GRIDfast

The advent of Automatic Meter Reading (AMR) is now giving utilities vast amounts of load information about their customers which, if properly managed, can not only be used to shape new regulatory strategies (one of the main drivers for AMI), but can be used to develop more accurate load models of the system allowing more detailed measurement and analysis of the distribution grid. In the past, utility companies determined customer power usage by sending a “meter reader” out to the meter on the house and writing down that number. Eventually, this myriad collection of data finds its way back to the billing system where it is entered into a collection system and a bill generated for the customers usage. This is typically a thirty day cycle for the utility with no real validity other than approximation on the usage patterns of customers and the coincidence of load between customers (to generate a system peak type measurement) making the load data of minimal value for system analysis without significant assumptions.

Utilities have longed for a way to determine and then to optimize and control the performance of the distribution grid. However, there were four major obstacles that prevented this grid performance optimization from occurring. These obstacles were grid data visibility, network model complexity, computing power to run the model in a near real-time mode, and the ability to incorporate distributed and renewable energy sources.

Grid Data Visibility: With the phenomenal acceptance of AMR and AMI, the utility company now collects all of those data points periodically (typically once every fifteen or thirty minutes) with an actual time stamp rather than once a month with no knowledge of usage at a specific time during the month. This results in 3,000 times more data than the utility company had before. Plus, it is digitally collected thereby dramatically improving the accuracy and the new “smart meters” that achieve this effort also offer two way communications between the customer and the utility company.

Model Complexity: The second reason is simply the complexity of the model. The GRIDfast is an incredibly sophisticated engine capable of modeling each and every component of even the largest distribution networks. There are hundreds of thousands of measurements and conditions that it must process to calculate the GRIDfactor. It has taken GRIDiant over four years to develop this model. GRIDiant performed numerous tests of its system with live data over the past several years to prove its value as a product for utilities of all sizes.

Computing Power: The third reason is computing power. This model has to run incredibly fast so that it can provide the Network Operations Center (NOC) with real-time data, grid profiles and GRIDfactors within a timeframe that allows the operators to make good decisions and make them in real-time. The computing power has only become available to do this in the last few years. Making data available to the system operators through the use of dashboards to alert the operators on system abnormalities (for example voltage deviations or over or under loaded conditions) allows the operators to react to system events in a proactive manner thereby reducing the stress on equipment and improving overall customer satisfaction.

Distributed Energy Sources: A new challenge in control and optimization is the ability to manage a distribution grid that will contain significant amounts of distributed energy sources. Distributed energy sources present new problems for the distribution networks in that they typically apply power downstream from a substation, apply a random amount of power based on available sun light or wind conditions, and are not controlled by but must be accounted for by the utility. The push for a “Smart Grid” requires the grid to accept renewable and distributed energy sources at the distribution level resulting in the operators of the distribution system needing to have better visibility on the loading and stress levels being incurred and the impact the distributed energy points has on the grid.

The following are the control devices used to achieve optimization:

Generator active and reactive power generation
Switchable capacitors and static VAR devices (SVDs)
Tap changing transformers and voltage regulators
Distributed generators (DG)
Demand response units (DR)
Storage devices (charging and discharging)

The following constraints are considered:

Voltage limits
Generator active and reactive power limits
Capacitor and SVD limits
Tap changer and voltage regulator tap limits
DG/DR and storage device generation limits

The GRIDfast engine is a state-of-the-art enterprise application that is based on a sophisticated non-linear mathematical model fine-tuned and perfected in seven major US utility operations. It runs on Linux for large-scale, massive data collection systems or on Windows for small municipals and coops in an SaaS (Software as a Service) environment. The code was written in Java. The overall structure of GRIDfast is illustrated below:

GRIDfast Architecture

New objectives, constraints and control device types can be easily added to GRIDfast due to its modular nature. Due to computational speed requirements, the engine modules are coded using the C/C++ language. Although J2EE would be the preferred development language, the speed requirement is crucial for real-time control of the grid health so C/C++ was chosen. Communication between the engine and the rest of the system is achieved using an XML messaging protocol or using Java Native Interface (JNI).

We are very pleased to state that the US Patent Office has awarded GRIDiant a patent for its GRIDfast optimization engine.