Developing countries commonly suffer from power shortages. When distribution companies seek to improve their networks, it can be difficult to be certain of how much power their customers would use if it was always available (the underlying demand). There is a risk that investors will be inefficient if the true level of underlying demand is significantly different from the forecast.
To address this challenge, we have developed energy demand modelling tools that apply a structural model to the time series of load on network feeders to estimate the current underlying total demand.
A common starting point for a distribution company energy demand study is to analyse the hour-by-hour loadings on feeders that serve their end-users. Time series data of hourly feeder loads over many days, months or years can be used to build up a picture of the daily, weekly and annual energy demand patterns respectively. The problem faced by distribution companies in Nigeria, and many other developing countries, is that due to the chronic shortages of power and in some cases unreliability of equipment, feeders are not always energized and so the underlying total demand is difficult to determine.
To combat this issue of sparseness in the data, we model the time series of a load of distribution company feeders using a statistical tool known as a “structural model”. The structural model approach calculates the total demand by forecasting the demand that would otherwise exist on the disconnected feeders if they were connected.