AI-driven Integration of Energy Storage Control and Building Monitoring

Case study of S&P Global Factory (Korea)
test
Client S&P Global Factory in Korea
Project Title Battery Control for S&P Global Factory in Korea
Duration/Time September 2018
Summary

We estimated that AI-driven integration with energy storage control and building monitoring can save 21% of its electricity cost.

 

Energy Storage for Smart Building Management

A growing segment in smart buildings uses energy storage systems such as batteries to manage building loads and to reduce electricity costs. Energy storage systems can reduce electricity costs through peak demand management, offsetting energy use, and integrating with distributed resources such as renewable generation. A building optimization strategy must manage the interplay of data between the energy storage system, schedulable loads, current and forecasted building operation and distributed generation. In order to be successful, data on how a building is currently and expected to consume electricity are crucial inputs for storage system software.

Potential Savings- A Case Study

Verdigris and S&P Global, Ltd partnered on a proof-of-concept to determine the potential savings from battery integration with building monitoring. Verdigris deployed systems at an S&P Global factory in Korea to measure real-time total energy consumption at the building, panel, and circuit level. The facility also has a battery system with energy storage software. Together the teams developed algorithms that integrated the S&P battery system with Verdigris data and models to control the battery charge and discharge sequence for optimum building management.

Part 1: Data, the Necessary Ingredients

There is a complex exchange of data from four areas: a) building energy measurements, b) building energy forecasts c) battery information acquisition and control, and d) electricity rate schedule.

  1. Building energy measurements
    The Verdigris data platform provides the current state of building energy and power consumption through high frequency measurements of energy, voltage, and power quality.
  2. Building energy forecasts
    Verdigris uses AI-driven models to forecast the load profile of a building. Neural network models provide day-ahead forecasts of the monitored load in 15 minute intervals. The model incorporates inputs such as historical energy use, weather data, and time of day. The model is regularly retrained and queried.
  3. Battery information acquisition and control
    S&P Global provided access to their data and models of the current and forecasted battery states. A key feature is the ability to submit control commands to the battery. A set of battery specifications establishes constraints in the optimization strategy:
    1. maximum power capacity of the battery
    2. minimum power capacity of the battery
    3. no flow of power back to the grid
    4. battery energy level should not max out its energy capacity
    5. energy stored in the battery should not be less than 10 percent of its energy capacity

      Battery specifications:

      Battery Max Capacity 18.0 kWh
      Battery Min Capacity 0.0 kWh
      Max Charging Power 9 kW
      Max Discharging Power 9 kW
      Initial Battery Capacity 1.84 kWh
  4. Electricity rate schedule
    There are two parts to electricity cost: energy and peak demand. The optimization goal is to minimize the electricity cost during a billing cycle. The optimization model accounts for both energy and demand costs.

    Electricity rate schedule:

      Energy Rate ($/kWh) Demand Rate ($/kW)
    Summer 0.17113 19.85
    Winter 0.13174 11.96

Part 2: Al-models, The Secret Sauce

The algorithm runs daily and generates a battery usage plan for the following day to optimize electricity costs. The battery control commands are automatically issued every 15 minutes.

Forecasted Building Profile Proposed Battery Charge & Discharge Plan
Forecasted Building Profile Image Battery Profile Image
Optimized Building + Battery Load Profile  
Combined Profile Image  

Part 3: Cost Savings and Smart Buildings, Just the Beginning

For this case study, we estimated that AI-driven integration with energy storage control and building monitoring can save 21% of its electricity cost.

Electricity cost summary:

Daily Forecasted electricity cost $258
Daily Forecasted electricity cost with battery integration $203
Estimated Monthly Savings $1650
Daily savings 21%

AI applications are advancing at a rate where we can envision even greater systems integrations that magnify the potential for electricity savings. For example distributed generation using photovoltaics, fuel cells, wind, geothermal, etc. can generated electricity where it is used. Additionally, the adoption of electric vehicles is bringing the transportation sector to the built environment and efficient, dynamic car charging will also require algorithmic integration.

Energy storage and control integrated with building monitoring is only the beginning. The future of smart buildings holds multiple needs for a Verdigris platform providing building data and AI-driven models.