Prescriptive approaches to power system planning are gaining momentum and available data is leading the way to new methods that leverage simulation software to the fullest extent.
System planning has largely been a craft that engineers have adapted technology to, rather than technology driving the decisions. As power system modeling software has evolved, methodologies have evolved with the technologies to support more detailed models. In the distribution world, models can be very complex with multiple voltages on a single feeder with many devices contributing to the performance of the system including capacitors, regulators, motors, roof top solar and storage.
Power system modeling software is expected to behave as the live system would in a time series state. To support this framework, models now have capabilities to not only model the voltage support devices and generation but also calibrate themselves with field measurement data. For many utilities, the adoption of AMI (Automated Metering Infrastructure) has allowed them to take a deeper dive into their modeling applications and for one utility in particular we have been able to compare field measured results in semi-real time with the modeling results through an analytics dashboard platform.
Hourly data is collected at various points on the distribution system and can be compared with the models to determine which hours represent discrepancy from the actual data collected. This allows planners to refine their model settings, load allocation methods and data to more closely mirror the actual system collected data. By comparing actual, field collected data, over time, modelers can hone in on potential data quality errors. We believe machine learning capabilities can be useful in calibrating models and load data to ensure that planning models are more accurate than ever before. By developing dashboards that quickly identify which feeders have the most discrepancy to the field collected data, engineers can prioritize the data QA/QC efforts while also using the dashboards to determine which feeders may be at the highest risk based on available contingency plans and asset condition.
Data scoring will be an important consideration moving forward as stakeholders, regulators and customers continue to demand and expect the infrastructure to support advances in technologies on the grid. If we can't trust the data, we can't make the best and most timely decisions and building that trust is about understanding the behavior of our applications relative to what our network data is telling us.