Title:  Smart Energy Management in Buildings

Principal Investigators: Prof. Krithi Ramamritham (Professor, Dept. of CSE, IITB) and Prof. Ronita Bardhan (Asst. Professor, CUSE, IITB)

Funding Agency: IPHEE (INITIATIVE TO PROMOTE HABITAT ENERGY EFFICIENCY) – Department of Science and Technology, Govt. of India.


Knowing that the world is facing a major energy crisis and the fact that around 40% of total energy expenditure in developed countries is contributed by buildings, we propose “Smart Energy Management in Buildings” that can reduce the energy consumption by obtaining various insights into a building’s behavior while using a minimum number of sensors.

  • A Building Management System (BMS) controls the appliances by tracking various pieces of information like environmental parameters (temperature, humidity, etc.), occupancy status and count, energy available and cost of control. But a smart BMS should also be able to perform tasks like reducing and optimizing power consumption, monitoring the status and health of the appliances in the building, maintaining expected energy consumption in different parts of the building, profile energy consumption of different areas, identifying zones with anomalous power consumption, to name a few. Performing such tasks requires the system to sense various parameters like temperature, humidity, power consumption, occupancy status, the status of appliances and many more. The simplest and obvious solution is to place sensors, which observe these facets, at all locations. But this approach entails developing solutions that involve:
  • Appropriate sensor selection, number of sensors required and their placement location.
  • The huge capital cost of initial deployment.
  • Upgradation/Modification cost as the system progresses.
  • Maintenance cost with respect to fault/failure handling.

These problems prompt us to ask the question: “Are there any alternatives to the (hard) sensor (deployment)?”.

  • The solution while “Providing Insights” should also provide answers to the queries, fired by the user, to gain insights of the building; example of different queries are: find out per capita power consumption, average temperature of a floor, occupancy profile activity of a lab, power consumption profile of different areas, appliance used per capita, effect of change in temperature on power consumption and appliance usage, and many more. Employing many sensors also generates a large amount of data. Besides, there may be missing values and corrupted data points in the generated dataset. Also while handling the data, the system should also take care of inconsistencies that might be present. These issues further complicate the process of answering the various queries (mentioned above) with the required accuracy. These issues prompted us to ask more questions: “Will the system provide a satisfiable answer to any query fired by the user? Does the system ensure a timely response from this sensor network data to meet the requirements? How do the inaccuracies inherent in various sensors propagate through the System?”

In this project, we address the above-mentioned questions in a systematic way such that the proposed solution is scalable to any building scenario while providing insights about a building using the minimum number of sensors.

Proof of Concept

As a first step to explore the idea of Observability and monitor its advantages, we will instrument a classroom complex in our department building with various sensors and actuation systems. The classroom complex will turn into a smart classroom complex, where appliances are algorithmically controlled based on the occupancy status, thereby eliminating undesired energy consumption. In the smart classroom system, we deploy motion sensors, PIR, and Camera to obtain the Occupancy status of the room. We have already deployed Temperature and Humidity sensors to smartly control the fans and ACs in the room. We have observed energy savings up to 20- 25% with this approach and were able to achieve these savings by following the principle of Observability. The payback period of the above-mentioned system is 12 to 18 academic months.

We are also looking forward to gathering energy and power consumption data from residential houses situated on our campus. Through this data, we can understand the behavior of residential spaces and come up with algorithms which can minimize the undesired energy consumption across various spaces.

Unique and Innovative Aspects of the Project:

The project introduces the innovative concept of “Facets of Observability”, wherein we utilize the relationship between different phenomena in a building to decrease the number of physical sensors to be deployed. We use these facets of Observability to obtain meaningful insights about a building. We follow a novel approach (applicable for any building scenario) of observing one facet to infer the value of other facets, by using the available information in a systematic way. Getting insights at lower granularity entails processing a large amount of data arriving from a large number of sensors. Using a minimal number of sensors, our approaches will give the manager extensive insights into the building.