Robust Asset and User Aware Dispatch of the Power Distribution Grid during Extreme Temperatures

The goal of this NSF-funded research project is to efficiently reduce residential energy demand and usage in the extremely hot situation. We mainly focus on resource-aware non-intrusive smart meter data analysis and pattern recognition. A series of functions, i.e., load disaggregation, the A/C system inference, and human presence detection, will be built to determine the operation state and (if applicable) consumption level of the A/C unit as well as detect occupancy in the house. With the consideration of processing capability and memory size of the local module, resource-aware models and algorithms with low computational complexity and memory requirement will be preferred. Furthermore, we will implement the proposed models and algorithms for the real-world concept demonstration. 


Smart Buildings

The rapid introduction of emerging Internet of things solutions allows building operators to make their properties smarter, more observable, controllable, and sustainable. Massive temperature sensors can be deployed in the building to monitor real-time fine-grained thermal distribution. The goal is to reduce the total energy consumption in the building, while considering human thermal comfort.  As a critical and necessary input to plan and optimize energy usage in smart buildings, an energy consumption prediction tool will be proposed by exploring computational models. Based on such a tool, we can forecast the energy budget needed for temperature adjustment in a certain building area. In addition to temperature, we will also consider the building structures, layouts, and characteristics, the thermal properties of building materials, zone/area functions, and the positions of HVAC vents. Our efforts respond to Chattanooga’s participation in DOE’s Better Buildings Challenge.


Flexible Data Acquisition, Compression, and Reconstruction in Advanced Metering Infrastructure

We design a general, flexible, and efficient framework for data acquisition, data compression, and data reconstruction in advanced metering infrastructure. Compressed distributed sensing is utilized to acquire load data from smart meters and transmit them to the central control unit. Different sparse binary measurement matrices are exploited for different time instances when data acquisitions are performed. Each sparse binary measurement matrix corresponds to one data gathering scheme using compressed distributed sensing. Joint reconstruction of the two-dimensional load profile is executed at the central control unit. Both spatial and temporal correlations are explicitly employed to facilitate data reconstruction with high accuracy and fidelity. Meanwhile, the desirable data compression ratio can be achieved.



 sgrt  sgrt1

Reconstruction Error with Noise                        Reconstruction Error without Noise

Method One: the method using one sparse binary measurement matrix for all the time instances;
Method Two: the method using different sparse binary measurement matrices for different time instances;
Method Three: Method Two plus time domain data compression;
Method Four: Method Two plus time domain data compression with data combination.

Predictive Analytics

From the perspective of big data, it is well known that smart grid is an important big data source. Massive sensors of different kinds, such as phasor measurement unit, smart meter, and so on, have already been deployed in smart grid to acquire large-scale real-time data related to power generation, power consumption, system state, and equipment condition. These data are crucial to smart grid and can be exploited to enhance observability and controllability of smart grid. Nowadays, our research focuses on exploring big data in smart grid and improve its performance by taking advantage of the state of the art big data analytics. 

With the framework of big data analytics, predictive analytics has gained more and more attentions recently. Predictive analytics based on machine learning and/or artificial intelligence can be used to build the powerful predictive models to forecast what will happen with high probability in the future. Predictive analytics can provide sophisticated and actionable insights to utilities for decision making. Take load forecasting in smart grid as a motivated example. Load forecasting predicts the future power load demand, which is extremely important for generation planning, demand response, and unit commitment. 

One hour load forecasting performance for ISO data.












 Training                                                                                                                     Testing