A Model for Solving Multiperiod Multireservoir Water Resources Problems with Stochastic Inflows.
Abstract
The model developed solves the multiperiod multireservoir water resources problem with stochastic inflows. Of unique importance is the development of a generalized network model which solves nonlinear nonseparable quadratic problems. Quadratic functions are used to measure the future value of water to the system. The nonseparable form stems from the realization that interaction exists between the benefits to be gained from a multireservoir system. Historically this interactive nature has been ignored due to the computational difficulty of measuring and solving such relationships. Also developed is a stochastic dynamic programming approach which utilizes the results to the network optimization as data for a least squares regression analysis. A quadratic function is fit to this data and is used to represent the future value of water to the system for the next period in the dynamic programming approach. This functional representation of the future value of water replaces the standard discrete matrix representation of dynamic programming and greatly reduces the dimensionality problems associated with the dynamic programming approach. In the end, this work represents a rare combination of generalized-nonlinear network flow programming, stochastic dynamic programming and regression analysis.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1981
- Accession Number
- ADA106769
Entities
People
- Douglas Dewitt Cochard
Organizations
- Air Force Institute of Technology