Recently, Power Analytics completed a three year microgrid project with ESTCP. Read here for more about the basics of the project (http://www.poweranalytics.com/estcp-story/) and here about the implemented technology (http://www.poweranalytics.com/estcp-technology/).
The conclusions drawn in all of our studies suggest that environmental indicators and weather variables should be monitored, as they might be used as input in a set of specific applications, such as electric load forecasting. However, while our work presents several studies of the relationship between weather variables and electric load, they are usually focused on large areas and regions and are not directly portable to smaller environments like microgrids. One of the advantages of these smart systems is they are capable of providing precise answers to local problems thanks to distributed intelligence, and, as such, the objective of this work is to particularize the correlation analysis to the microgrid scale using an adequate data set, and present a microgrid design to take advantage of this data in real time.
The research completed demonstrated that potential migrogrid integration is possible within the confines of the cyber security policies and standards. With changing regulations focused on the reduction of energy consumption from nonrenewable resources, and the transition to renewable energy resources, energy producers such as photovoltaic arrays, or additional forms of distributed energy resources, would ensure reliability, energy surety, and energy reliability to the bases.
The solar array data use cases identified the need to create a solar energy production database. Because there existed no prior data on what a solar panel could produce during the time periods for the use case, a new data set had to be created. The data set required real-time point of array solar irradiance data, however, the only available data set found was “near real-time”. This issue was somewhat problematic as the data set would not exactly matched up, instead, the data sets had to be summed up to one value per day to match the newly created solar data set. With the newly created solar data set, hourly peak shaving could not be fully determined, however, as the recorded energy consumption data was time-of-day sensitive, and most of the values occur during hours of sunlight, it can be assumed that additional solar panels would reduce the energy consumption from the utility during those hours.
Therefore, some technological challenges need to be solved in order to allow a full implementation of adaptation intelligence into microgrids, opening a set of research directions. First, it is necessary to develop intelligent hardware capable of running software implementing the agents, both to complete the monitoring sensor networks and the intelligence in the microgrid. Second, most of the algorithms currently employed in power grids are designed to operate at a nation-wide scale, and it will be necessary to adapt them for operation in small environments. This is especially true (as shown in this study) for load forecasting algorithms. There is still a lot of work to do in order to be able to predict loads at small microgrids, and even single nodes, but it is also true that new control algorithms have to be designed to operate at a extremely small scale (with nodes being represented even by single devices). For this, the study reported along this work represents a very important step, since local weather variables will be an extremely valuable input for node load/production forecast.