The doctoral dissertations of the former Helsinki University of Technology (TKK) and Aalto University Schools of Technology (CHEM, ELEC, ENG, SCI) published in electronic format are available in the electronic publications archive of Aalto University - Aaltodoc.
|
|
|
Studies on Long-Term Inflow Forecasting
Jarkko Koskela
Dissertation for the degree of Doctor of Science in Technology to be presented with due permission of the Faculty of Engineering and Architecture for public examination and debate in Auditorium R1 at Helsinki University of Technology (Espoo, Finland) on the 26th of June, 2009, at 12 noon.
Dissertation in PDF format (ISBN 978-951-22-9933-1) [1777 KB]
Dissertation is also available in print (ISBN 978-951-22-9932-4)
Abstract
This thesis aims to improve knowledge of long-term inflow and streamflow forecasts. A special focus is on the development of a new long-term forecast model and on the evaluation of long-term inflow forecasts.
In the first part of the work, a new categorical long-term forecast model is developed and its performance is investigated in four case studies. The forecasts are based only on the current hydrological state of the basin and thus, weather forecasts are not utilised. By using the k-Nearest Neighbour Rule (k-NRR) or the minimum distance classifier (MDC), the forthcoming period is classified into a wetness class based on the hydrological state of the basin on the forecast date. Inflow forecast is finally based on this classification. The results show that for a lake with a large basin (Lake Päijänne case study), this forecast model could be used in real-time inflow forecasting and the results are comparable with the forecast accuracy of the multiple linear regression models. For small basins (<10 km2) and in Lake Pyhäjärvi, the use of the new model for long-term discharge forecasting gave satisfactory results on April 1. On October 1, long-term forecasting turned out to be difficult irrespective of the forecast model.
In the second part of the work, long-term inflow forecasts are evaluated based on their length and accuracy. The study is based on two cases: a single multipurpose reservoir Lake Pyhäjärvi in Säkylä and a multipurpose lake-river system, River Kymijoki. The evaluation method is based on artificially generated inflow forecasts and on the optimisation of the release sequences based on these forecasts. The results are in line with the outcome of similar international studies: if the live capacity of the lake-river system compared with the annual inflow is small, short and accurate forecasts should be aimed at. For large systems, a long forecast period should be used without focusing as much on forecast accuracy. The main finding, however, is related to approximation of the potential hydropower production increase in Finland by supposing that forecast accuracy could be improved and the optimal forecast periods used. In the two case studies it was possible to increase hydropower production up to 0.7-9% compared with the status quo during the study period, if perfect inflow forecasts had been available. However, the realistic possibilities to increase hydropower production in Finland by improving forecast accuracy were approximated to be 0.5-2% at the maximum. At the same time problems related to floods and droughts would decrease.
Simulated annealing is used as the optimisation algorithm in the operation of the systems, and the evaluation of the performance of this algorithm was one of the special objectives of this study. The algorithm was flexible and reliable.
Keywords: water resources management, hydropower, long-term inflow forecasting, pattern recognition, forecast accuracy, simulated annealing, lake-river system
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
© 2009 Helsinki University of Technology
Last update 2011-05-26