Many existing and planned manufacturing and commercial facilities are designed with extensive central control and/or monitoring systems for their HVAC equipment. A quasi-real-time expert system installed in a manufacturing plant to increase the effectiveness of the data collected is described. This system consists of two major components. The first is a statistical analysis preprocessor which screens the incoming data and estimates system operating parameters. The second component is a rule-based expert system which analyzes the collected data in terms of the estimates and expected operating conditions. Hourly reports are prepared which identify possible problems requiring immediate attention. Daily reports are also prepared that summarize the urgent items, report Iower priority anomalies, and monitor actual vs. estimated energy consumption. The details of a predictor based on the use of a singular value decomposition algorithm and sample rules are given.