The number of windows detected increased from 38 for to 64, which means the new camera can detect more windows.

Comparison of performance

In general, the new camera can catch more windows with lower cost and higher accuracy. Using the new camera will increase the amount of windows for analysis.

Based on the methodology presented in the previous part, we were able to estimate the number of windows that our camera could capture, posing at Manhattan’s downtown. The number of windows in that area is 72,751,118. Out of those windows, we could observe 160 windows with acceptable illuminated pixel size and signal to noise ratio. This gave us the penetration rate of 0.0002%. This is a conservative number for certain reasons. The issue of obstruction was excluded and the camera’s field of view was assumed to expand to infinity for the simplicity of the problem, so the actual number of windows in the camera would, in fact, be smaller. Also, the area of our study, Manhattan is unique. We would not expect to see as many tall buildings as  in Manhattan in any other part of the non-first countries. Variations even exist among different neighborhoods within New York city. One example is the big gap between Brooklyn and Manhattan. With the camera pointed at the Verrazano-Narrows Bridge, covering only Brooklyn, there are about 38,011,368 windows in the camera view based on our estimation, close to 50% of those in the Manhattan’s case. In the less urban environments such as Mumbai, Deli, the obstruction issue will be less severe and there are abundant incandescent lights, so the penetration rate will be more significant. There are currently 12 PMU in New York, whose costs range from $40,000 to $180,000. With our methodology, we will need 6 cameras to have a 360-degree view of the neighbor, which costs $30,000 and is still less expensive than the minimum expense for one PMU.  

With the new camera, the instrument cost decreases from $5000 to $1500, which results in 70% cost reduction. The number of windows detected in the same area increased from 38 windows to 64 windows, which means 68% more windows. In a word, the cost for capturing the same amount of windows is reduced by about 82% with the new camera. Besides, the new camera have less shuttle time so that the intensity of light obtained by the new camera is more accurate.

The two most common ways to gather information about the grid dynamics, smart meters to monitor consumption patterns (usually at the building level) and Phasor Monitoring Units (usually at the city or city sub-network level) to monitor grid health are expensive and tend not to have to much granularity (temporal and spatial for smart meters and spatial for PMUs) in the information they provide. Hypertemporal imaging (HTI) from a vantage point in the city can be a cheaper way to have more granular data about the grid dynamics. Although the information is not exactly the same as the one generated by smart meters, it is possible that with some amount of ground truth data, the phase changes identified by the HTI technique can give valuable insights of the energy consumption in different portions of the city. On monitoring the grid health, the HTI technique can also be an alternative to the expensive and centralized PMU monitoring, by providing the same nature of information. This type of solution would be especially valuable for cities in developing nations that may not have the resources to implement the more expensive equipment to monitor grid dynamics. Our cost effectiveness analyses showed that the implementation of HTI technology would be 3.67 times cheaper than the average PMU implementation costs. The penetration rate will be improved in the less developed countries, where the obstructed view problem is minimized. The tests performed with the different upgraded instrumentation showed that since it allows the capturing of more windows at a given scene, it can make the costs per effectively monitored window even lower than the previous camera. This makes this technology even more promising in terms of its potential for monitoring grid dynamics and its feasibility for cities with constrained budgets.

More tests still have to be implemented with the new camera in order to fine-tune its parameters and eventually improve its performance even further. Besides that, ground truth data to correlate the phase changes with specific types of load changes in the systems is the key if we want to apply this technique to gain valuable insights about energy consumption patterns in the cities. Once we acquire that type of data, it is theoretically possible to build a model that would map phase changes to the turning on and off of certain appliances and detect any collective change in energy consumption happening at the neighbor level, thus attaining valuable information about energy consumption dis-aggregation. To achieve that goal, it is indispensable to establish a partnership with electricity providers to collect the required data and training such model on such real ground truth data.

Example of image processing and phase change detection