A Residential Energy Control Algorithm Assessment Tool for Smart Grid: Multi-Criteria Decision Making Using the Analytical Hierarchy Process

For homes to become active participants in a smart grid, intelligent control algorithms are needed to facilitate autonomous interactions that take homeowner preferences into consideration. Many control algorithms for demand response have been proposed in the literature. Comparing the performance of these algorithms has been difficult because each algorithm makes different assumptions or considers different scenarios, e.g., reducing the peak load, minimizing cost in response to the variable price of electricity, minimizing energy, or achieving a balance between overall energy savings, ensuring comfort, and minimizing cost. A comprehensive framework for assessing the performance of these algorithms that considering simultaneously considers multiple objectives and users’ subjective preferences has not previously been studied and it is necessary to be able to compare their performances. To overcome these limitations, a flexible assessment framework using the Analytical Hierarchy Process was developed to compare and rank residential energy management control algorithms. The framework is a hybrid mechanism that derives a ranking from a combination of subjective user inputs, representing preferences, and objective data from the algorithm performance related to energy consumption, cost and comfort. The Analytical Hierarchy Process results in a single overall score used to rank the alternatives. Testing and validation of the assessment framework is illustrated by applying the assessment process to six residential energy management control algorithms. The control algorithms were developed and tested using a simulation model of the Net-Zero Energy Residential Test Facility located on the campus of the National Institute of Standards and Technology in Gaithersburg, MD. The Net-Zero Energy Residential Test Facility is a research house that is comparable in size and aesthetics to the houses in the greater Washington DC metro area. One algorithm was designed to match a real heat pump controller used in the house model. A second was the same as the first with relaxed comfort deadbands. Four others used linear integer optimization with varying optimization objectives to generate forecasted heat pump control actions. The algorithms were compared by analyzing their performance over a year based on energy consumption, cost, and comfort as measured by predicted mean vote and predicted percentage of dissatisfied. Successful implementation of the assessment framework produces a figure of merit that enables policy makers, control algorithm engineers, and other stakeholders to compare the performance of residential energy management control algorithms.