Summary

The Carlisle Public Schools (CPS) shows an estimated drop of energy usage of 3.5% for 2015 versus 2014. The value is estimated because of an apparent outage in the Schneider EPO system that shows zero usage for the months of February through May of 2015. Our estimate uses the 7 months of data available in both years to calculate the drop from 2014 to 2015 and then interpolates the missing months. The net is that we calculate a CPS usage of 791MWh in 2015 versus 820MWh in 2014. As a reference, the CPS electricity footprint is roughly 6x the size of Gleason Library.

In our Sensor Review below, we looked at HW Valves and selected CO₂ sensors. In previous reporting, we calculated a $1,700 savings opportunity associated with better VFD speed control during unoccupied hours. Many CO₂ sensors appear to have calibration issues, and while we don't see any health concerns based on the data, but under-reporting on CO₂ could result in under-ventilation. In some cases (Robbins UV-19, UV-24, the calibration appears off on the high side which could result in excessive ventilation and energy costs.

We constructed Load Profiles based on the EPO main meter data for the months of October through December for the past four years, nominally equal to the first half of the school year. We see a consistent drop in energy use over the four years.

Given the upcoming changes in peak demand electricity pricing, we focus on peak demands and will continue to explore opportunities to reduce peak demand charges. The month of September is consistently a high month in terms of kWh, but also tends to have the highest peak demand day in any year mostly due to the start of school in the early days of the month and, in the case of 2015 unusually warm weather. We see excellent weekend and off-peak setback at CPS and calculate a base load of approximately 35kW.

Sensor Review

Carlisle Schools Hot Water Valves

During 2015, it was observed that the valves are opening during the unoccupied hours and costing the school approximately $1700 per year in unnecessary VFD speed. This is reported here as a reminder to ensure that the valves do not continuing to operate 100% open during the unoccupied winter months.

Carlisle Schools CO2 Sensors

+ Expand this section

CO2 Sensor Calibration – based on observations from August – October 2015.

The following CO2 sensors should have the calibration verified because the minimum values are below ambient CO2 levels of approximately 400 ppm. There has not been any attempt to calculate the energy cost resulting from sensors that need calibration.

Corey RTU-1

Minimum values below 200 indicate that the sensor requires calibration. The CO2 levels do not indicate excessive energy consumption and it appears the space is adequately ventilated, but there is the opportunity for an under-ventilated condition to occur.

Grant EUV-5

Minimum values below 200 indicate that the sensor requires calibration. During September, the CO2 levels exceeded 1,000 ppm once. This doesn’t indicate excessive energy consumption, but there is the opportunity for an under-ventilated condition to occur.

Unit Vent 10

Minimum values below 300 indicate that the sensor requires calibration. During September and October the CO2 levels exceeded 1,000 ppm a few times. This doesn’t indicate excessive energy consumption, but there is the opportunity for an under-ventilated condition to occur.

Unit Vent 4

Minimum values below 200 indicate that the sensor requires calibration. During October the CO2 levels exceeded 900 ppm a few times. This doesn’t indicate excessive energy consumption, but there is the opportunity for an under-ventilated condition to occur.

Robbins UV-19

Frequently the minimum values never drop below 500 ppm during the unoccupied hours and the CO2 levels are frequently reach the 1000 ppm set point so there is the possibility that the space is over-ventilated and consuming excessive energy.

Robbins UV-24

Frequently the minimum values never drop below 500 ppm during the unoccupied hours and the CO2 levels are frequently reach the 1000 ppm set point so there is the possibility that the space is over-ventilated and consuming excessive energy.

Spaulding RTU-7

Minimum values below 300 indicate that the sensor requires calibration. During October the CO2 levels approached 800 ppm once. This doesn’t indicate excessive energy consumption and it appears the space may be adequately ventilated, but there is the opportunity for an under-ventilated condition to occur.

Load Profiles

We see a consistent drop in average demand at CPS over the past four years. The charts below show the average daily profile for 113 days in the months of September through December, nominally equivalent to the entire first half of the school year for the past four years.

Most notable is the 25kW drop in the base load from 2012 to 2013. Thereafter, loads show a consistent average decline while maintaining a similar daily schedule.

2012 Average demand, 110.3kW

2013 Average demand, 95.7kW -13.2%

2014 Average demand, 94.9kW -0.8%

2015 Average demand, 90.0kW -5.2%

Peak Demands

Below we show the hourly average kW demand for the months of October through December 2015. We would have looked at a full year, but the EPO portal is reporting zero values for the 3 months March, April and May 2015. So we focused on the first half of the current school year.

Discussion

By clicking on "wd", we remove all of the weekday values and see that the School retreats nicely to a relatively low state on weekends. At the same time, there are some very high peaks during the occupied states that should be further inspected for cases of coincident demands on equipment that might be better scheduled to mitigate the peaks. The default chart shows that 5% of the hours (110 hours) consumed 12.3% of the total energy in the period, and a little exploration shows the highest concentration of the peak demands to be concentrated between 9am and 3pm, as expected in a school.

Billing Simulation

In the table below, we have taken historic main meter EPO data for CPS and made an estimated calculation of the electric utility bill. The calculations are based on the Eversource B7 schedule which has a peak demand charge component as well as energy charges based on kWh. Looking at the past year, we can see that the demand charge component is approximately 80% of the average monthly bill.

We used the rates in effect on the December 2015 Eversource bill for the Carlisle Public Schools.

(74.137)

Discussion

Click on any of the peak demand days shown above to show the Top 10 peak demand days in each period. With 5' interval data from EPO, the tendency is that the peaks are all clustered in a few days in the month since one peak is usually a good predictor of a peak in the next interval.

In September 2015, school opened on September 8th which was a 90°F day, and as a result this single day had 23 of the top 25 peak demands in the entire year. The top peak demand for the year was at 12:23 on that day with a demand of 349kW which is nearly 4x the average load for the entire year. With anticipated changes coming in peak demand pricing, controlling peaks will be an essential component to good stewardship of the electricity footprint going forward. AEI will continue to focus on this topic.

Predictive Model

Coming Soon. We have enough main meter data where it may be possible to study the conditions (temperature, humidity and current load) that lead to historic peak demands. We will perform the regressions on available independent data to see if there is a good fit. A model that has a reasonable expectation of predicting a peak demand event 4 to 24 hours in advance could be a useful tool to help establish simple demand response protocols that are easy to implement.

Next Steps and Recommendations

We will continue to explore for discrete O&M opportunities at the HVAC level. Additionally, if we could gain access to the Eversource billing rates for CPS, we would be able to understand the impact of upcoming changes to the demand charge and how it specifically affects CPS, as well as Gleason Library and Town Hall.

In the near term, we will be updating this page with more extensive weather regression work and continue to explore ways to develop a high-confidence predictive model that could be used to implement a simple set of DR rules to counteract peak demands.

Twitter