|
|
Enron Mail |
Hi Tom:
Thanks for the prompt and accurate evaluation of what MODREG does. I know about you leaving Enron and what your schedules looks like, so I think I will try to make the best use of time and resources in order to have an SDDP model up and running for Argentina. In the meeting we had with Vince K., and after reviewing the calibration exercise run with MODREG we have decided to: 1. Set up the Argentinean system into the SDDP model * I would like to know what information you need in order to get this done 2. Run the calibration exercise with the SDDP model in order to compare results. * Feed the model with actual data (hydro generation, Exports) from January to June and get the output for Prices, Thermal Generation, Exports/Imports Please find attached the presentation regarding MODREG and the calibration results. I will be talking to the consultants in order to get more information in some topics we need further research such as. Transmission losses, forced dispatched units and gas availability. Please contact me so I can start gathering all you need to get the model running Regards Jorge Gumucio From: Tom Halliburton/ENRON@enronXgate on 09/28/2001 02:29 PM CDT To: Remi Collonges/SA/Enron@Enron, David M Rosenberg/SA/Enron@Enron, Jorge Gumucio/SA/Enron@Enron cc: Vince J Kaminski/ENRON@enronXgate, Zimin Lu/ENRON@enronXgate Subject: ModReg for Argentina Folks, Here's some preliminary comments on the model we are using for Argentina. Please correct me on any misunderstandings! ModReg was written by consultants in Buenos Aires. Documentation is available in Spanish only. All data input is by means of text files, which require column orientated input, ie numbers must go into specific columns. Detailed documentation has been provided specifying the data format for each file. A weekly time step is used, with five load blocks per week, and a time horizon of one year, typically. Each model run requires about 6 hours of processing. ModReg is a modified version of the model used by the grid operator. It is coded in Fortran and requires a CPLEX licence. CPLEX is a powerful linear programming solver, developed by ILOG. As the model is written in Fortran, the very good higher level interface to CPLEX is not available, so the code will have to access CPLEX at a lower level. We have no documentation describing how the model algorithms. Source code is not available at present. ModReg appears to model around 30 busses and 71 lines. Lines are characterised by from bus number, to bus number, maximum MW capacity in each direction, a percentage loss factor, cost of transmission in each direction, and some other details. Demand data is specified by week and load block for each bus. Thermal plants have a heat rate and maintenance schedule specified, with the possibility of multiple fuels. I understand that hydro generation schedules for each historical inflow condition are an input to the model as they are calculated elsewhere. From the above it appears that we have an unsatisfactory situation: ModReg is difficult to use we do not know in detail what the model does we do not have documentation describing the algorithm because source code is not available, we are completely dependent on the consultants. If Enron is to continue modeling Argentina, we should as a minimum have a model that we understand, is reasonably easy to use, and is documented. A graphical user interface is really needed to make ModReg easier to use. Documentation is of course essential. We should not use a model developed by consultants without documentation. These steps would enable us to have more confidence in model results, and shortens the learning curve to enable more people to use model. SDDP is likely to be a suitable replacement. It would require reseroir inflow data rather than the calculated generations. The regional loads etc could be modeled using the DC loadflow feature, with zero impedance lines. Transmission losses in SDDP are a quadratic function of power flows, so some guesses would have to be made for line resistence values. Setting up an SDDP model to represent all the aspects included in ModReg would be a time consuming job, that would require a good undertanding of the data for both models. Some direct cut and paste of data from one model's files to the others will be possible with a suitable column oriented text editor. However the end result would be a model with the desirable attributes of documention, GUI interface, and flexible output. Tom Halliburton
|