Chung-Ang University scientists develop new framework for home energy management systems


Seoul, South Korea, June 8, 2022 /PRNewswire/ — Our steady advance into the era of smart grids and smart homes presents new opportunities for more efficient use of energy. Fortunately, many of the decisions and actions needed to optimize our energy use can be delegated to home energy management systems (HEMS), which effectively manage the energy consumption of household appliances by scheduling the start-up of machines. washing and strategically turning on and activating air conditioners (AC). stopped. In general, a HEMS works to minimize electricity bills while taking into account user preferences and comfort.

Usually, handcrafted models that use abstract equations to represent devices and distributed energy resources are used to program HEMS. But these optimization models and methods are not very versatile and give sub-optimal solutions. An alternative is to use centralized machine learning, where data from thousands of users is collected, sent to a central server, and used to train a model from scratch. This strategy, however, can be expensive, computationally complex, and susceptible to hacking.

To tackle these issues simultaneously, Associate Professor Dr. Dae Hyun Choi and doctoral student Sangyoon Lee from Chung Ang University, South Korea, proposed a new data-driven strategy. The researchers developed a framework for HEMS based on federated deep reinforcement learning (F-DRL), combining the advantages of various machine learning techniques. Their study was published in Volume 18, Number 1 of IEEE Transactions on Industrial Computing in January 2022.

The key word of note in F-DRL is “federated,” which indicates a decentralized form of machine learning. In the proposed framework, each home has an EMS connected to various devices and appliances, which collects data on the energy consumption of its users and tries to optimize a schedule for the appliances by creating a local model. These local models are all uploaded to a global server, which averages them to produce a global model. Then, each HEMS replaces its local model with the global model and re-trains it using local data. This process is repeated several times, gradually improving the accuracy of the global and local models. “In a typical centralized DRL model, the global server must have access to data from all local devices to generate the global system model. This causes data privacy issues for local devices,explains Dr. Choi, “However, in our federated DRL method, the system does not require user data sharing because only local and global model parameters are exchanged. In turn, this helps prevent local data leaks and protects user privacy.

The researchers tested their approach through simulations, highlighting its optimal performance when planning the operation of various devices in different homes. “To our knowledge, this is the first HEMS framework based on a federated DRL capable of managing the energy consumption of several smart homes and ensuring consumer comfort while taking into account their preferences in a distributed way,observes Dr. Choi. Besides its low computational complexity and relatively fast training process, the proposed framework can easily support the addition of additional devices in each home.

Dr. Choi envisions a more comprehensive version of this framework that also takes into account electric cars and energy exchanges between households. We certainly keep our fingers crossed for the safe optimization of energy consumption in the future.


Original Article Title: Federated Reinforcement Learning for Energy Management of Multiple Smart Homes with Distributed Energy Resources

Log: IEEE Transactions on Industrial Computing


About Chung-Ang University


Media contact:
Seong Kee Shin
[email protected]

SOURCE Chung-Ang University


Comments are closed.