An analysis of the optimization disaggregation algorithm in the estimation related to energy consumption of appliances in buildings

Generally speaking, energy monitoring system plays an important role in both energy use benchmarking and energy conservation service for large-scale buildings. However, it is often necessary for us to disaggregate the metered data according to types of energy use. Aiming at eliminating unclear concepts between monitoring and disaggregation and incorporating existing algorithms' advantages into the process in which this estimation is generated, the optimal disaggregation algorithm is developed. Based on the assumption of mutually independent ends, a Bayes process is degenerated into maximization likelihood estimation. Following this, the process is expressed by the least squares. With a new algorithm, an unbiased and overall convergent posterior estimation is provided for each terminal category, which is powerfully supported by theoretical analyses and numerical calculation. Actually, it is proved that the disaggregation algorithm is quite appropriate and cost-efficient for sub-item energy monitoring systems in large-scale buildings. Meanwhile, this thesis also focuses on discussing the conditional probability of overall lessening estimation errors specifically. In this way, posterior estimations of overall accuracy are available when prior estimations are concordantly biased with much different accuracy.