Power & Energy Research | IPERC

Power & Energy

Intelligent Microgrid Innovations

With our academic background and military tenacity, IPERC has and continues to be a leading innovator in intelligent power and energy research. Over the years, we have been awarded numerous research and development contracts, resulting in multiple innovations being brought to market, including the fundamental architecture approach.

This unique approach, honed by IPERC team members, works beyond the standard "master-slave" approach to form a distributed control architecture using a colony or hive concept in which each node has the ability to serve as the master at any time, as needed. The distribute controls architecture can be expanded into an intelligent network that includes numerous nodes working together to deliver a more efficient and economical flow of energy.

IPERC is currently developing, testing and demonstrating multiple innovations in the areas of power line sensing, microgrid communications, active detection and software modeling. For more information on the ongoing research in these areas, please Contact Us online or call (800) 815-6183.

Energy Efficiency

Residential Fuel Cells

Fuel cells are devices that produce electricity with virtually no pollution. Three proton exchange membrane fuel cell systems have been installed and operated to provide electrical power and heat for three separate residences. The 5 kW fuel cells are configured as standby units with waste heat used for domestic hot water and space heating. The fuel cell systems operate at steady state providing a preset level of power to the grid. In case of loss of grid power, an automatic transfer switch changes the home power source from grid to fuel cell (with battery) to follow the load. IPERC has experience in installing, monitoring and determining operating strategies for fuel cells and other forms of distributed generation.

Intelligent Control for Combined Heat and Power

Control strategies implemented in construction do not consider the changes in buildings and equipment from year to year, season to season or even day to day. As a result, much of the potential cost savings of using HVAC equipment is lost. Optimal control has not been implemented because of difficulties accommodating the complex interactions between equipment. Equipment behavior is highly non-linear and it varies from one location to another, requiring experts for fine-tuning and control. Even for experts with vast experience in installing HVAC equipment, models are complex and require significant effort to calibrate. Furthermore, as equipment ages or undergoes retrofit, models that describe equipment behavior must be changed, requiring further expert assistance.

The West Point power plant is equipped with a modern neural network-based supervisory controller that "learns" plant characteristics and then notifies plant operators when items of equipment should be operated. In the winter of 2002-2003, the plant recorded $161,000 reduction in operating costs.

Pattern Recognition

IPERC's software uses a variety of artificial intelligence techniques to perform complex pattern recognition and prediction. Our prediction projects range from electrical and thermal loads to satellite imagery. The artificial intelligence techniques, in conjunction with our custom controllers, make for a powerful method to minimize your energy consumption and maximize your profit.

Key Benefits

  • Complex, non-linear models where underlying relationships are not well understood
  • Models that self-calibrate to system changes
  • Experienced in a wide variety of applications


Predicting Snow Coverage Area

Predicting Snow Coverage Area

The image is a satellite image of the King's Basin, located in California. For this problem, the goal is to determine where the snow is located on the ground (Snow Coverage Area (SCA)). If one knows where the snow is located, then it is fairly straightforward to determine the probability of flooding due to spring runoff. Unfortunately, any regional cloud cover prevents the satellite image from being useful for locating SCA. To overcome this problem, a neural network was used to predict where the snow is located on the ground. The synthetic image (right) was produced using a neural network, producing a better image than previously thought possible.

Predicting Electrical Demand

Electrical demand prediction can save users and producers of electricity millions of dollars per month if it were possible to predict when its use would be required. A neural network was used to predict the electrical load within an average of 1.6% at the United States Military Academy, West Point, NY. The graphic below shows the electrical demand profile that was created using simple, Predicting Electrical Demandeasily obtainable information. Shown on the graph is electrical demand over time. Peaks are indicative of daytime electrical demand and valleys occur at night. In this example, the first two peaks were recorded on Thursday and Friday with lower peaks following on Saturday and Sunday. Eight additional days are shown with the last peak being a Monday.

Predicting Ice Jams

Predicting Ice Jams

Breakup ice jams occur during periods of thaw when increased discharge due to snowmelt and/or precipitation cause the forces on an ice cover to exceed its strength, resulting in the breakup of the ice cover. The broken ice is transported down the river until the river's transport capacity is exceeded. This forms an accumulation that obstructs flow, creates backwater and can cause flooding. Breakup ice jams can create significantly more flooding than traditional river flooding due to the reduction in channel width and rapid rise in water levels, similar to flash floods. These rapid increases in water level can make it difficult to plan or execute ice jam mitigation measures such as evacuation or blasting. Depending on the jam characteristics, a prediction method might significantly increase warning time. Table 1 shows three methods that were used to predict ice jams (smaller numbers are better). In all cases, we see where neural networks provide better predictions than traditional statistical methods for predicting ice jam occurrences.