Expanding Renewable Energy Networks under Changing Climate Conditions: A Three-Stage Robust Optimization Approach using FICO Xpress
The FICO Xpress Best Paper Award was established in 2022 to recognize exceptional research in Mathematical Optimization, Operations Research, and related fields. Among a multitude of high-quality submissions, the scientific committee had the challenging responsibility of selecting a winner. The Award goes to …*drumroll* …:
Moreira, A., Pozo, D., Street, A., Sauma, E., & Strbac, G. (2021). Climate‐aware generation and transmission expansion planning: A three‐stage robust optimization approach. European Journal of Operational Research, 295(3), 1099-1118.
In this blog post, we explore the fascinating research conducted by Alexandre Moreira and his team of international coauthors, providing insights into challenges and solution approaches in the realm of Climate-Aware Power Production Planning.
Uncertainty management poses a great challenge in modern power systems planning and operation. In contrast to the power systems of the 20th century, which relied on a limited number of controllable power stations and exhibited predictable demand, 21st-century power systems are much more complex. They consist of numerous small, non-controllable stochastic power sources that are influenced by climate patterns undergoing significant changes.
This new paradigm introduces short-term uncertainties that utilities and system operators must address when planning their assets and infrastructure. These uncertainties primarily stem from the generation output of intermittent renewable units and system element outages, which can be triggered by unpredictable and destructive natural disasters. Considering these uncertainties is crucial for determining the appropriate reserves required during real-time operation to meet the energy demand of all customers.
Moreover, long-term uncertainty can substantially impact the behavior of renewable energy resources, which are increasingly integrated into power systems. In various regions across the globe, the El Niño and the Southern Oscillation (ENSO) play a major role in influencing long-term climate patterns. For instance, in Southern California, El Niño and La Niña events significantly affect wind speed variability, while in Australia, the ENSO strongly influences wind and solar patterns. In the Northeast of Brazil, wind speeds during El Niño events have been observed to exceed historical averages, whereas during La Niña conditions, wind speeds tend to be lower. In Southern South America, the ENSO has a notable impact on wind speeds and, consequently, wind energy production over monthly and seasonal timescales, particularly in spring. Specifically in Chile, the effects of the ENSO on wind speeds vary across regions and months. For instance, the warm phase of the ENSO (El Niño) tends to reduce wind speeds in central Chile and increase them in the south during September to November.
Figure 1: Extreme weather conditions impact renewable energy generation. How can we make expansion planning robust against unknown future likelihood changes of such events?
By addressing both short-term and long-term uncertainties, power system operators can effectively plan and operate their assets, ensuring reliable energy supply to meet the demands of customers in this evolving energy landscape. To tackle this issue, we have developed a hierarchical multilevel optimization model that takes into account the impacts of the El Niño and La Niña events, part of the El Niño-Southern Oscillation (ENSO) phenomenon.
Our model considers both long-term and short-term uncertainties. In the long term, we focus on the expansion of generation and network infrastructure, while acknowledging that climate conditions, such as wind speeds, can be influenced by ENSO. We analyze the impact of ENSO on wind power generation in terms of average production, standard deviation, and spatial correlation. On the other hand, in the short term, we address wind variations and failures of generating units and transmission lines during system operation.
To handle uncertainties at different time scales effectively, we propose a three-stage robust model formulated as a five-level optimization problem. This model provides a comprehensive framework for decision-making in both the long and short terms. In the long term, the key decisions revolve around generation and transmission expansions, made without knowledge of future climate conditions, renewable generation profiles, or system element availability.
Our model introduces an adaptive energy and reserves scheduling approach in the short term, incorporating information about the current climate state. This enables system operators to adapt their short-term day-ahead forecasts based on the more long-term information available. By considering the proposed optimization model, we ensure both cost efficiency and compliance with a user-defined deterministic security criterion, accounting for short-term wind variability.
To solve our optimization model, we have developed a variant of the nested column-and-constraint-generation algorithm, ensuring global optimality within a finite number of steps.
Figure 2: Out-of-Sample analysis – Costs in bars (associated with the left vertical axis) and lines (associated with the right vertical axis) indicating Expected Value and CVaR95% of system imbalance related to the expansion plans obtained via stochastic and robust models.
Through extensive case studies, we have demonstrated the benefits of considering different climate patterns when planning power system expansions. We compared various approaches, including stochastic and robust models, and highlighted the superior performance of the proposed robust model in terms of cost and efficiency. By conducting a realistic case study based on the Chilean power system, we showcased the impact of different climate states on wind generation patterns and emphasized the cost savings achievable through the incorporation of conditional short-term uncertainty in operational decisions.
