Some analytical tasks can require weeks of development and testing and may include totally original statistical processes, designed specifically for the project concerned.
In order to justify the level of government funding that would be needed over the next 5 years, a water company required an economic model of its sewage treatment works, covering their operational and capital investment costs.
A new power plant was commissioned with a variety of power and heat generation systems. A model was required that would not only represent the plant's whole operation but also enable the operators to run each system in the plant in the most cost-effective manner, given short-term forecasts of the electricity and gas market prices.
Sewer blockages are an unpleasant fact of life. Their treatment is expensive and time-consuming and can be carried out in a number of different ways with several techniques. A model was needed that would allow each water company to enter its own data and to generate a cost-benefit comparison of alternative ways of maintaining a high level of customer service.
Doubts have arisen about the validity of the government-sponsored forecasts of housing requirement. Using a technique known as Monte Carlo, I designed and built a model for estimating the uncertainty in the forecasts. I then used this model to demonstrate the great danger of planning 20 years ahead on the basis of highly volatile and unreliable forecasts.
Stephen's contributions were invaluable and we are likely to refer to them in later sessions (ie of the Examination in Public of the draft Regional Spatial Strategy for the South West)'. Geoffrey Sworder, leader CPRE South West Housing topic group http://www.cpresouthwest.org.uk/
A major electricity supply company required a system to enable it to forecast the consumption of electricity by a portfolio of commercial and industrial users. I designed, developed and built a totally original forecasting tool, in a spreadsheet, with thousands of lines of VBA code. This acted as a test-harness and was later used as a standalone system for evaluating the cost of errors in forecasting consumption by new customers. The main algorithm was re-coded in a database and is currently saving them over a million pounds each year.
