Just over two years ago, PSR won an international competition to develop an advanced energy analysis model for the North American Pacific Northwest region. This region has a very diverse set of generation sources, including the Bonneville Power Administration (BPA) hydroelectric plants, large wind and natural gas generation capacity, and electrical interconnections with California, other West Coast states and Canada. The specifications of the model, called Genesys (GENeration Evaluation SYStem), were highly ambitious: to optimize the operation of the generation system and interconnections considering the hourly operational details from all sources, the uncertainty of wind and hydroelectric production, and the operational and environmental constraints of hydroelectric plants, of the most complex in the world, for thousands of scenarios. To meet the specifications, it would be necessary to solve hundreds of millions of large-scale optimization problems, a world record for models of this type.
The solution proposed by PSR for Genesys was based on three components at the frontier of algorithms and high-performance computing: (i) adaptation of PSR’s SDDP stochastic operating model, used in dozens of countries on all continents; (ii) a cloud computing architecture with tens of thousands of processors working in parallel to perform all analyzes in a few hours; (iii) a cloud-based environment, PSR Core, which integrates data input, model execution and distributed graphical analysis of the results, which contain tens of Terabytes of data.
The model was extensively tested and validated for about a year, together with the NWPCC’s (Northwest Power and Conservation Council) technical team, responsible for the project, and in presentations to dozens of companies and agencies in the region, such as the BPA and the US Army Corps of Engineers. In the middle of last year, Genesys was used to prepare the recently published 2021 Northwest Power Plan, a regional expansion plan comprising the states of Idaho, Montana, Oregon and Washington, whose objective is to balance the energy and environmental needs of the region, with an emphasis on preserving the Columbia River benefits for the future (https://www.nwcouncil.org/about). The scenarios analyzed included changes in the portfolio of the main agents with the entry of renewables, robustness in energy efficiency, dependence on external resources and paths to decarbonization (https://www.nwcouncil.org/2021-northwest-power-plan/).
The plan analyzed more than 12,000 scenarios, using 24,000 cloud processors. In two days, almost 200,000 CPU hours were used, and more than 200 million large-scale integer optimization problems (millions of decision variables and constraints) were solved, resulting in more than 10 terabytes of data for post-processing on the PSR Core.
John Ollis, NWPCC’s Planning and Analysis Coordinator, comments “PSR has gone above and beyond to make sure that we have a state-of-the-art tool to answer questions we previously thought outside the realm of what could be modeled. During the Council’s most recent regional power plan, PSR provided invaluable and timely assistance to Council staff by creating a tool to help understand the operations of a heavily constrained hydropower system under different policy outcomes and fast changing power markets dominated by renewable energy. The PSR tools and team have surpassed my expectations time and time again and I would recommend them to anyone.”
According to Joaquim Garcia, who led the development of the Genesys algorithms, “the success of this project is due to the multidisciplinary team at PSR, which combines the best of mathematical theory, high-performance computing, large databases, cloud computing and graphics tools.” For Mario Veiga, founder and current Chief Innovation Officer of PSR, “Genesys illustrates the high quality of PSR tools, which combine the most modern computational and mathematical technologies, and which for more than thirty years have been exported to over seventy countries in all continents.”