More precisely, in our study, we examined different climate states, including La Niña, El Niño, and neutral conditions, each characterized by specific wind generation patterns. We compared four different approaches to determine the expansion plan for the case study, all of them conveniently implemented in the mathematical modeling and programming language FICO Xpress Mosel:
In contrast, our proposed 3-stage robust approach, with both n-1 and n-2 security criteria, took approximately 15 hours and 59 hours (about 2.5 days), respectively, to obtain proven optimal solutions with FICO Xpress Solver. As depicted in Figure 2, our proposed methodology exhibited slightly superior performance in terms of system power imbalance with the n-1 security criterion and considerably superior performance with the n-2 security criterion.
Furthermore, the investment and scheduling costs associated with our proposed methodology using the n-1 security criterion were roughly half compared to the costs associated with the 3-stage stochastic formulation. Even with the n-2 security criterion, the costs were significantly lower.
These findings highlight the effectiveness and efficiency of our proposed robust approach in addressing uncertainty and ensuring system reliability while achieving cost savings in power system expansion planning. Furthermore, our results demonstrate that the representation of conditional short-term uncertainty for each climate state followed by conditioned operational decisions can change the expansion plan and significant cost savings can be achieved by considering the impact of different climate states on the renewable generation patterns while planning the expansion of power systems.
Our solution algorithm has been implemented in Xpress Mosel using the IDE FICO Xpress Workbench, leveraging the efficiency of Xpress Solver and the parallel execution capabilities of Mosel. This allowed us to evaluate the operation of power systems for different climate states in parallel iterations, enhancing the overall efficiency of the algorithm.
By integrating climate uncertainty considerations into power system planning and operation, we can pave the way for more resilient and efficient energy systems, capable of meeting the challenges of an evolving energy supply landscape.
We at FICO congratulate Alexandre Moreira, David Pozo, Alexandre Street, Enzo Sauma, and Goran Strbac on winning the first ever FICO Xpress Best Paper Award. Amidst numerous strong contenders, their remarkable research stood out prominently. Like many other research teams around the world, they are members of our Academic Partner Program. Join our FICO Xpress Optimization Community and learn about our free community licenses for everyone and our fully-fledged free licenses for all academics. The call for the 2023 edition of the FICO Xpress Best Paper Award is out now: Consider submitting your Xpress-related research paper!
The FICO Xpress Best Paper Award was established in 2022 to recognize exceptional research in Mathematical Optimization, Operations Research, and related fields. Among a multitude of high-quality submissions, the scientific committee had the challenging responsibility of selecting a winner. The Award goes to …*drumroll* …:
Moreira, A., Pozo, D., Street, A., Sauma, E., & Strbac, G. (2021). Climate‐aware generation and transmission expansion planning: A three‐stage robust optimization approach. European Journal of Operational Research, 295(3), 1099-1118.
In this blog post, we explore the fascinating research conducted by Alexandre Moreira and his team of international coauthors, providing insights into challenges and solution approaches in the realm of Climate-Aware Power Production Planning.
Uncertainty management poses a great challenge in modern power systems planning and operation. In contrast to the power systems of the 20th century, which relied on a limited number of controllable power stations and exhibited predictable demand, 21st-century power systems are much more complex. They consist of numerous small, non-controllable stochastic power sources that are influenced by climate patterns undergoing significant changes.
This new paradigm introduces short-term uncertainties that utilities and system operators must address when planning their assets and infrastructure. These uncertainties primarily stem from the generation output of intermittent renewable units and system element outages, which can be triggered by unpredictable and destructive natural disasters. Considering these uncertainties is crucial for determining the appropriate reserves required during real-time operation to meet the energy demand of all customers.
Moreover, long-term uncertainty can substantially impact the behavior of renewable energy resources, which are increasingly integrated into power systems. In various regions across the globe, the El Niño and the Southern Oscillation (ENSO) play a major role in influencing long-term climate patterns. For instance, in Southern California, El Niño and La Niña events significantly affect wind speed variability, while in Australia, the ENSO strongly influences wind and solar patterns. In the Northeast of Brazil, wind speeds during El Niño events have been observed to exceed historical averages, whereas during La Niña conditions, wind speeds tend to be lower. In Southern South America, the ENSO has a notable impact on wind speeds and, consequently, wind energy production over monthly and seasonal timescales, particularly in spring. Specifically in Chile, the effects of the ENSO on wind speeds vary across regions and months. For instance, the warm phase of the ENSO (El Niño) tends to reduce wind speeds in central Chile and increase them in the south during September to November.
Figure 1: Extreme weather conditions impact renewable energy generation. How can we make expansion planning robust against unknown future likelihood changes of such events?
By addressing both short-term and long-term uncertainties, power system operators can effectively plan and operate their assets, ensuring reliable energy supply to meet the demands of customers in this evolving energy landscape. To tackle this issue, we have developed a hierarchical multilevel optimization model that takes into account the impacts of the El Niño and La Niña events, part of the El Niño-Southern Oscillation (ENSO) phenomenon.
