Before an energy model justifies a retrofit, it should match the plant's meters: ±5% NMBE and 15% CV(RMSE) on monthly data.
The usual bar is ASHRAE Guideline 14. Against monthly data, a model counts as calibrated when its normalised mean bias error (NMBE) is within ±5% and its coefficient of variation of the root mean square error, CV(RMSE), is within 15%. Against hourly data the limits are ±10% and 30%.
The two numbers catch different faults. NMBE is net bias: a model can hit the annual total by being 20% high in the dry season and 20% low in the wet season, and NMBE alone will not notice. CV(RMSE) measures the month-by-month scatter, where the seasons show. In a tropical plant the wet season is usually the one a model gets wrong, because dehumidification and reheat are easy to underestimate.
This matters before a retrofit because the savings you claim are the difference between two model runs. If the baseline run does not match the meters, the difference is a guess with decimals.
In practice that means 12 months of metered data per system, not only the site bill; submeters on the chillers and large AHUs; real schedules and setpoints; and a monthly comparison before anyone looks at options.
Our energy modeling targets a 30% reduction in HVAC energy, and a target like that means little without a calibrated baseline. To talk it through: contact@humiley.com or +84 28 7777 6668 (Viet Nam office).
Topics




