How Accurate is Sense vs. Utility Metering?

[Dcdyer]

@kevin1- You inspired me to analyze my home data. It’s been a while since I have used EXCEL charting…

I compared my daily electrical consumption with my local power company meter and the internal personal energy monitor called SENSE. I now have almost a year of data stored. My electrical provider is CenterPoint Energy of Texas.

Setup Steps:

1. I extracted the daily readings from the ‘SENSE V4.0 web version’ for each month in 2018.
2. I downloaded my power company’s (CenterPoint Energy) meter readings from WWW.SMARTMETERTEXAS.COM
3. I combined all of the .CSV files into a single Excel file for 2018. (and developed another file for 2017)
4. I setup my calculations in Excel:

• kWh (daily) difference = SENSE kWh – Power Company kWh
• I removed any bad data. (-1.0 < ‘kWh diff’ < 1.0 [between -1 and 1]) If SENSE did not download all of the results for that day, then I excluded that data point. Generally the problem was a loss of my SENSE WiFi connection.
• % error = kWh (daily) difference / kWh Power Company (daily) * 100
• And an ‘offset value’ from the average baseline defined later.

Plotting Charts in Excel (Windows 10).

First, I plotted the SENSE values against the Power Company values using a simple graph. This allowed me to determine which days might be bad data. Then I plotted a linear regression. If the slope is 1.00 and the R2 factor is 1.00, then the data points would be a perfect correlation.

Second, I plotted the ‘kWh difference’ values as a Histogram. I was looking for a perfect single bell curve. What I observed was two distinct bell curves indicating two separate trends. My first plot showed a nice bell curve.

I kept reducing the range and discovered two bell curves.

Third, I plotted the ‘kWh difference’ values as a scatter point chart and looked for any trends. Two distinct baselines were observed. (It is hard to see the two dashed lines that show the average baselines on this chart at -0.137 and +0.011)

Fourth, I calculated the average for each baseline, then I calculated the ‘offset’ from the baseline. ‘offset’ = kWh difference - baseline average, I plotted those values in a Histogram, Then I summed all the outliers (values that were not close to the bell curve).

Observations:
On April 23rd, 2018 there was a shift on the offset baseline of 0.148 kWh and has stayed consistently at the new baseline. I explored several possible reasons for the baseline shift.

1. New monitor software update from SENSE during this time period. (That had not happened.)
2. Change in power consumption. Warmer weather, more A/C usage. (2017 data did not show a baseline shift when the weather changed.)
3. Temperature change inside the house. My thought was maybe the SENSE monitor (located inside the house in a climate controlled room) was operating at a different condition. (We keep the same setpoint on our thermostat year-round).
4. Maybe my home power company meter had been changed out or their meter was at a different operating condition. (No changes were noted.)
5. On 4/23/2018 I added a TD-69 delay timer relay to my SENSE monitor. At the same time I reworked my CT clamp installation. I used the foam insulation from a ½-inch pipe insulation as a spacer to center my clamps and to position them so that they are 90-degrees perpendicular to the service cable in the breaker panel. When I initially installed the CT clamps (Sept. 2017) they were not at a true perpendicular angle or perfectly centered around the cable. I also taped the CT clamps in closed position, making certain they were completely closed. The new foam insulation ‘spacers’ were not allowing the clamps to remain closed.
6. My charts indicated that prior to re-positioning my CT clamps, that I could expect a higher number of outlier data points.
   Pre- 4/23/2018: 111 data points with 37 outliers
   Post-4/23/2018: 124 data points with 30 outliers

Recommendations:

1. Verify your CT clamps are completely closed and at a 90-degree angle to the power (or service) cable when doing your initial installation. Changing the clamp positions may give you a change in the readings.
2. My SENSE monitor now tracks the same values as the Power Company meter more frequently after making this very minor change. The SENSE unit currently reads an average of +0.011 kWh/day more consumption than my power company meter where previously it read -0.137 kWh/day or less consumption.
3. SENSE currently tracks the power company meter with an average 0.024 % daily error between the two readings.
4. It is also possible that the new TD-69 relay is helping to automatically reboot the SENSE unit (and reconnect to the WiFi sooner) after a power outage. Prior to 4/23, I was manually rebooting the SENSE monitor 1.5+ hours after the unit failed to communicate. SENSE text and message “off-line” alarms do not send until the device has been off-line for over an hour.