Summary
In this study, the authors employ a simulation-optimization approach to develop a model for reservoir operation. The system under examination is Chennai, India, and the surrounding reservoirs which supply its water system. In order to overcome the computational difficulty of simulation-optimization models in these types of problems, the authors incorporate a backpropagation neural network which is trained to approximate the simulation model. As for the optimization side of the model, the Hooke and Jeeves nonlinear approach is used.
Chennai is plagued by droughts on a frequent basis. Although the city receives 1300 mm (51 in) of rain per year, most of this received is during the monsoon season, which occurs between October and December. The ultimate goal of the authors' study is to reduce water deficits through optimal reservoir operation. If the erratic rain pattern makes these deficits unavoidable, the authors' secondary goal is to make the deficits less critical by providing water on a consistent basis, even though it may not completely meed demand. This process was referred to as "hedging."
Discussion
To be truthful, I wasn't able to grasp the authors' explanation of how an artificial neural network works. Nevertheless, I found their approach to be interesting, since it was able to take more complicated factors into account, such as evaporative losses and training errors related to the model. Although their approach was obviously more detailed than our homework, I found the dynamic programming approach to this problem much more enlightening. With worldwide demand continuing to grow and supplies becoming more erratic and scarce, however, a neural network approach may become more necessary in the future.
In this study, the authors employ a simulation-optimization approach to develop a model for reservoir operation. The system under examination is Chennai, India, and the surrounding reservoirs which supply its water system. In order to overcome the computational difficulty of simulation-optimization models in these types of problems, the authors incorporate a backpropagation neural network which is trained to approximate the simulation model. As for the optimization side of the model, the Hooke and Jeeves nonlinear approach is used.
Chennai is plagued by droughts on a frequent basis. Although the city receives 1300 mm (51 in) of rain per year, most of this received is during the monsoon season, which occurs between October and December. The ultimate goal of the authors' study is to reduce water deficits through optimal reservoir operation. If the erratic rain pattern makes these deficits unavoidable, the authors' secondary goal is to make the deficits less critical by providing water on a consistent basis, even though it may not completely meed demand. This process was referred to as "hedging."
Discussion
To be truthful, I wasn't able to grasp the authors' explanation of how an artificial neural network works. Nevertheless, I found their approach to be interesting, since it was able to take more complicated factors into account, such as evaporative losses and training errors related to the model. Although their approach was obviously more detailed than our homework, I found the dynamic programming approach to this problem much more enlightening. With worldwide demand continuing to grow and supplies becoming more erratic and scarce, however, a neural network approach may become more necessary in the future.
4 comments:
I agree. The explanation of Artificial Neural networks in the article was very unclear. Dr. Z's explanation on monday cleared up a few of my questions, but I look forward from gaining more insight during Wednesdays discussion.
Well I guess a class dedicated to neural networks wont be a bad idea. Whose with me?
Andrew, don't expect much. :)
P.S. Another amazing post, Ian.
Post a Comment