Area: Climate Modeling, Policy
Problem class: Large-scale Monte Carlo
Technologies: SaaS, GAMS, GAMS Engine
This case study examines the successful use of GAMS Engine SaaS by Rhodium Group to develop a comprehensive climate outlook ahead of the COP28 summit in Dubai. Rhodium Group leveraged advanced modeling techniques, including the RHG-GEM model and Monte Carlo simulations, to provide robust projections of energy, emissions, and temperature through the end of the century. GAMS Engine SaaS played a crucial role in providing flexible and extensive computational resources to support these simulations.
Climate Change Projections: Climate change remains one of the most pressing challenges of our time, requiring detailed and accurate projections to guide policymaking. Rhodium Group’s climate outlook aims to provide these projections by accounting for various uncertainties in economic and population growth, commodity prices, and technology costs.
RHG-GEM Model: The RHG-GEM model, initially a modified version of the U.S Energy Information Administration’s (EIA) World Energy Modeling System, and then developed into a completely new model, provides insights into global energy and climate dynamics across 16 world regions. It integrates the Electric and Emerging Clean Technology module (REEM) and its outputs are linked to the Finite Amplitude Impulse Response (FaIR) model to deliver detailed projections.
Monte Carlo Analysis: To address uncertainties, Rhodium Group employed a Monte Carlo Analysis with Latin hypercube sampling, which involved 4726 simulations in 2023, and 4950 simulations in 2024, each requiring approximately three hours of runtime. The scale of this effort demanded a highly scalable cloud solution.
GAMS Engine SaaS : GAMS Engine SaaS provided the necessary computational power and flexibility for the task. Horizontal scalability and custom backends allowed Rhodium to run 1200 simulations simultaneously and analyse their data in the cloud, significantly improving efficiency and performance.
High Uptime: Despite the high computational demands, GAMS Engine SaaS maintained high reliability and uptime, demonstrating the robustness of its fundamental design and infrastructure.
Cost Efficiency: The pay-as-you-use licensing model of GAMS Engine SaaS allows users to avoid substantial investments in hardware and maintenance, providing a cost-effective solution for organizations with varying computational needs.
The use of GAMS Engine SaaS by Rhodium Group exemplifies how advanced computational tools and scalable cloud solutions can address the complex challenges of climate modeling. This enabled Rhodium Group to produce critical climate projections, and as an added benefit, also drove significant scalability improvements in GAMS Engine SaaSfor all users.
Climate change stands as one of the most significant challenges of the coming decades, posing threats to ecosystems, economies, and societies all around the world. As global temperatures rise and the impact of climate change becomes more obvious, there’s an urgent need for fast and informed action from policymakers and stakeholders across multiple sectors; and high-quality, robust projections are at the heart of the issue.
These projections are essential to develop strategies that can mitigate risks and capitalize on opportunities as we transition into a sustainable future. However, they are also inherently accompanied with uncertainties stemming from various sources:
Economic and Population Growth: Future economic conditions and demographic trends significantly influence greenhouse gas emissions and energy consumption, yet they’re quite difficult to predict
Commodity Prices: Fluctuations in the prices of energy commodities, such as oil and gas, can impact the feasibility and adoption rate of alternative energy sources.
Clean Technology Costs: The costs of developing and deploying clean technologies are subject to rapid change, influenced by technological advancements, policy decisions, and market dynamics.
Rhodium’s main tool for delivering climate projections is the Global Energy Model (RHG-GEM), which is an advanced adaptation of the U.S. Energy Information Administration’s (EIA) World Energy Modeling System (WEPS). This model is designed to provide detailed insights into global energy and climate dynamics by dividing the world into 16 distinct regions and allowing for region-specific analysis. As an output, the RHG-GEM produces projections for energy use, emissions, and temperature changes through the end of the century.
These projections are essential for understanding the long-term impacts of current and future energy and climate policies. By simulating various scenarios, RHG-GEM equips stakeholders with the insights needed to anticipate trends and make informed decisions to effectively address climate change.
