“20th PSCC 2018 papers submission and review platform

Full Program »

Twenty Years of Application of Stochastic Dual Dynamic Programming in Official and Agent Studies in Brazil – Main Features and Improvements on the Newave Model

The Brazilian generating system presents a high share of renewable energy sources, which includes hydropower, biomass and wind power, accounting for more than 80% of country´s electricity consumption. It is hydro dominated and characterized by large reservoirs presenting multi-year regulation capability, arranged in complex cascades over several river basins.
Since 1991, CEPEL, the Brazilian Electric Energy Research Center, has been proposing to apply stochastic dual dynamic programming (SDDP) algorithms to de-termine the optimal allocation of the generation resources, especially hydro and thermal sources, in the long and mid-term operation planning. In 1993, CEPEL extended the original SDDP formulation to take into account serial correlations of the inflows to the reservoirs. This led to considering another state variable (hydrological trend) in addition to the original one (reservoir storage level in each equivalent energy reservoir (EER). In that year, CEPEL also started the implementation of these SDDP algorithms in a computational model called NEWAVE, to solve the long term hydrothermal schedule problem for the Brazilian system.
In 1998, an institutional framework for the Brazilian electrical sector was issued leading to the need of calcu-lation of the so-called “initial contracts” to phase in the existing contracts from the older to this institutional model; these initial contracts were computed by solving a long-term hydrothermal coordination problem in a specific procedure. On the other hand, a new transmission line started its operation in that year, connecting the North/Northeast and South/Southeast subsystems, which were operated independently until this date. Due to the dimensions of these now four large and interconnected subsystems, the only existing methodology to deal with the associated long-term hydrothermal coordination problem was the SDDP. Therefore, the year of 1998 is the milestone for the first official use of the NEWAVE model in Brazil.
The Brazilian System Operator (ONS) and Whole Sale Energy Market (MAE/CCEE) were created and since the year of 2000 the NEWAVE model has officially been used by the ONS to dispatch the hydro and thermal power plants and by MAE/CCEE to calculate the electricity spot prices. Also, a comprehensive procedure was established to validate the new releases of the NEWAVE program, through a task-force co-coordinated by ONS and CCEE with participation of all agents, and which requires a formal approval by the Electricity Regulator. The year of 2000 also marks the use of the NEWAVE program by the Brazilian government in the ten-year expansion plan studies.
A new institutional framework was issued in 2004. A new entity (EPE) was created which is in charge of the studies associated to the energy expansion planning, under the coordination of the Ministry of Mines and Energy (MME), and NEWAVE program was kept as the official tool in these expansion studies. The concept of power plant assured energy was also kept and public auctions were introduced as a procurement mechanism to guarantee capacity expansion through PPAs with the distribution companies; in both cases, the NEWAVE model was adopted to calculate associated figures of merit. Because of the strategic uses of this model, a Standing Committee for Analysis of Methodologies and Computational Programs of the Electric Sector was established in 2007 by MME.
This paper describes the key features and recent en-hancements of the SDDP algorithm implemented in the NEWAVE model in terms of mathematical modeling, computational aspects and real decision making, such as: a multivariate selective sampling to define the sub-tree of streamflow scenarios to enhance the robustness of the results with regard to variations in its sample size; two risk aversion methodologies to improve the system security – the Conditional Value at Risk (CVaR) and the Risk Aversion Surface (SAR); representation of more than one ERR per subsystem in order to better represent distinct hydrological behaviors of plants belonging to the same subsystem; an iterative a procedure to select and keep only the active Benders cuts that compose the LPs, resulting in a high performance with the same accuracy; methodologies to estimate the energy deficit probability to assess whether electricity rationing should be implemented; synchronized parallel processing strategies to distribute the LP problems among the processors and reduce the computational time; an hybrid approach that allows a combination of individual hydropower plants representation, e.g., in the first years, with its representation by EERs, e.g., in the late years of the plan-ning horizon.
These features and improvements allowed keeping the NEWAVE model in the state-of-the art and its broad range of practical applications in official and agent studies in the Brazilian electrical sector along these twenty years.

Author(s):

MARIA ELVIRA MACEIRA    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

DEBORA PENNA    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

ANDRE DINIZ    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

CESAR VASCONCELLOS    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

ROBERTO PINTO    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

CRISTIANE CRUZ    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

ALBERT MELO    
CEPEL - ELECTRIC ENERGY RESEARCH CENTER
Brazil

 

Powered by OpenConf®
Copyright ©2002-2014 Zakon Group LLC