Demand-Supply Optimization with Risk Management for a Multi-Connection Water Reservoir Network

TL;DR

This paper introduces a novel convex programming framework for long-term demand-supply optimization in multi-connection water reservoir networks, effectively managing transfer flexibility and risk to improve water resource planning.

Contribution

It presents a new convex optimization approach that handles multi-connection reservoirs and risk considerations, unlike previous models with fixed transfer paths and static plans.

Findings

Framework efficiently finds optimal solutions.

Handles complex multi-connection transfer networks.

Shows promising results in experiments.

Abstract

In this paper, we propose a framework to solve a demand-supply optimization problem of long-term water resource allocation on a multi-connection reservoir network which, in two aspects, is different to the problem considered in previous works. First, while all previous works consider a problem where each reservoir can transfer water to only one fixed reservoir, we consider a multi-connection network being constructed in Thailand in which each reservoir can transfer water to many reservoirs in one period of time. Second, a demand-supply plan considered here is static, in contrast to a dynamic policy considered in previous works. Moreover, in order to efficiently develop a long-term static plan, a severe loss (a risk) is taken into account, i.e. a risk occurs if the real amount of water stored in each reservoir in each time period is less than what planned by the optimizer. The multi-connection function and the risk make the problem rather complex such that traditional stochastic dynamic programming and deterministic/heuristic approaches are inappropriate. Our framework is based on a novel convex programming formulation in which stochastic information can be naturally taken into account and an optimal solution is guaranteed to be found efficiently. Extensive experimental results show promising results of the framework.