Histograms and Ranked Correlations

Select

the TrendPlot,

scroll

down to the Unit 2 Temperatures and Flows and

perform

a Multi-Select on flows from “Unit 2 Steam Flow” to “Unit 2 CTG200 Fuel Gas Flow”. On any of the selected trends,

right mouse-click

and

select

“Analyze” and “Histogram”.

Histograms display the range of data, by populations (height of vertical bar) distributed in bins. Select and unselect “Accumulation” that totalizes the bin volumes across the range (e.g. what value is 95% of the values).

Select All

and

Right Mouse-Click

on one of the plots and

Select

“Restrict Values”. Note also that there is also an option to “Set Bin Count” if you need more or less resolution.

Occasionally when comparing common values it is useful to display histograms on a common range.

Set

range to include the maximum data (Select all four tags and “Use Min/Max from Selection”). Alternatively

set

(manually change) the range (Minimum and Maximum Value) from 0 (eliminate below 0 data) and leave 118,130 (which is preset to that overall Max) and

select

OK.

You may select and range different histograms to different ranges. Select desired tags.

Undo this range set, with another

right mouse-click

on the histogram and

selecting

“Show All Values”.

Select

the histogram tab and the filter Icon in the display (on the right histogram header). Add the filters for High and Medium Load. Select OK. You may unselect (i.e. clear) plots with Shift mouse-click.

This may seem like all the data.

Select

the filter icon again and

select

“Show Unfiltered Data”.

Select

OK.

You will find unfiltered is most of the data but unfiltered in histogram is 'all' data. So the stack is all versus high load and medium load data. Move your cursor over different sections of the vertical stacked bars. Each pattern is shown (not hidden by overlay plots), so check the quantity of patterns represented (the percent is the percent of the section data) and you will see small differences, while most patterns available are either in the selected high or medium times.

Select

the original “Correlation” plot. We will study Unit 2 Inlet NOx.

Observe

what appears more or less correlated with Inlet NOx (remember inlet NOx is before the SCR and Exhaust NOx is after SCR which is used to reduce emissions). The larger diagonal is 100% (Inlet NOx versus NOx/itself) and stronger correlations are larger boxes such as Inlet O2 (the first blue box – a negative/reflective correlation of -0.559 (-55.9%).

Scroll

down,

observe

the second column to look for other correlated variables. From the 100% NOx versus O2,

right mouse-click

and

select

“Analyze” and “Ranked Correlation”.

Unit 1 Efficiency shows a high correlation with Unit 2 NOx (but we know to ignore this, it also indicates there are only 830 common rows used, compared to other results with closer to 50,000 patterns to evaluate). Correlation can be misleading and human consideration is a better way to confirm or discover real information in your data. After Inlet O2, information on the duct burner and pilot manifold appear relevant. To look at a delay search (increasing correlation search area

select

Delay “Search” and

enter

-15 to 0 at an interval of 1. Then

select

‘Compute’.

Correlation increases slightly with the broader search range. Ignore Unit 1 data – in particular those variables with very few rows (i.e. 100% correlated, but only using 2 rows). Only small changes occur here and combustion is a rapid process operation.

Compute again after selecting the “High Load” and “Mid-Load” filters. Results differ measurably after we remove transient and outliers. This is probably a better evaluation.

Double-click

both filters.

Double-click

the selector <All of> to change to “ANY”. Select “Calculate” again.