Newly paved roads in Massachusetts are under "guarantee" for 5 years, which prohibits digging, meaning National Grid may not be able to address Cathodic Pipe Protection (CPP) compliance issues as required until the guarantee period is over for the assets needing attention.
National Grid leverages machine learning techniques to predict the probabilities of compliance issues throughout the network and prioritize monitoring efforts and work orders. National Grid illuminates the path from noisy data to a working output lighting the way as teams in the field keep gas pipelines safe and reliable. They also highlight early successes and the program’s positive effect on community relations.
- Data Science can be a messy process, finding the right approach takes patience and creativity
- Focusing on the end-user is paramount to ensuring model adoption
- Probabilities can be an effective output of classification models, as long as they are calibrated