This work proposes a framework for designing resilient RES for buildings. Buildings are major consumers of energy and, specifically, electricity. Globally, buildings represent over one third of global energy consumption and emissions [1]. In the United States, they represent 40% of energy use and 75% of electricity consumption [4]. In Saudi Arabia, buildings consume approximately 80% of the total electricity generated, with 70% used for cooling [5]. The energy consumption per square meter of a building depends on its usage profile (e.g., residential, office building, data center), building envelopes, efficiency policies, and weather conditions. Zero-energy buildings [3] are, in general, efficient buildings that generate and store electricity to supply their own load demand and potentially export to the grid. The supply side has two main components: 1) the renewable generation technologies; and 2) the battery electrical storage systems. On the building side the characterization of the load demand is paramount to design generation and storage systems that meet the load demand profiles.
Weather conditions have a double impact on net-zero buildings. On the one hand, they influence the energy consumption used for space cooling or heating. On the other hand, the weather strongly impacts the power output of RES. Therefore, for a given building in a specific location, the weather input used in the RES design is paramount to obtaining a resilient system that can cope with extreme weather events.
The proposed methodology focuses on designing resilient RES for buildings considering the uncertainty of climate evolution due to uncertainty of socio-economic pathways, model uncertainty, and internal variability [6]. The methodology involves: i) selecting appropriate climate models for the region of interest and downscaling their coarse results to high-resolution weather data; ii) developing a building simulation to predict the electricity demand (cooling, lighting, appliances) as a function of the weather input; iii) a RES simulation integrated with the load demand forecast; iv) a simulation-based multi-objective optimization to determine the best system configuration; and v) a method to incorporate extreme weather events from various climate models into the optimization to enhance system resilience.
The building simulation and RES modeling were performed in the simulation software TRNSYS, which is interfaced with a derivative-free optimization library. The original optimization problem renders a two-stage stochastic program to minimize the RES’ life cycle cost (LCC), where the weather parameters and load demand are the discrete random parameters. The first-stage variables are the capacities of a photovoltaic system, wind turbines, and storage, while the second-stage variables consist of the battery operations and the dump of excess electricity. In this setup, each climate projection (covering the same time horizon) is a scenario in the stochastic program. However, to formulate a stochastic program using TRNSYS, we reformulate the optimization problem from an L-Shape structure, where blocks of variables and constraints for each scenario occur in parallel in the structure, to a structure with the scenarios as a sequence. This re-organization ensures that all scenarios share the same first-stage variable values, at the cost of considering additional time-linking constraints for battery storage balances between consecutive years, which are not considered in the original stochastic program.
Given the potential evolution of the climate conditions, we adopt a rolling horizon approach, where we design a system for 10 years of weather and load demand scenarios and evaluate the design for the following 15 years. If the design obtained in the first step does not lead to unserved energy in the following 15 years, then the initial design is selected. Otherwise, additional capacities are determined.
Incorporating climate scenarios covering periods of 25 years from multiple models results in extremely large stochastic programs. Therefore, we propose a method to generate first-stage solutions that are feasible over several climate models and their scenarios. The method follows the following steps:
1. Initial stochastic program: solve a reformulated stochastic program using the hourly weather dataset from the four scenarios of the MPI-ESM-HR climate model for 25 years (using the multistage investment method). A RES design is obtained for that weather dataset.
2. Post-optimization operation simulation: simulate the RES design obtained in step 1 using the scenarios of other climate models, such as the BCC-CSM2-MR climate model.
3. Identification of unserved energy periods: find the hours/days when the RES fails to supply the total electricity demand, necessitating electricity from an external grid. Arrange these periods based on the highest cumulative value of electricity required from the external grid for successive hours.
4. Identification of extreme weather events: classify as extremes the days preceding the grid-reliant periods, which resulted in inadequate power generation and insufficient charging of the storage.
5. Weather data substitution: replace the original weather data of the MPI model with the calendar days corresponding to the selected extreme days. Re-optimization: rerun the optimization using the modified weather dataset. A new RES design solution is obtained.
6. Re-simulation: Simulate the operation of the new RES design using the modified weather data across the 25-year weather datasets from all scenarios of the BCC-CSM2-MR climate model.
7. Iterative procedure: if the RES is feasible over the whole hourly 25-year weather dataset of all the scenarios, then the procedure stops; otherwise, return to step 3.
8. Cross-model simulation: using the newly feasible solution for all scenarios from the two climate models, MPI-ESM-HR and BCC-CSM2-MR, repeat the procedure by simulating this solution for other climate models (CMCC-CM2-SR5 and MRI-ESM2).
We apply this methodology to design a resilient RES dedicated to two buildings located in Saudi Arabia–a residential and an office building with complementary load demand. As an initial step, we performed a comprehensive study on estimating the electricity demand for all climate scenarios using alternative climate models. These simulation results indicate that the electricity load increases throughout the century across most scenarios. The optimization results focus on the first-stage variables and the corresponding expected LCC of the RES over 25 years. We also address each socio-economic scenario individually to estimate the cost of climate change. For the end of the century, a high-emission scenario has a 20% increase in the LCC compared to the LCC obtained with a sustainable scenario. Detailed results indicate also the periods of unserved energy across multiple scenarios. The computational results show that incorporating the extreme weather events from various scenarios to obtain a feasible design over multiple models and scenarios can increase the expected LCC up to 271%, compared to the LCC obtained from one stochastic program with scenarios from only one climate model.
These findings provide useful insights into the value of stochastic programs and the high cost of resilience, which further motivate integrating decisions on building efficiency improvements and including demand side management options to flex load demand and help to address extreme weather events.
[1] Malheiro A., Castro P.M., Lima R.M., Estanqueiro A., 2015. Integrated sizing and scheduling of wind/PV/diesel/battery isolated systems. Renewable Energy 83, 646–657.
[2] https://www.iea.org/energy-system/buildings
[3] https://www.energy.gov/eere/buildings/zero-energy-buildings-resource-hub
[5] Demirbas, A., Hashem A. A., Bakhsh A. A., The cost analysis of electric power generation in Saudi Arabia, Energy Sources, Part B: Economics, Planning, and Policy 12 (2017) 591–596.
[6] John A., Douville H., Ribes A., Yiou P., 2022. Quantifying CMIP6 model uncertainties in extreme precipitation projections. Weather and Climate Extremes 36, 100435.