• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content

Loading…

Flash Player 9 (or above) is needed to view presentations.
We have detected that you do not have it on your computer. To install it, go here.

Like this presentation? Why not share!

Artificial Intelligence Applications For Building Energy Management Applications [John Egbuta]

on

  • 1,049 views

 

Statistics

Views

Total Views
1,049
Views on SlideShare
1,049
Embed Views
0

Actions

Likes
0
Downloads
0
Comments
0

0 Embeds 0

No embeds

Accessibility

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    Artificial Intelligence Applications For Building Energy Management Applications [John Egbuta] Artificial Intelligence Applications For Building Energy Management Applications [John Egbuta] Presentation Transcript

    • UHI RESEARCH AND POST GRADUATE CONFERENCE John Egbuta (PhD Candidate : Systems Engineering-University of Aberdeen) Artificial Intelligence Applications for Building Energy Management Applications
    • The Research Question
      • How can Artificial Intelligence be used to Automate the Building Energy Module by Integrating the Building Energy Management System and Micro-generation systems on the same platform in order to achieve the Zero Energy Building status?
    • The Current Paradigm and the Proposed Shift
      • Michael Mozer’s Cost function for the Neurothermostat
      • My Proposed Cost Function for the Controller in this PhD Endeavour
        • Where relevant variables are quantized in Great British Pounds and defined as follows
          • Ec (Ut)= Energy cost as the result of the control decision (Ut)
          • M(Xt) = Misery as a result of the environmental variable (Xt) i.e. Indoor Temperature
          • Ed(Ut) = Energy debit from Micro-generation Systems
        • Note: At the moment, no known controller combines these three variables in a cost function!
    • Current Modelling Effort
      • MATLAB Model of the Model Predictive Controller
          • Relevant Parameters
            • Yr = Desired Response
            • Ym = Neural Network Model Response
              • Which is compared to the desired response to generate an error that is minimized over the Control and Cost Horizons
            • Yp is the output of the plant
            • U is the control input
      CONTROLLER OPTIMIZATION BLOCK ENERGY PLANT BEMS NEURAL NETWORK MODEL Yr Ym Yp u u’
    • Modelling: Optimization Block
      • Relevant Parameters
        • N1 and N2 are the Cost Horizons
        • Nu is the Control Horizon
          • The 3 horizons above refer to the number of discrete steps over which the error from prediction and control are minimized.
        • Yr = Desired Response
        • Ym = Neural Network Model Response
        • A backtracking line search routine is used to step through the input (u) during the error minimization process.
    • Modelling : BEMS Neural Network
      • Network Architecture
        • Feed-forward Network (Multiple Layer Perceptron) with tansig and purlin transfer functions
          • Inputs(3)
          • Hidden Layer (20 Neurons)
          • Outputs(3)
            • All inputs and outputs are TDL (Tapped Delayed Lines (TDL) are used to enable correct prediction based on past inputs to the network )
    • Modelling :Energy Plant
      • Emulated Linear Model of Proportional Control Action of a Heat Plant
        • Linear Representation of the Plant is
          • V = KE + M
            • Where:
              • V is the final Temperature output by the plant
              • K is the Gain, which is equal to the 100/Proportional Band
              • E is the error which is the difference between the Control Point and the Set Point
              • M is the bias, which is the output of the plant when the final control element is at 50% of its range
    • Final Model of Model Predictive Controller
    • Modelling Results: Plant Identification
        • There is a very close scaled relationship between the input and output data that was generated from the Model.
          • This is means the data can be used to train the Neural Network
    • Modelling Results: Neural Network Training and Validation of Generated Data
      • Performance Evaluation
        • Mean Square Error: 0.0023425
        • Plant to Neural Network Tracking: Good
      Training data Validation data
    • Controller Simulation Results
      • Green Signal is the Reference Signal (which has no variable states at the moment)
      • The Blue signal is the Plant Output Optimized to track the reference signal
        • At the moment, more work is being done to optimize this signal
        • Also, if a derivative or Integral block were included in the design of the plant model, the plant signal will track the reference signal better
    • Conclusion
      • In the end, this research endeavour will investigate how the Building Module could become
        • An Intelligent Entity (i.e. more ADAPTIVE and less reactive)
        • An Energy System (i.e. an energy asset)
    • Thank You
      • Appreciation
        • Greenspace Research LCBL
        • Hydrogen Labs at LCC
        • Dr. Alasdair Macleod
        • University of Aberdeen Supervisory Team