Benefits of Continuous Commissioning Provides automated monitoring Looks at every single point every single day Uses mathematical functions to find anomalies Prioritizes these anomalies based on cost and time in the queue • Issues work orders through email • Tracks issues till they are resolved • •
The Goal… Should Never Have To Look At Another Trend Graph 80 70 60 50 40 30 20 258 AH 1 OUTSIDE TEMP 10 258 AH 1 MIXED TEMP 258 AH 1 RETURN TEMP 0 Confidential 1/23/07 17: 16 A 1/P 1 1/23/07 14: 52 A 1/P 1 1/23/07 12: 28 A 1/P 1 1/23/07 10: 04 A 1/P 1 1/23/07 7: 40 A 1/P 1 4/08/09 2
Example Large Chain Convenience Store • Each site has around 60 interesting points, both commercial refrigeration and HVAC • There are 1, 200 sites being monitored which relates to 72, 000 points • Let’s say conservatively it would take 5 minutes to look at each point every day (not including reporting, dispatching and prioritizing) • That means it would take 750 highly skilled engineers working 8 hour shifts to do what can be done automatically every day
What Systems Should be Monitored? • All types of chillers (Centrifugal, Absorption, Rotary Screw, Scroll, Reciprocating) • All types of unitary systems • All types of Air Handling Units • All Types Of Variable Air Volume Boxes • Commercial refrigeration • Lighting • Power
Proposed Process • Hierarchical • Sensors • Rule Based Engineering • Neural Nets 4/08/09 Confidential 5
Why use neural-networks? • Problem: The only thing the industry knows about an asset is how it performed in a laboratory when it was tested. (example ARI Standard 550/590) • Once installed in the field, all bets are off. • Need a method that uses the ARI/DOEII standards as the starting point yet able to construct a true model of performance. • Even with a general form of the model though, system identification can be tricky and rarely automated • Hand tuning of models/parameters is not practical for scalability/robustness
Why use neural-networks? • Need a Universal Approximator – That can be automatically tuned
Why use neural networks? • Constant improvement • Use multilayer feedforward neural networks – Proven mathematically in 1989 to be a universal approximator of any continuous nonlinear function
Building KW • Goal: spot anomalies in day-to-day energy use – Lights left on overnight – High-demand activities at wrong time – Heat/Cool during unoccupied times • Available data: – Hourly weather data – interval building data (KW)
Building KW • Neural Network Inputs – Weekday (enumeration) – Hour of the Day – Barometric pressure – Outdoor dry-bulb temperature – Outdoor web-bulb temperature – Solar radiation • Neural Network Output – Predicted (KW)
Building KW
Building KW When Things Go Wrong
In Conclusion Recommended Process • Institute automated continuous commissioning prior to on site commissioning • Do on site commissioning to include those recommendations found in the automated process • Automated monitoring will identify both things within the site personnel capability and when a professional needs to come back (no time base commissioning)
Questions?
Скачать презентацию Benefits of Continuous Commissioning Provides automated monitoring Looks
1b0a2550ae048c779d1c2b47307a6d62.ppt