Abstract: A method can include detecting a positive transition in a time trace disaggregation process (TTDP), creating a new instance j, capturing a positive transition data block for the instance j, creating a time trace feature description for power, reactive power, and current, and matching the instance j to a device type by performing a quick match based on at least one quick match parameters. Responsive to a positive identification based on the quick match, the instance j may be assigned to the device type.

Abstract: A method can include collecting time traces for N calibration instances of a device and identifying a key feature for the device based on the time traces. The method can also include using an equation to determine a key feature variation based on a predicted value and a measured value for each new qualified instance of the device. The method can also include generating an alert based on the key feature variation.

Abstract: A method of identifying multi-phase devices in a time trace disaggregation process can include detecting a multi-phase transition, identifying multi-phase devices based on the detecting, and processing the multi-phase devices separate from single-phase devices. The method can also include initializing a multi-phase array and, for each of a plurality of supply legs, determining whether a qualified transition exists. If so, the method can further include determining whether there is a pending multi-phase transition. If so, the method can further include determining whether the pending multi-phase transition is positive. If so, the method can further include creating a multi-phase instance. The method can also include storing the multi-phase instance in the multi-phase array.

Abstract: A method can include receiving raw data having power samples that correspond to various types of electrical devices. A cycle detector can process the power samples to detect a masking data cycle and determine masking parameters that characterize the masking data cycle. The power samples can be stored in a circular data buffer, and the masking parameters can be stored in a circular cycle parameter buffer. A cycle characterization can determine whether there is masking data within the raw data and, if there is, call a removal process to remove the masking data.