Fluxys operates as an independent natural gas transportation company at the centre of the European gas market. The Fluxys network and the Zeebrugge Hub are the very heart of crossborder natural gas flows in Western Europe, which has made Fluxys a key player in the gas operations services. The Zeebrugge hub, owned and operated by Fluxys, has an annual throughput of around 1412.6 MMMSCF (40 BSCM) - connecting to 17 neighbouring networks and about 11% of the total demand for Western Europe. In the wake of gas market deregulation in Europe, Fluxys is committed to safe, reliable and flexible transportation of gas. To effectively anticipate the requirements of gas demands on its network, Fluxys use a Neural Network based forecaster. This strategic tool furnishes Fluxys with the information required to make critical decisions, resulting in greater flexibility, agility, and increasing customer service. Although traditional methods of forecasting demand, such as Regression Analysis or Neural Networks, have proven to be very useful in the management of gas supply, they are highly dependent on the availability of influencing factors such as weather, calendar, economics and production plan. The quality of data available to the forecasting tool as well the utilization of that data within the tool plays a key role in the level of accuracy of the forecast. As such, if necessary data is unavailable, of poor quality or used inappropriately then the accuracy of the forecast is adversely affected, which has a direct impact on the business. This paper examines various ways in which the data quality and utilization can be improved to more accurately predict gas demand. Methods to improve forecasting include intelligent selection of the Neural Network model, combination of weather effects, data cleaning (data pre-processing) and within-day forecasting after abrupt changes (data postprocessing). These techniques involve calculated manipulation of both the input and the output data with a view to enhance the accuracy of the forecast. Fluxys incorporates this methodology as an integral part of the forecasting system to reduce supply risk and to drive operational efficiencies by significantly reducing forecast unavailability. Each of these methods is described and the effects of applying each method and subsequent improvement of forecast are given. We also show that the correct use of each of these methods can lead to a consistent forecast accuracy of more than 95%.
Technological advancements have allowed forecasters to choose forecasting methods and techniques suited to their particular application within the business. Accurate Short-Term forecasting not only demands choice of proven methods but also involves direct user intervention to streamline the utilization of the data. Using appropriate and timely data sets enhance the effectiveness of the chosen forecasting mechanism - the result is increased accuracies and productive business planning. Achieving a reliable daily forecast is rather easy but achieving the same quality of hourly forecast is a real challenge for any gas dispatching team.
