Segmented, Constrained, Non-linear, Multi-objective, Dynamic Optimization Methodology Applied to the Dairy Cow Ration Formulation Problem

摘要

In the last meeting of this series of modelling workshops we reported on the development of an optimization scheme for dealing with dairy cow ration formulation, in an optimal way, in the situation where constraints were both non-linear and discontinuous in their arguments. Indeed, we illustrated the implementation of this approach in the Cornell Pennsylvania Miner (CPM)-Dairy ration formulation system. The methodology underpinning CPM-Dairy, the Cornell Net Carbohydrate and Protein System (CNCPS) 'evaluator', projects dairy cow needs on a daily basis, and the solutions obtained by CPM-Dairy may be thought of as least-cost daily formulations. If we set ourselves the objective of, for example, balancing cost against production returns in an optimal way against the background of maintaining the health status of the animal over some period, e.g. the anoestrus period, we might want to examine a more dynamic approach in which the consequences of the time-varying pattern of dairy cow health indices are continuously monitored. In this presentation we report on an approach that addresses this problem, and we outline the issues that confronted us as we developed workable solutions.