An approach to estimating hydro power system income gain from computerized water scheduling

Abstract

Hydro power schemes operating in a free electricity market seek to maximize profits by differing generation rates to take best advantage of fluctuating selling prices, subject to the constraints of keeping storage lakes within their operational bounds and avoiding spillage losses. Various computer algorithms can be used in place of manual scheme operation to aid this maximization process, so it is desirable to quantify any profit gained from a given algorithm. A standard approach involves applying the algorithm to a period of past river flow records to see how much additional scheme income might have been obtained. This process requires the use of a hydro power scheme model, which inevitably can only approximate operational details, so the anticipated income gains are likely to be biased estimates of actual income gained from implementation of the algorithm. In addition to preliminary algorithm evaluation, it is desirable that hydro scheme managers have methodology to confirm anticipated income gain. Such confirmation can be difficult because true income gains are typically in the order of a few percentage and may not be easily distinguishable from background noise. We develop an approach, which allows estimation of true income gain for the situation where a change is made from manual to computer control of hydro power scheme operations, or upgrading from one maximization algorithm to another. The method uses a regression model to describe the former period of scheme operation. Postimplementation residuals from the regression predictions then provide estimates of actual income gain. The method can be sensitive to small but consistent income gains. Also, there is no requirement to construct any hydro scheme simulation model so bias effects should be considerably reduced. The approach was developed in the context of evaluating an income-maximization algorithm applied to a small hydro power scheme in the Kaimai Ranges of New Zealand. However, the methodology seems sufficiently simple and general to be applicable, with modification, to other power schemes moving toward increasing income through operational changes.