Presented at: MODSIM 2001, International Congress on Modelling and Simulation, Canberra, Australia, 10- 13.12.2001 Reservoir Operation by Fuzzy Reasoning A. Jolma a , E. Turunen b , M. Kummu a and T. Dubrovin c a Helsinki University of Technology, P.O.Box 5200, 02015 HUT, Finland (Ari.Jolma@hut.fi) b Tampere University of Technology, Tampere, Finland c Finnish Environment Institute, Helsinki, Finland Abstract: The operation of lakes and reservoirs is currently under scrutiny in Finland due to the aging of the old operation permits, climate change, and changing environmental values. Simulation is a viable tool for studying reservoir operation but difficult to use when the operation does not have clear explicit rules. Fuzzy logic has been shown to be useful in modeling simple decision making which requires intelligence. Two fuzzy logic based methods, namely case-based and rule-based reasoning were examined and applied for modeling the operation of five lakes of a river basin. Different mathematical structures and interpretations were tested including the use of fuzzy similarity based approaches. The best results were obtained using a rule based ap- proach with a simple mathematical structure and interpretation of the fuzzy and as a product. The rule base in the model was allowed to be inconsistent and each rule had an associated weight or strength. The case-based approach performed rather well in comparison with the rule-based approach and in some cases better in some respect. Both models can mimic the human operator reasonably well in easy circumstances but have problems especially in the downstream lakes with large discharge to storage ratios. The fuzzy logic based models, case- based reasoning or rule-based reasoning require further work to be applicable in studying practical problems like the adjusting the operation to changes in environmental values and climate. Keywords: Reservoir Operation; Fuzzy Logic; Total Fuzzy Similarity; Simulation 1. INTRODUCTION The actual operation of many Finnish lakes is often a compromise between the goals of the operator and the rules set out in an operation permit. Often neither the operator's own rules nor the operation permit state explicitly what the release should be in a given situation. The operation is thus very much based on the experience and skill of the managers. This forms an obstacle for modeling and simula- tion of the operation of the lake. Currently many operation permits are being re- vised. A strong motivation for the revision are the changing environmental values and changing uses of water resources. A thorough analysis of the op- eration and its impact requires simulation under existing and predicted conditions. This paper describes an attempt to model the op- eration of a system of five lakes using a fuzzy logic based approach. The attempt is based on general knowledge and on water level and release data. The goal of the modeling at this stage is to mimic the human operator. We have tested several tech- niques and mathematical formulations including the total fuzzy similarity of Turunen [1999]. Several approaches to apply fuzzy set theory to reservoir operation have been described in the lit- erature. These include fuzzy optimization tech- niques, fuzzy rule base systems, and combinations of fuzzy approach with other techniques. Applica- tions can be found in the work of Fontane et al. [1997], Huang [1996], and Saad et al. [1996]. Fuzzy rule base control systems for reservoir op- eration are presented by Russell and Campbell [1996] and Shrestra et al. [1996]. The fuzzy rule base can be constructed on the basis of expert knowledge or observed data. Methods for deriving a rule base from observations have been presented by Bárdossy and Duckstein [1995] and Kosko [1992]. 2. FUZZY REASONING OF RESERVOIR RELEASE 2. 1 Introduction