Meter Data Management Q&A Series: Data and Analytics Architectures

How distributed do you recommend for data and analytics architectures?

.

Unfortunately there is not a single answer to fit all scenarios; just as there will be no two utility Smart Grids that will be exactly the same. Smart Grid data comes from a variety of distributed sources – from customer homes with smart meters and HANs to T&D networks and utility backend systems. In addition, new distributed business applications are now emerging such as microgrids, distributed generation and PHEVs. 

Utilities are faced with questions on how to manage and use this information effectively by creating data and analytics architectures that can correlate information from these distributed sources to provide actionable intelligence. Also, we need to determine the appropriate level of central vs. distributed analytics; i.e. should analytics done at one central location in the utility backend databases/data warehouses, or can some parts of it be done in the field, closer to the devices and systems that create this data.

This decision depends on the type of the raw information and its reuse potential across more than one type of analytics. It may be beneficial to process information from smaller devices in the field at concentration points such as a substation instead of sending large amounts of raw input data to the utility back end systems. This also reduces the time between data creation and distribution operation decisions since there is a reduced need for backend systems to process large amounts of information. Telecommunication costs will certainly be a factor in this equation as well.

As an example, substation-related analytics can be applied at the substation level itself and only aggregated data and summarized fact tables need to be sent back to the utility for analytical purposes. The utility can directly use this information in a near real-time manner instead of waiting for the next analytics run across all the available data – which can be the next day in most utilities. Also, power factor analysis at the substation level can help optimize the quality of electric power down to the meter level directly.

Another factor that influences this decision of doing analytics in the field is if there is a real-time/near real-time decision that needs to be taken based on it. For example, analyzing customer usage data to identify candidate DR programs does not have any real-time application, but analysis of load of transformers on a hot summer day is near real-time.

While current technology constraints only provide such computing power at the substation level, future grid devices may have be able to pre-process the information before sending it upwards in the communication network. For example intelligent transformers that analyze their own load based on seasonally adjusted thresholds and send alerts in case they are crossed, home energy management system can calculate customer savings based on their usage profile and available DR programs and provide approximate savings to customers for making informed decisions.