How Machine Learning and Big Data Can Help Manage Electricity Bills

Michael Tobias
3 Minutes Read
  • Home
  • Blog
  • How Machine Learning and Big Data Can Help Manage Electricity Bills

    When a building gets an energy audit, a key challenge is finding out how electricity consumption is distributed among appliances and equipment. Buildings normally have a main power meter, individual meters for apartments and commercial spaces, or a combination of both. However, adding a power meter to every single circuit in the installation is very expensive.

    An important step during energy audits is analyzing the individual energy consumption of appliances and equipment. However, since power bills only provide general data, energy consultants must estimate how electricity consumption is broken down. The following are some common methods:

    • Estimating electricity consumption (kilowatt-hours) based on nameplate information and usage schedules. This approach is suitable for lighting fixtures and other devices that operate at constant power. However, the method is not very accurate for equipment that runs at variable power, such as air conditioners.
    • Energy consultants may install temporary power meters for specific equipment, to measure consumption over a period of time, typically days or weeks. This information is then used to estimate annual consumption. However, some information must be assumed, such as seasonal variations in workload.
    • Another common approach is energy modeling. The building, its equipment and the local climate are simulated with advanced software to estimate energy consumption. Simulations can be very accurate, but they still rely on assumptions.

    Installing a permanent power meter for every circuit in a building is possible. However, the cost of such a project is difficult to justify from a financial standpoint. A promising solution is using artificial intelligence to analyze and break down data from a main power meter.

    Reduce power and gas bills with energy efficiency.


    Energy Disaggregation A.K.A. Virtual Submetering

    Each electrical device has a unique energy consumption pattern. For example, lighting fixtures have a constant power input, while refrigeration compressors use power intermittently as they cycle on and off. All this data is mixed together when power meters measure consumption, but it can be broken down and analyzed with machine learning - a form of artificial intelligence.

    • Artificial intelligence can “learn” the consumption pattern of each electrical device.
    • When a power meter measures the combined consumption of all appliances and equipment, AI can break down the data by device.

    Disaggregation has very promising applications in energy management: it can achieve electric submetering for different devices, without having a physical power meter for every circuit. Using this data, energy consultants can suggest the most effective measures to save energy, without having to rely on less accurate estimations.


    Disaggregation can also be an added value service, provided by electrical companies who have invested in advanced metering infrastructure:

    • Normal electricity bills only provide general information about variables like energy consumption, peak demand and power factor.
    • However, an electricity bill with disaggregated consumption is extremely useful for consumers. They can know exactly which devices are increasing their power bill, and take action.

    Power companies can also analyze data from thousands of users to identify which loads are creating the largest burden on the grid. This way, they can create effective demand response programs. Traditionally, electric companies have simply upgraded their grid capacity to keep up with demand. This leads to a constant increase in electricity prices, since power companies must recover the cost of infrastructure upgrades.

    Virtual Submetering as an Added Value Service

    Disaggregation is especially useful for the residential sector, where there is a limited budget for conventional energy audits and submetering. With this concept, each homeowner can get a detailed breakdown of energy consumption each month. However, disaggregation is also promising for businesses, who have traditionally used physical submetering to manage their consumption. Disaggregation is often called “virtual submetering”, since it accomplishes the same function as submeters without using physical devices.

    Real estate companies who use submetering for tenant spaces can also provide added value with disaggregation. Each month, tenants can get a power bill that provides insight on how to reduce consumption. 

    • Disaggregation is especially useful for buildings who split consumption equally among tenants, or based on square footage.
    • By adding power meters with virtual submetering for tenants, electricity can be charged based on the exact consumption. Tenants are also provided with useful information to reduce consumption.

    When power bills are split equally or based on floor area, there is little incentive for tenants to save energy. With this traditional approach, efficient tenants are forced to split their savings, while tenants who waste energy affect everyone else.


    Contact Us


    Leave Comment

    Please avoid adding links in comments. Any comment with external website links will not be published.

    Join 15,000+ Fellow Architects and Contractors

    Get expert engineering tips straight to your inbox. Subscribe to the NY Engineers Blog below.

    Looking for more information?
    Contact Us