A key component of RHG-GEM is the Electric and Emerging Clean Technology module (REEM). Developed using the GAMS-based TIMES model framework, REEM applies a Linear Programming (LP) methodology to perform multiple detailed analyses:
In addition to energy and emissions forecasts, RHG-GEM provides comprehensive climate policy projections, mapping the anticipated evolution of climate action in response to political and socioeconomic changes. This capability enables stakeholders to address critical questions such as “What trajectory are we on?” by offering a detailed view of the potential future impacts of current policies and actions.
With its sophisticated and dynamic modeling features, RHG-GEM is an essential resource for policymakers, researchers, and industry leaders seeking to understand the complexities of climate change and design effective strategies for a sustainable future.
In order to develop a reliable climate outlook, Rhodium Group had to account for numerous uncertainties intrinsic to economic model inputs, such as population developments, commodity prices, and technology costs. To manage these uncertainties effectively, they employed Monte Carlo Analysis, a statistical technique that involves running a large number of simulations to capture a range of possible outcomes; all while ensuring thorough and evenly distributed input sampling of the input parameter space using Latin hypercube sampling
The scope of the Rhodium Group’s efforts is evident in the volume of simulations conducted. A total of 4,725 simulations were performed, each requiring approximately three hours to complete. Running this many simulations on local hardware is impossible for most organizations due to the substantial computational demands. A scalable cloud solution was essential.
In late 2022, Rhodium Group partnered with GAMS to address their significant computational needs. Anticipating the growing demand for large-scale computing, GAMS had developed the Engine SaaS to simplify and enhance cloud-based modeling. This tool proved to be the ideal solution.
GAMS Engine SaaS uses Kubernetes on AWS infrastructure, delivering a robust and scalable system for handling heavy computational workloads, which enabled Rhodium to execute their intricate model code—comprising Python glue code, compiled Fortran binaries, and GAMS code—seamlessly, with up to 1,200 simulations running simultaneously. This high-throughput capacity was critical in meeting the stringent requirements of their climate modeling efforts, providing timely insights ahead of the COP28 summit in Dubai.
The simulations present a stark outlook for the planet’s climate: they project that the global temperature rise is likely to exceed the crucial 2℃ threshold, emphasizing the urgent need for intensified climate action. For further details and analysis, Rhodium offers comprehensive articles and reports through the following links:
Working with a customer like Rhodium, that required exceptionally high compute demands, drove significant enhancements to Engine SaaS. Prior to this collaboration, we were well-prepared in terms of information security and vertical scalability. However, Rhodium’s extensive parallel computational requirements for Monte Carlo analysis highlighted challenges in horizontal scaling. For example, we repeatedly saturated all z1d.2xlarge instances available in the AWS US-East-1a availability zone.
To address this issue, we integrated AWS resources across multiple availability zones into our compute cluster, successfully overcoming the resource constraints. This advancement significantly improved our system’s ability to handle large-scale simulations, benefiting all our customers.
Another forthcoming improvement was the introduction of custom data backends. With GAMS Engine, job results are typically transferred via the REST API, but in Rhodium’s case, the high volume of data created a bottleneck. Our development team introduced custom data backends, enabling Rhodium to avoid transferring large datasets to and from the cloud by utilizing S3 buckets. This enhancement facilitates post-processing directly within AWS, which reduces data transfer times and increases efficiency.
GAMS Engine SaaS demonstrated remarkable reliability during the heightened demand in 2023, achieving an impressive uptime of 99.961%, as verified by an external monitoring tool. This performance underscores the strength of our infrastructure and our dedication to delivering dependable services.
GAMS Engine SaaS provides a straightforward licensing model based on a pay-as-you-go approach, ensuring customers are charged solely for the hardware resources consumed by each job.
A significant advantage of GAMS Engine SaaS is the elimination of large upfront investments in infrastructure and ongoing maintenance. This is particularly advantageous for organizations with variable computational needs. For example, the Rhodium project would have required over $1 million in investments to run 1200 parallel simulations on pre-configured AWS Outpost racks—an impractical solution for most users.
With GAMS Engine SaaS, users can scale computational resources on demand, paying only for what they use while avoiding the complexities and costs of maintaining hardware infrastructure. This approach offers flexibility and cost-effectiveness, making high-performance computing accessible to a broader range of users and applications.
Additionally, the flexible REST API enables users to run any GAMS job in the cloud with minimal programming effort.