The default view is to show a 7-day composite, or "folded" view of the average hourly loads in the period. This has the effect of showing what a typical week looks like. One gets a visual indication of the base load in the building, and the effects of occupancy during weekdays, and how well the building sets back at nights and on weekends.

You can zoom into a small piece of the chart by clicking and dragging horizontally across the chart. Click the ALL button to reset the zoom.

The dark line is the average 15-minute value for the day and time-of-day shown on the x-axis. If the Period is set to "Last 30 Days", then by definition each point is the average 4 or 5 loads found in the 30-day period for the given day of week and time of day since a given day and time-of-day can only happen 4 or 5 times per month. If the Period selector is set to "2017 FULL YEAR", then by definition each point on the line is the average of 52 or 53 load values since, for example, 9:15 on a Monday occurs 52 or 53 times in a given year.

The dark bands around the line are the +/- one standard deviation in the data. This is a measure of how tighly the loads surround the average value. The lightest blue band is the minimum and maximum loads for the given day and time-of-day in the period.

Switching to a 1-Day profile has the effect of showing a composite 24-hour view of all the days in the Period. Be careful with this view since by defintion it has the effect of mixing weekdays and weekends together, since each point on the line in a 30-day period represents the time-of-day average of 30 load values, some on weekdays and some on weekends. With a 1-Day Profile, it makes sense to use the Include Days of Week filter to look at a more appropriate day type for your purposes.

You can switch to a traditional Time Series view to get to the actual load values for any date and time in the period. In this case, the standard deviation and min/max bands will disappear because the data is no longer folded - you're looking at specific load data for a precise moment in time. (In some cases where AEI is measuring building levels at 1-minute intervals, the Detail view will still show the bands because we are averaging 1-minute values into 15-minute groups).

Finally, if you choose any FULL YEAR Period, the 12-Month 1-Day profile gives you an idea of how a composite day looks for each month in the year. This is most appropriate to see how the daily average profile changes during the different seasons. If the building is a school and is shutdown during the summer, this view can help you see that it was set back as expected.

QUALITATIVELY, what we are looking for with these Time-Of-Use Profiles is a visual indication that the building is on a schedule, is setting back, and it is reacting to weather in ways that are expected. We specifically look at the standard deviation bands during the overnight hours, for example, as an indication of whether the unoccupied hours are tightly controlled. If the building is unoccupied on weekends, do the load levels set back as expected compared to the nighttime set back? We can also see when the building starts and stops each day, and whether cooling or heating loads might be accompanied by higher volatility and increased peak demand risks.

Regression lines have certain predictive capabilities, but in the case of building energy as a function of weather the correlations are typically too weak to be of significant value. Nearly by definition, loads in a building are most highly correlated with work schedules (occupancy), and this independent variable is not usually available. As a proxy for the effect of occupancy, the selectors above allow you to filter out unoccupied hours and days, and this typically will improve the apparent correlation with weather by discounting the stronger effects of occupancy.

Does a building engage in meaningful demand response on hot summer days? Switching the x-axis to ISO NE Grid will show the building's loads versus ISO NE loads at coincident times. It makes sense that when ISO New England is north of 22,000 MW, we would expect a building that has DR protocols to react in a way that shows lower loads at the high end of the ISO NE loads. In the case of Boston City Hall (switch to July 2017 and x-axis ISO NE grid), we do not see this effect which suggests that the building is indifferent to grid conditions. In fact, it could be the case that the building flatlines at its max capacity on hot summer days (switch x-axis to Weather) because we see it's highest loads around 2,000 kW can occur in a wide range of temperatures above 70°F. It would seem that the building is at its limit during high cooling loads, and any meaningful demand response could go beyond the acceptable limits of building comfort.

Boston Copley Library is a great example of a "bi-modal" building when the chart is set to include All Hours. In a bi-modal building, the reaction to weather or other outside influences is negligible, and the building appears to have just two states - on or off. In the case of BPL, the nighttime hours are concentrated around the 500 kW mark, while the daytime hours tend to be all around 1,500 kW. This is also obvious when looking at the 7-Day Profile for the building. It's as if the building is controlled by a single on/off switch that coincides with the scheduled times of operation.

