Multireservoir Modeling with Dynamic Programming and Neural Networks
01 Apr 2001-Journal of Water Resources Planning and Management (American Society of Civil Engineers)-Vol. 127, Iss: 2, pp 89-98
TL;DR: The multireservoir model based on the dynamic programming-neural network algorithm gives improved performance in this study.
Abstract: For optimal multireservoir operation, a dynamic programming-based neural network model is developed in this study. In the suggested model, multireservoir operating rules are derived using a feedforward neural network from the results of three state variables' dynamic programming algorithm. The training of the neural network is done using a supervised learning approach with the back-propagation algorithm. A multireservoir system called the Parambikulam Aliyar Project system is used for this study. The performance of the new multireservoir model is compared with (1) the regression-based approach used for deriving the multireservoir operating rules from optimization results; and (2) the single-reservoir dynamic programming-neural network model approach. The multireservoir model based on the dynamic programming-neural network algorithm gives improved performance in this study.
Citations
More filters
TL;DR: The main aim of this study is to explore the efficiency and effectiveness of genetic algorithm in optimization of multi-reservoirs and found it effective and can be utilized as an alternative technique to other traditional optimization techniques.
Abstract: Application of optimization techniques for determining the optimal operating policy of reservoirs is a major issue in water resources planning and management. As an optimization Genetic Algorithm, ruled by evolution techniques, have become popular in diversified fields of science. The main aim of this study is to explore the efficiency and effectiveness of genetic algorithm in optimization of multi-reservoirs. A computer code has been constructed for this purpose and verified by means of a reference problem with a known global optimum. Three reservoirs in the Colorado River Storage Project were optimized for maximization of energy production. Besides, a real-time approach utilizing a blend of online and a posteriori data was proposed. The results obtained were compared to the real operational data and genetic algorithm was found to be effective and can be utilized as an alternative technique to other traditional optimization techniques.
96 citations
Cites background from "Multireservoir Modeling with Dynami..."
...Chandramouli and Raman (2001) extended the study of Raman and Chandramouli (1996), developing operating rules for multireservoir systems....
[...]
TL;DR: The results indicate that FR is useful to derive operating rules for a long-term planning model, where imperfect and partial information is available, and ANFIS is beneficial in medium-term implicit stochastic optimization as it is able to extract important features of the system from the generated input-output set and represent those features as general operating rules.
96 citations
TL;DR: In this article, a novel intelligent reservoir operation system based on an evolving artificial neural network (ANN) is proposed, where the parameters of the ANN model are identified by the GA evolutionary optimization technique.
94 citations
Cites methods from "Multireservoir Modeling with Dynami..."
...Raman and Chandramouli [24] and Chandramouli and Raman [4] proposed the use of ANN to generate operational strategies trained based on the optimal results from a deterministic Dynamic Programming (DP) model, for the case of a single and multiple reservoir system, respectively....
[...]
TL;DR: The Q‐Learning method in reinforcement learning is demonstrated on the two‐reservoir Geum River system, South Korea, and is shown to outperform implicit stochastic dynamic programming and sampling stochastically dynamic programming methods.
Abstract: [1] Although several variants of stochastic dynamic programming have been applied to optimal operation of multireservoir systems, they have been plagued by a high-dimensional state space and the inability to accurately incorporate the stochastic environment as characterized by temporally and spatially correlated hydrologic inflows. Reinforcement learning has emerged as an effective approach to solving sequential decision problems by combining concepts from artificial intelligence, cognitive science, and operations research. A reinforcement learning system has a mathematical foundation similar to dynamic programming and Markov decision processes, with the goal of maximizing the long-term reward or returns as conditioned on the state of the system environment and the immediate reward obtained from operational decisions. Reinforcement learning can include Monte Carlo simulation where transition probabilities and rewards are not explicitly known a priori. The Q-Learning method in reinforcement learning is demonstrated on the two-reservoir Geum River system, South Korea, and is shown to outperform implicit stochastic dynamic programming and sampling stochastic dynamic programming methods.
89 citations
Cites methods from "Multireservoir Modeling with Dynami..."
...Application of more robust rule inference methods such as fuzzy rule-based systems [Shrestha et al., 1996] or artificial neural networks [ Chandramouli and Raman, 2001 ] may enhance the successful application of ISO....
[...]
TL;DR: A parallel dynamic programming algorithm to optimize the joint operation of a multi-reservoir system using a peer-to-peer parallel paradigm based on the distributed memory architecture and the message passing interface (MPI) protocol is developed.
81 citations
References
More filters
TL;DR: The objective of this paper is to review the state-of-the-art of mathematical models developed for reservoir operations, including simulation, which include linear programming, dynamic programming, nonliner programming, and simulation.
Abstract: The objective of this paper is to review the state-of-the-art of mathematical models developed for reservoir operations, including simulation. Algorithms and methods surveyed include linear programming (LP), dynamic programming (DP), nonliner programming (NLP), and simulation. A general overview is first presented. The historical development of each key model is critically reviewed. Conclusions and recommendations for future research are presented.
1,345 citations
TL;DR: A neural network is developed to forecast rainfall intensity fields in space and time using a three-layer learning network with input, hidden, and output layers and is shown to perform well when a relatively large number of hidden nodes are utilized.
675 citations
TL;DR: In this paper, the authors present a survey of dynamic programming models for water resource problems and examine computational techniques which have been used to obtain solutions to these problems, including aqueduct design, irrigation system control, project development, water quality maintenance, and reservoir operations analysis.
Abstract: The central intention of this survey is to review dynamic programming models for water resource problems and to examine computational techniques which have been used to obtain solutions to these problems. Problem areas surveyed here include aqueduct design, irrigation system control, project development, water quality maintenance, and reservoir operations analysis. Computational considerations impose severe limitation on the scale of dynamic programming problems which can be solved. Inventive numerical techniques for implementing dynamic programming have been applied to water resource problems. Discrete dynamic programming, differential dynamic programming, state incremental dynamic programming, and Howard's policy iteration method are among the techniques reviewed. Attempts have been made to delineate the successful applications, and speculative ideas are offered toward attacking problems which have not been solved satisfactorily.
524 citations
TL;DR: A broad array of computer models have been developed for evaluating reservoir operations as discussed by the authors, and selecting a modeling and analysis approach for a particular application depends upon the characteristics of the reservoir characteristics.
Abstract: A broad array of computer models has been developed for evaluating reservoir operations. Selecting a modeling and analysis approach for a particular application depends upon the characteristics of ...
494 citations
TL;DR: In this article, a back-propagation neural network is trained to predict the peak discharge and the time of peak resulting from a single rainfall pattern, and the neural network was trained to map a time series of three rainfall patterns into a continuum of discharges over future time by using a discrete Fourier series fit to the runoff hydrograph.
Abstract: Spatially distributed rainfall patterns can now be detected using a variety of remote–sensing techniques ranging from weather radar to various satellite–based sensors. Conversion of the remote–sensed signal into rainfall rates, and hence into runoff for a given river basin, is a complex and difficult process using traditional approaches. Neural–network models hold the possibility of circumventing these difficulties by training the network to map rainfall patterns into various measures of runoff that may be of interest. To investigate the potential of this approach, a very simple 5 × 5 grid cell synthetic watershed is used to generate runoff from stochastically generated rainfall patterns. A back–propagation neural network is trained to predict the peak discharge and the time of peak resulting from a single rainfall pattern. Additionally, the neural network is trained to map a time series of three rainfall patterns into a continuum of discharges over future time by using a discrete Fourier series fit to the runoff hydrograph.
316 citations