Forecasting with a Flash: improving Openreach service delivery with lightning data

Openreach Limited, a wholly owned subsidiary of British Telecommunications (BT), builds and maintains the UK’s digital network. Openreach operates in a strongly regulated business with strict minimum service level targets. MeteoGroup provides weather data that helps Openreach meet and surpass those targets.

Weather impacts fault intake and repair planning

Openreach works on behalf of more than 600 communication providers, including major businesses such as Sky, TalkTalk, Vodafone and BT.  It employs more than 31,000 people. The Openreach copper and fibre broadband networks are the largest in the UK and available to more than 32 million homes and business premises. 

The Openreach Service Delivery team, which employs around 18,000 engineers, looks after the local access network, completing over 30,000 provision and repair tasks per day on average. A significant proportion of these tasks are repair activities, fixing faults in the network. With most of the local access network being either overhead or near-surface underground, it is susceptible to environmental factors, with adverse weather conditions causing a significant source of day-to-day variation in Openreach’s fault intake. This leads to planning uncertainty, with potential adverse effects having an impact on service delivery, e.g. the ability to fix faults within the Minimum Service Level targets set by Ofcom, the industry regulator. This was the motivation for developing a 14-day repair forecasting model for repair intake, with MeteoGroup supplying all observed and forecasted weather data. The model utilises forecasts for key weather variables, such as precipitation, wind speed and humidity, and has been demonstrated to work well throughout the year.

Weather API

Peter Anstey, General Manager, Production Management at Openreach: “Together with MeteoGroup we are improving our forecasting of potential network faults based on detailed weather forecasts.  This enables us to better plan the availability of technical resources and resolve issues for our customers more quickly.  We operate in a highly regulated market and improving the service we offer to our customers is extremely important to us. That’s why we’re continuously working to improve the way we plan and allocate our engineering resources.  Better forward views of demand mean that we can deploy the right skills to the right place at the right time to deliver for our customers. Using weather data, both historical and forecast helps us to better predict faults.”

Forecasting Lightning

However, lightning storms, which occur in the UK primarily during the summer months from May to September, caused additional peaks in fault intake that are not covered by the existing data and models. This led to poorer forecasts, and therefore plans, during periods following lightning storms. This was the motivation for an additional project to incorporate lightning information into the forecasting system. Meteorologically, lightning is one of the harder parameters to forecast due to the capricious nature of thunderstorm phenomena. Therefore, the initial phase of the project focused on estimating the impact of observed lightning on future fault intake. As the timing of fault arrivals is determined by a combination of physical effects, customer behaviour, and business processes, adverse weather conditions can impact fault intake many days into the future. Hence, there is merit in being as accurate as possible in profiling the impact of lightning and incorporating lightning-strike data into the forecasts communicated to the planning teams within Openreach.

Working in close collaboration with MeteoGroup, lightning detection data is sourced from a network of advanced ground-based sensors operated by NowCast. MeteoGroup delivers a near-continuous stream of data consisting of the timing and location of detected lightning strikes, along with additional metrics such as amplitude, polarity and type (i.e. cloud-to-cloud or cloud-to-ground). Whilst other sources of lightning data are commercially available, the key advantage of the LINET network is that it is able to achieve extremely high precision, with lightning strikes typically geo-located to within 50m of occurrence, and in many instances to within 10m.

Peter Anstey: “We use spatial mapping techniques to calculate the proximity to infrastructure and map strikes to Openreach planning regions. The high accuracy of the lightning location is key to the feasibility of this approach. From this mapping, we can calculate a range of different aggregates of lightning metrics, depending to types of strikes, distance to infrastructure, and time of day.”

Continuous enhancements of weather model

The forecasting model was used to communicate the expected repair intake profile to Openreach’s resource planning teams during and after the severe lightning storms of the first half of 2018. As a result, Openreach adjusted its repair plans to mitigate against the impact of lightning-strikes on repair service levels over the periods following the lightning storms. Anstey concludes: “We operate with a strong service imperative, being right the first time results in significant service improvements.”

“We aim to improve our service performance year on year.  But in a regulated marketplace we need to ensure we do that efficiently.  The better we forecast, the better we can deliver service and do so efficiently.  We are now in the process of further enhancing our lightning-enabled forecasting model by integrating it fully with our strategic forecasting and planning system.”

The teams in Openreach, BT Applied Research, and MeteoGroup will continue working together to improve the models and forecasting algorithms and to enhance the data used in the forecasts.

 

Together with MeteoGroup we are improving our forecasting of potential network faults based on detailed weather forecasts.  This enables us to better plan the availability of technical resources and resolve issues for our customers more quickly.