Our model considers both long-term and short-term uncertainties. In the long term, we focus on the expansion of generation and network infrastructure, while acknowledging that climate conditions, such as wind speeds, can be influenced by ENSO. We analyze the impact of ENSO on wind power generation in terms of average production, standard deviation, and spatial correlation. On the other hand, in the short term, we address wind variations and failures of generating units and transmission lines during system operation.
To handle uncertainties at different time scales effectively, we propose a three-stage robust model formulated as a five-level optimization problem. This model provides a comprehensive framework for decision-making in both the long and short terms. In the long term, the key decisions revolve around generation and transmission expansions, made without knowledge of future climate conditions, renewable generation profiles, or system element availability.
Our model introduces an adaptive energy and reserves scheduling approach in the short term, incorporating information about the current climate state. This enables system operators to adapt their short-term day-ahead forecasts based on the more long-term information available. By considering the proposed optimization model, we ensure both cost efficiency and compliance with a user-defined deterministic security criterion, accounting for short-term wind variability.
To solve our optimization model, we have developed a variant of the nested column-and-constraint-generation algorithm, ensuring global optimality within a finite number of steps.
Figure 2: Out-of-Sample analysis – Costs in bars (associated with the left vertical axis) and lines (associated with the right vertical axis) indicating Expected Value and CVaR95% of system imbalance related to the expansion plans obtained via stochastic and robust models.
Through extensive case studies, we have demonstrated the benefits of considering different climate patterns when planning power system expansions. We compared various approaches, including stochastic and robust models, and highlighted the superior performance of the proposed robust model in terms of cost and efficiency. By conducting a realistic case study based on the Chilean power system, we showcased the impact of different climate states on wind generation patterns and emphasized the cost savings achievable through the incorporation of conditional short-term uncertainty in operational decisions.
More precisely, in our study, we examined different climate states, including La Niña, El Niño, and neutral conditions, each characterized by specific wind generation patterns. We compared four different approaches to determine the expansion plan for the case study, all of them conveniently implemented in the mathematical modeling and programming language FICO Xpress Mosel:
- A 3-stage stochastic formulation minimizing expected cost in both long and short terms.
- A 3-stage stochastic formulation minimizing maximum cost in the long term and Conditional Value at Risk (CVaR95%) of imbalance in the short term.
- Our proposed 3-stage robust approach with an n-1 security criterion, which ensures the system can handle the failure of any single generator or transmission line without load shedding.
- Our proposed 3-stage robust approach with an n-2 security criterion.
In contrast, our proposed 3-stage robust approach, with both n-1 and n-2 security criteria, took approximately 15 hours and 59 hours (about 2.5 days), respectively, to obtain proven optimal solutions with FICO Xpress Solver. As depicted in Figure 2, our proposed methodology exhibited slightly superior performance in terms of system power imbalance with the n-1 security criterion and considerably superior performance with the n-2 security criterion.
Furthermore, the investment and scheduling costs associated with our proposed methodology using the n-1 security criterion were roughly half compared to the costs associated with the 3-stage stochastic formulation. Even with the n-2 security criterion, the costs were significantly lower.
These findings highlight the effectiveness and efficiency of our proposed robust approach in addressing uncertainty and ensuring system reliability while achieving cost savings in power system expansion planning. Furthermore, our results demonstrate that the representation of conditional short-term uncertainty for each climate state followed by conditioned operational decisions can change the expansion plan and significant cost savings can be achieved by considering the impact of different climate states on the renewable generation patterns while planning the expansion of power systems.
Our solution algorithm has been implemented in Xpress Mosel using the IDE FICO Xpress Workbench, leveraging the efficiency of Xpress Solver and the parallel execution capabilities of Mosel. This allowed us to evaluate the operation of power systems for different climate states in parallel iterations, enhancing the overall efficiency of the algorithm.
By integrating climate uncertainty considerations into power system planning and operation, we can pave the way for more resilient and efficient energy systems, capable of meeting the challenges of an evolving energy supply landscape.
We at FICO congratulate Alexandre Moreira, David Pozo, Alexandre Street, Enzo Sauma, and Goran Strbac on winning the first ever FICO Xpress Best Paper Award. Amidst numerous strong contenders, their remarkable research stood out prominently. Like many other research teams around the world, they are members of our Academic Partner Program. Join our FICO Xpress Optimization Community and learn about our free community licenses for everyone and our fully-fledged free licenses for all academics. The call for the 2023 edition of the FICO Xpress Best Paper Award is out now: Consider submitting your Xpress-related research paper!
Updated 3 years ago
Version 3.0