• Main Page
  • Products
  • DECOMP – Short Term Operation Planning Model for Interconnected Hydrothermal Systems
  • VOLTAR
  • PRINT
  • SEND
  • A+ A-
DECOMP – Short Term Operation Planning Model for Interconnected Hydrothermal Systems

Introduction
In view of the large complexity of the characteristics, conditions and constraints for the hydrothermal generation systems, the generation planning task is performed in different phases. For the mid term planning performed by the NEWAVE model, the time horizon comprises up to five years ahead, discretized in monthly steps, where the key objective is to set the value of water in the reservoirs along time, according to the storage levels in the reservoirs. In this sense, a so-called “operation policy” is obtained, which is composed of a set of cost-to-go functions for each time step.
 
In the short-term planning, a horizon of up to 12 months discretized in weekly and monthly steps may be considered, in order to determine the individual generation targets of hydro and thermal plants as well as the energy interchanges between subsystems (system areas), considering the expected cost-to-go function provided by the mid-term model at the end of its time horizon. At this phase, a scenario tree to represent the uncertainties in the inflows to the reservoirs and the generation of intermittent sources can be taken into account.
 
The DECOMP program has been developed by Cepel’s Department for Energy Optimization and the Environment (DEA), for application in the short-term planning of hydrothermal systems, and, together with the NEWAVE model, is the official tool used by the Brazilian Independent System Operator (ONS) to set the weekly/monthly operation program (PMO) and by the Market Operator (CCEE) to set the weekly prices in Brazil, in three load blocks. As a consequence, the main objective of the program is to determine the generation targets for each hydro/thermal plant, subject to stochastic inflows, in order to meet the demand and minimize the expected operation cost along the planning period, and taking into account the conditional value-at-risk (CVaR) measure.
 
The prices are obtained based on the marginal costs provided by the model for each system area, which comes from a sensitivity analysis of the system costs according to an increase in the demand. The problem is formulated as a linear program, representing the physical and operation characteristics of the hydro plants in an individualized way. The uncertainty on the inflow to the reservoirs is considered by means of a scenario tree produced by the GEVAZP model, where for each node a given probability is assigned (Figure 1). Constraints on the anticipated dispatch of LNG thermal plants are also considered, and nonlinear characteristics are modeled by using piecewise linear functions.
 
The problem is decomposed into a subproblem for each time step and scenario, and the overall problem is solved by applying multistage Benders decomposition, also known as dual dynamic programming. At the end of its time horizon, the DECOMP model considers the cost-to-go function provided by the NEWAVE model. As a result of its solving strategy, DECOMP obtains a cost-to-go function for each node of the scenario tree, and in particular the cost-to-go function at the end of the first week, which is used as a boundary condition to provide the water values in the reservoirs for the DESSEM model. This latter model aims to obtain the daily operation dispatch of the system and is being validated by the ISO and CCEE for official use to set the hourly prices and operation in Brazil, starting in January 2020.
 
 
In order o provide flexibility in terms of problem formulation, the DECOMP model provides the following features:
 
 

Time and scenario representation

 

• Weekly time steps for the first month, and monthly stages from the second month on, with representation of the uncertainties in the inflows to reservoirs and generation of intermittent sources, through a scenario tree, with a horizon of up 1 year;
• representation of a load duration curve within each scenario/time step;
• Integration with medium-term operation planning models (NEWAVE), through its future cost function.
 
 

Reservoir operation

 

• Water balance in the reservoirs for each stage and scenario, considering the water delay time between cascaded plants, evaporation in the reservoirs, operation of spillways and filling of dead volume for the plants which are starting its operation;
• Several operation constraints to the reservoirs, such as storage limits, minimum/maximum inflow/outflow, as well as water withdrawals/returns due to other uses;
• Waiting volume for flood control.
 
 

Hydropower Generation

 

• Variation of efficiency of hydro generation with the water head, by means of a multi-dimensional piecewise linear hydro production function that represents hydro generation as a function of turbined outflow, storage and spillage;
• maximum turbined outflow limits as a function of the storage level, representation of the scheduled maintenance and forced outage rates of the hydro units.
 
 

Thermal Generation and other sources

 

• Inflexibility of thermal generation, and anticipated dispatch constraints for the operation of LNG plants;
• Uncertainty on the generation of intermittent sources, as for example wind and solar plants;
• Energy interchanges with neighborhood systems.
 
 

Transmission system

 

• Modeling of a multi-area system, with interchange limits among them;
• consideration of special electrical constraints that allow the representation of dc power flow limits in some transmission lines, as well as voltage control in some regions of the system;
 
 

Computational features

 

The DECOMP model applies parallel processing to speed up the CPU time to solve the problem. The current version of the program is released on a Linux Linux environment, and can be run in a local computer or remotely, by using clusters of computers or cloud computing.

Contact

Contact the responsible area via email:


 decomp@cepel.br