Network

NETWORK / 29 / FEBRUARY 2019 Forecasting now: Long term fore- casting The Bayesian-based long-term gas forecast- ing system developed by Newcastle Univer- sity has been integral to the NGN system op- erator since 2015, and is regularly reviewed and updated. The long-term forecasts for NGN's network predict gas demand daily for the North East and Northern regions for the next 10 years. They provide an answer to how much gas demand there will be for each year, determining the peak and annual demands, and then allocate gas demand for each day of each year for the 10 year period. The allocation of daily demand, driven by the Bayesian forecasting approach, must account for changes in demand between weekdays and weekends, and include variances due to seasonal, holidays and bank holiday periods. The implementation has been a major success, the approach produces more accurate demand forecasts which means the network is operated more efficiently. Forecasting now: Short-term offtake profile forecasting Oƒake Profile Forecasting (OPF) is the analysis process whereby gas demand for an area is aggregated across the gas sites (oƒakes) directly connected to the national transmission system, and which bring gas into the distribution network. Twenty three oƒake sites each require a rolling 48 hour profile forecast made up of 48 hourly supply numbers to be derived for publication to the national control centre. The existing process uses general, standardised profiles to de - velop that hourly view but this by its nature fails to account for local conditions and the unique behaviours and performance of an individual oƒake site. The new Bayesian approach, currently being trialled by the University and NGN, allows short-term forecasts to be adjusted based on previous actions, knowledge and data to calibrate the tool and make forecasts more robust. The more accurate the short- term forecasts, the less additional interven - tions are needed to manage the network, resulting in operational savings. Full imple- mentation is envisaged in Q1/Q2 of 2019. The future But what does this and similar ways to optimise the gas network mean for helping network operators decarbonise? Currently biomethane and in the future hydrogen will drive the decarbonisation of the gas indus- try to support the UK climate change obliga- tions through to 2050. This will ensure that the maximum volumes possible are injected into the gas grid. Bayesian statistical approaches are a solution. Accurate forecasting of demand will become ever more important as the gas industry manage complex mixtures of natural gas, biomethane and hydrogen; and develop their knowledge of deploying 100 per cent hydrogen networks, displacing natural gas and transitioning to a low-car - bon system for heat. During the transition process, forecasting for hydrogen demand will emerge alongside that of the exist- ing process, with accuracy, reliability and repeatability key to ensuring an efficient process. In optimising the network for hydrogen distribution there are also opportunities for system integration. For example, if an offshore wind farm generates more electricity than needed during a particular time of day and the grid is over capacity, converting that excess electricity to hydrogen (referred to as Power to Gas), offers demand side response to the electricity network, enabling more renew - able power onto the grid, whilst simultane- ously delivering a degree of decarbonisation into the gas mix. A forecasting tool that can account for the system needs across the gas and electricity sector would provide valu - able insight into the temporal and spatial behaviours of whole energy systems. Conclusion Bayesian statistical analyses can provide the whole energy sector, not just gas, with the means to make the most of their data (live and historical), system experience and process knowledge. Decarbonisation is without a doubt a major challenge for the gas industry, as 45 per cent of our energy demand in the UK comes from heating alone, of which 85 per cent is supplied by gas. If we are going to make best use of low- carbon energy vectors like hydrogen and biogas then we need in place ro - bust statistical methods for forecasting that support optimisation within the network and reduce the need for more reactive, costly methods for ensuring gas demand is met in the short and long-term. Bayesian statistical methods make possible not only more accurate fore- casting but help safeguard infrastruc- ture and enable a smooth transition to greener fuels. The industry stands to benefit from these statistical approaches in making the UK's energy infrastructure low-carbon, sustainable and secure. AUTHORS: Matt Linsley, Industrial Statistics Research Unit (ISRU) manager, Newcastle University Keith Owen, head of systems development and energy strat- egy, Northern Gas Networks Brett Cherry, science writer for Newcastle University

• Cover