Have you ever wondered how much you spent on your AC (Air Conditioner) last night? Have you ever thought how to reduce your electricity bill in summers? Every time summer comes up, and these are some of the thoughts going through our mind, but we hardly find any answers for them. Therefore to reduce their electricity bill, people start limiting their AC usage to fixed number of hours every day. But Why? Doesn’t there exist ways to optimize the AC energy consumption by changing some of those numerous settings given on an AC remote? They are tough!! Even I used to prefer to run my AC at a constant temperature (generally the lowest possible temperature I can set in my AC). But unknowingly I was increasing the contribution of my AC towards my electricity bill by 2-3 times. As shown in the figure below, when we run our AC at very low temperature (22℃ in this case), its all effort is wasted in achieving 22℃ instead of maintaining a comfortable temperature range for you. Such actions not only increased my electricity bill but also compelled me always to sleep with a blanket during the night time. I know that’s funny, but most of us practice the same.


As per a famous saying by Lord Kelvin: “If you can’t measure it, you can’t improve it”. We are living in a technological era where before even booking a cab, we never forget to check the amount we are going to spend on our travel. If that amount is out of our budget, we prefer to go for a lower budget cab or try to take cab driver through the shortest possible way. Information acts as a feedback and motivation to make optimal and cost-efficient decisions. However, when it comes to our electricity bills, we still wait for a month to know the amount of electricity we consumed. Besides, we have to rely on our power distributors for this information. But the question is, do we need to depend on them?

To better understand the current scenario, when I looked into the market I found there exist very few power monitors [1][2][3][4] that can work on 220V (voltage at which AC works). Besides customers need a plug-and-play solution that just works which is not the case with any one of them. So, there is a scope to develop such sensors but then why they are still missing in the market? One of the possible reasons is the cost-to-savings ratio. The cost involved in developing these sensors is high in comparison to potential savings by the customer. Further, the aim of an AC is to maintain room temperature and these power monitors shortfall if the user wants to know the quality of AC output i.e. room temperature.

Working on these shortcomings of existing technology, we decided to monitor and analyze the output (Room Temperature) instead of input (AC Power Consumption). We performed a study across seven different apartments and developed an estimation algorithm to report user their AC energy consumption by only using their room temperature information. Our algorithm estimated with an average accuracy of 85%, decent enough to encourage further improvements. Is it the best we could do? Definitely No! Given the existing knowledge and literature of Thermodynamics and Machine Learning, we decided to go one step further. We merged the two branches to learn a theoretical thermal model using machine learning algorithms to predict AC energy consumption for users before they use their AC. Now if I want to use my AC for next two hours at a particular temperature, PACMAN (that’s what we call our system) will tell how much you will consume for a given weather condition and that too with an accuracy of 83%. Isn’t that cool!

A temperature sensor on a device (that could be your Smartphone or Raspberry Pi) with PACMAN installed in it, and you are all set to know all the information you need regarding your AC. This temperature sensor based device is economical, easy-to-deploy and a smart move to make your AC energy efficient. To further demonstrate our work, please go through demonstration video of our current implementation. For further details of our work, please feel free to go through our paper titled “Non-Intrusive Estimation and Prediction of Residential AC Energy Consumption”. It is available on IEEE Xplore or you can also visit the Publication section to download the PrePrint version of the article. Currently, we are working on numerous limitations of the proposed study and various other exciting projects on optimizing the energy consumption of Air Conditioners, especially in Indian context. Please feel free to share your feedback, suggestions, comments, or queries either in the comment section below or you can also contact me by either way mentioned at the Contact page of my Website. 🙂

– Milan Jain, Ph.D Scholar, Energy Group, IIIT Delhi

NOTE: Numbers presented in the article indicates the performance of our algorithms on the dataset collected during the study. This study was performed In-Situ, and no specific instructions were given to users during the data collection i.e. they were free to use AC any time of the day at settings convenient to them.