is regularly stored and full at the end of heat storage, but in CADR, heat is stored when the emission factor is low. In addition, this method was effective in reducing CO₂ emissions, even with a load prediction error of 10%. The future research focus will be to calibrate the equipment model for MPC, expand the type of energy buffer such as batteries, and analyze the contribution of the proposed method to the power systems.