Again, we caution that the use of a regression line for building energy should be done with a high amount of skepticism and not without the calculations of standard error that should accompany any analysis. Given that, it may be useful to understand the implications of the regression line in a marginal way. In August 2017, the regression line equation for daytime/weekday loads at Boston City Hall versus Weather reads y = 14.07x + 787.61 with a correlation of r² = 67%. Translated into English, the takeaway is that for each degree increase in outside weather, the building load increases by 14 kW. The equation would predict a load of 2,054 kW at 90°F, but the problem is that a 95% confidence on that estimate comes with a ± 264 kW range which may be too large to use in any tactical way. Just the same, if the building knows that tomorrow's 90°F weather prediction would cause it to hit its 95th percentile load at the same time the grid might also peaking, there is significant value in that knowledge notwithstanding the margin of error that comes with it.

This is the underlying data used to drive AEI ISO Alerts. It is a simple time-series plot of the relevant ISO or RTO for the building, and it shows interval load readings retrieved via an API provided by the ISO or RTO, along with a forecast if available. The weather data comes from Weather Underground and uses the most reliable weather station in the same ZIP code as the building. Forecasts from the ISO vary from 2 to 10 days into the future, and the weather forecast is generally available for 10 days into the future.

You can zoom into a small piece of the chart by clicking and dragging horizontally across the chart. Click the ALL button to reset the zoom.

While it might seem helpful to plot the building load on the same chart, that visual is already available in the Time-Of-Use Profiles tab by selecting the ISO choice in the Channel Two selector under Options and Filters.

This Grid Coincidence plot may be the toughest visual to comprehend, but it's our most powerful tool for assessing the relative risk associated with the timing of a building's peak load and how closely it coincides with peaks on the grid. For a building that is subject to ICAP tags or ordinary monthly demand charges, knowing when the building is likely to peak is as important as knowing the load at that time.

The visualization uses a standard box and whisker for visualizing the median, quartiles, and extremes of a data set to show the distribution of the data and the central tendency. Moreover, we can look at two different distributions and quantify how much they overlap. In the case of a building's energy use compared to the grid, we plot the time-of-day for each daily peak in a given month - blue for the building and purple for the grid - and the degree to which the distributions overlap can be quantified. In our case, we define the degree of overlap as a risk that the daily peaks of the building coincide with the daily peaks of the grid.

First note that the y-axis is calibrated for the time of day on a 24-hour scale. The x-axis shows a pair of plots for each month in the selected Period - blue for the building, and purple for the relevant ISO. Each point in each plot represents a single day in the month, and the point is plotted on the y-axis at the time of day when the maximum load occured for that day. The diameter of each point is not very important, but suffice to say that larger diameters are suggesting higher loads. If you've followed this far, then it should make sense that in each plot there are 30 or 31 points - one for each day in a month, or fewer if we filter for just weekdays or weekends. Given the different times that a peak might occur, the "box" in each plot captures the middle 50% of those times, and we consider that to be the central tendency for the building or grid. The black mark inside the box is the median time-of-day for all of the 30 or 31 peaks in the month.

Finally, we make a Risk calculation which measures the degree of overlap between the plots for the building and the grid, and we scale the calculation to a range of 0 to 100. A Risk calculation of 100 indicates a high degree of overlap - a high risk that the building peaks at the same time as the grid. A lower risk means less overlap, and indicates that it's less likely that a building peak will occur near the same time as the grid.

QUALITATIVELY, and in the very broadest terms, Risk factors below 50 typically mean there is good separation between the building's peak time-of-day and that of the grid. Values above 50 can mean a separation of 2 hours or less. And with any Risk calculation, it's important to remember that on any given day the peaks may coincide regardless of what the Risk calculation suggests.

The Billing tab shows a pro forma table of billing values according to standard utility rate tariffs. The billing engine takes usage and peak values from the EPO or real-time interval data and applies them to the calculations specified by the rate schedule.

Since by definition we are using time-of-use interval data, this implies that the building is on a TOU rate schedule, such as Eversource B2 or B7, or National Grid G-3. Other non-TOU rate schedules are supported when a building can provide interval data on its own, as shown in our examples for the BME and BMU Schools which use AEI Soft Start Real-Time to obtain 1-minute interval data for these buildings that are on the National Grid G-2 schedule because they have a maximum demand below 200 kW. In practice, with EPO or real-time interval data through AEI Soft Start, any rate schedule can be used for a billing calculation because the minimum inputs of kWh usage are always available, as well as the peak demand if it's needed for demand charge calculations.

