Distribution of bulk resources in the South Island


By Alex Dunn & Tao Vink





This project was completed as a requirement of the University of Canterbury MSCI Honours course in 2009. We were approached by Contact Energy to investigate potential changes to their distribution strategy of a specific physical commodity which, due to confidentiality agreements will be referred to as “resource”. We used optimisation to solve the Vehicle Scheduling Problem (VSP) and then simulation to analyse possible changes to Contacts’ current distribution strategy.


Problem Situation


·        The problem is restricted to the South Island of New Zealand.

·        There are two depots, which can receive and store an assumed unlimited supply of resource.

·        Due to the geographical layout of the depots and customers, the problem can be divided into two sub-problems, North and South, with a depot in each.

·        There are about 60-100 customers in each sub-problem, approximately 10-15 of which will require a delivery on a given day.

·        There are a limited number of trucks available to service these customers, with limited capacities.

·        Stock-outs occur when a customer runs out of resource. These should be avoided if at all possible.


Complicating Issues

There are a number of issues which complicate this problem:


·        Demand is highly seasonal.

·        Demand can vary significantly in the short term.

·        Some locations can only accept deliveries during certain ‘time-windows’, if a truck turns up outside these windows, then they will have to wait, or come back later.

·        Some locations take priority over others.

·        The storage capacity is different at each customer location.



Our Mission


We have been given the task of investigating potential ways to improve distribution, with the goal of reducing long term delivery costs while maintaining a high level of customer service. Delivery cost reductions can occur in two main ways:


1)      By increasing storage capacity at a number of the customer locations.

2)      By lowering the inventory re-order point for specific customers, allowing larger deliveries less frequently.


Implementing either of these two changes at any location will come at a cost (whether it be the cost of a new warehouse, or the cost of having more stock-outs) and so we must conduct a cost-benefit analysis regarding each change. This involves finding the costs associated with each change, and weighing them up against the potential long term savings that could result from that particular change.





Rounded Rectangle: Gather Data
Rounded Rectangle: Gather data on historical routes, historical demand, storage capacities of customer and delivery trucks. Inter-location driving distances were calculated and longitude and latitude positions collected for graphing.




Rounded Rectangle: Model

Rounded Rectangle: Build a VSP optimisation model in AIMMS, resembling the problem as closely as possible. This allows inventory levels to be evaluated and deliveries to be scheduled accordingly to avoid stock outs. 
This replaces the manual process and allows for fast, consistent scheduling of trucks and trip costings.






Rounded Rectangle: Perform a large number of simulated runs using the model, with current, unchanged reorder points and storage capacities, up against the same model but with changes to the reorder points and storages. (see diagram below)
Rounded Rectangle: Scenario Runs




Rounded Rectangle: Calculate the costs of each of the changes made above, along with the associated delivery costs. Calculate the long term cost savings attributable to each of the changes, taking into account the investment needed to implement such a change.
Rounded Rectangle: Analyse Results







The following is a diagrammatic representation of the simulation process. The same demand scenarios are fed into two identical copies of the model, with one using the current parameters while the other uses hypothetical parameters. Only one parameter is changed at a time in order to isolate the resulting changes in delivery costs associated with that parameter. Capital Expenditure (CAPEX) is taken into account when determining the long term feasibility of each of the parameter changes.


























Rounded Rectangle: The VSP model
Rounded Rectangle: The VSP model we created in AIMMS, with the ability to import from Excel and graph the results (which will look something like the diagram below).




Rounded Rectangle: The numerical results of the simulations, as well as a recommendations report ranking the parameter changes in terms of cost-effectiveness, with their associated potential cost savings.
Rounded Rectangle: Results of the Simulations





Example: Output graph from VSP model showing which trucks to send to which locations on a given day.





We would like to acknowledge the continued support of our supervisors, Dr John F. Raffensperger and Dr John George and all members of the department for their feedback, as well as our classmates.


Thanks also to Mark Armstrong, Bob Kooge and Trisha Upton from Contact for their help and for sponsoring the project.