The emergence of smart grids requires a systematic approach to verifying certain scenarios. We wish to make a framework where the effects of an imperfect network can be seen on the performance of a control algorithm. We will then attempt to improve the performance by taking network conditions into account.
In this report we define a scenario for the smart grid, wherein a general architecture for the communication, control, and power distribution is defined. We then describe the control algorithm that will be used in the defined scenario. This control algorithm is a market controller that attempts to regulate the power consumption during peak hours, by storing thermal energy in refrigerators when there is low power demand and using this stored energy when there is a high power demand. A performance metric have been defined which shows the impact of an imperfect network in the simulations.
In order to improve performance of the control adaption two methods have been considered, one where the control parameter value is scaled according to the packet loss, and one where the control message is transmitted several times, based on the probability of information mismatch. The first method shows that performance can be increased compared to not using adaptation of the control algorithm, and the second method shows that although there is improvements in transmitting the control message several times it does not make sense to use mismatch probability in this scenario. The first method is very specific to the defined scenario, while the second method can be generalized.