If a building has a third-party supplier contract, the terms of that contract are incorporated into the billing engine. Upon request, the Billing tab can be configured for multiple rate tariffs for customers who wish to evaluate different tariffs to make a case for a rate change with their utility.

Move the mouse cursor over any column in the billing table, and the values in the column will color-coded from low (green) to high (red).

Certain setback ratios are helpful for tracking a building's ability to reduce energy consumption during unoccupied hours. In a typical commercial space that is only open for business during daylight hours on weekdays, approximately 70% of all hours in an average week are unoccupied, so it makes sense that a strong reduction in energy consumption during these hours can have a big impact on energy usage and costs.

We calculate the setback and volatility ratios above to provide guidance on how well a building is managing the unoccupied hours. But we want to be clear that there is no specific ratio that is best because the underlying components have their own characteristics, and simply dividing them to arrive at a ratio doesn't necessarily mean that one ratio is better than another.

Night Setback

In this table, we show the median load for each month for both the Weekday Occupied hours (6am to 6pm), and the Nighttime hours (6pm to 6am). The ratio is the nighttime median divided by the weekday daytime median, expressed as a percent. A relatively low percentage would seem to be preferable, because that means that the building reduces its usage during the unoccupied nighttime hours. But as mentioned above, it is possible to have a low ratio if the weekday occupied median load is very high.

Move your mouse cursor over the WD Occ column, and you will see all weekday occupied loads color-coded with a scale from low (green) to high (red). For a building with high summer cooling loads, you should expect to see orange/red cells that correspond to higher weekday cooling loads during the summer months. Typically, if a building requires mechanical cooling overnight, you will see the same color-coded pattern when you position the mouse cursor over the Nights column. Finally, if you move the mouse cursor to the Ratio column, you should expect to see a consistent ratio over time if the overnight and daytime loads move up and down together throughout the year. If the ratio shifts significantly over time, it's then worth inspecting the overnight and daytime loads to see which one of them is having the larger effect on the ratio - is the base load increasing such that WD Occ is gradually moving higher? Was new scheduling introduced to drive the Nights loads lower, resulting in a lower (better) setback ratio? It may require extra research to unravel the root cause of the changes over time, but with this presentation of the main meter interval data, you have a consistent measurement over time that can alert you to changing patterns of energy usage.

Weekend Setback

Similar to the table for Night Setback, in the Weekend Setback we compare the median load for all weekend hours to the weekday occupied median load. For a building that is closed on weekends, we would expect a very high ratio since the building might expected to have a programmed weekend setback that is even more aggressive than its weekday nighttime setback. Conversely, for a building that is open 7 days per week, the ratio could be significantly higher. In the case of schools that are in shutdown for the entire summer, it would not be unexpected to see a high ratio during the summer months because the weekday occupied loads are also low since the building may be in a 'weekend' mode for the entire summer.

Nighttime Volatility

We like to look at the variability of loads in unoccupied hours because a high variability - or volatility - hints at the possibility that the building is not running on a schedule during the unoccupied hours. In this table, we measure the night time median load in the Nights column, and the Night IQR is the middle 50% range of the loads around that median. When the ratio is high, that means that there are large load swings or "noise" happening during the hours when we would expect loads to be controlled to minimal levels.

In this screen image below, we are looking at the Nighttime Volatility for Boston Copley Library.

In 2015, we notice a volatility ratio of 60%. Given a median load of 734 kW, this volatility means that 50% of the nighttime loads are between 734 kW ± 30% or ± 220 kW. So while it might be the plan to have a reduced nighttime load of approximately 750 kW, we note that the actual loads range from 530 kW to as high as 970 kW for 50% of the time. The year 2016 had no EPO data, and then the problem seems to get worse in 2017, although we note the median load is lower. Finally, in YTD 2018, the nighttime median loads have jumped significantly, but the IQR of 199 kW shows that the volatility is lower - perhaps everything is turned on, leaving less equipment to randomly volley on and off through the night. It turns out that the Library has certain nights during most months where the overnight loads are more similar to occupied hours, as if the building is open and occupied all night.