We develop a scalable, computationally efficient method for the task of energy disaggregation for home appliance monitoring. In this problem the goal is to estimate the energy consumption of each appliance based on the total energy-consumption signal of a household. The current state of the art models the problem as inference in factorial HMMs, and finds an approximate solution to the resulting quadratic integer program via quadratic programming. Here we take a more principled approach, better suited to integer programming problems, and find an approximate optimum by combining convex semidefinite relaxations with randomized rounding, as well as with a scalable ADMM method that exploits the special structure of the resulting semidefinite program. Simulation results demonstrate the superiority of our methods both in synthetic and real-world datasets.

Frequency-dependent modeling of devices and systems is a common practice in several fields, such as power systems, microwave systems, and electronics systems. The modeling process usually involves converting the tabulated frequency-response data into a compact equivalent circuit model. The main drawback of the currently existing methods such as vector fitting is that the obtained model is often nonpassive, leading to unstable simulations. In order to overcome this problem, this paper proposes a genetic programming (GP) approach to generate equivalent circuits with guaranteed passivity. The proposed method starts with a nonoptimal initial equivalent circuit. Both the elements and the topology of this circuit are then evolved by the proposed GP-based method, and an accurate equivalent circuit is obtained. Key ideas and detailed algorithms are presented in this paper. Finally, the performance of the proposed method is verified by using different case studies.

One of the important issues in range query (RQ) retrieval problems is to determine the key's resolution for multi-attribute records. Conventional models need to be improved because of their potential degeneracy, less-than-desired computability and possible inconsistency with the partial match query (PMQ) models. This paper presents a new RQ model to overcome these drawbacks and introduces a new methodology, stochastic programming (SP), to conduct the optimization process. The model is established by using a monotone-increasing function to characterize range sizes. Three SP approaches - the wait-and-see (WS), here-and-now (HN) and scenario tracking (ST) methods - are integrated into this RQ model. Analytical expressions of the optimal solution are derived. It seems that HN has advantage over WS because the latter usually involves complicated multiple summations or integrals. For the ST method, a nonlinear programming software package is designed. Results of numerical experiments are presented that optimized a 10-dimensional RQ model and tracked both middle-size [100] and large-size (1,000) scenarios.