Time Merge and Analyze Data

For our TimeMerge,

select

both displayed (and utilized) Time variables ,

leave

“Use earliest start date”, and “Use latest end date”. The “Time Statistics” button can be helpful to select other Merge parameters.

Select

and review “Show Time Statistics”.

Time is sorted (all true), so

leave

Out-of-order times as is (Cut data). Time statistics around intervals are quite complex and obscure our intervals for merged time. Observing the original data capture is every 5 seconds for process data, 30 seconds for Emissions analyzer data.

Set

the Interval to 5 seconds (on other cases you could choose 30 seconds or a different frequency), Maximum time gap to 12 minutes, Zero interpolated value certainty after 5 minutes and interpolation method to “Linear Extend”. There are multiple “Constant’ intervals (duplicate times) so

set

Duplicate times to “Use Average”.

Select

OK.

Data is now on a common time basis so that any analysis or machine learning that assumes a time alignment between different sources can use the data.

Open

Transforms as shown in the following image or

Click

the Dataset icon and click [Transforms].

Edit

the variable name, e.g. !Unit2 Efficiency! and move your cursor to the remaining Transform definition window. You may delete, edit or restart any variable name with just an initial exclamation mark locating the right variables (the name needs to exist, as misspelling will be an error on an undefined variable).

Change

to divide “Unit

2

Generator Total Real Power (kW)” by “Unit

2

CTG

2

00 FUEL GAS FLOW (Lb/Hr) – FT2108S”. Then

select

“Add” again. We will investigate what influences turbine efficiency. When complete (confirm two transforms are added at the bottom) select “Show Content Assistant”.

A keypad and common transforms are displayed. You will see a broad range of functions available under an array of categories (Common, Math transforms, String transforms, Date & Time transforms, Generated: clip/cut/TimeMerge, Status: tests, checks and ways to set cell status, Column: a variety of column transforms, Move: moving window transforms, Signal: frequency domain transforms). Each transform is available by initially typing in a “$” plus the beginning of the transform name or locating an available transform in the Content Assistant window. Note also that If/and/or/not (e.g. nested if statements) can also be created. Search filters for variable names or transforms are available.

Select

‘Close” at the bottom of the window.

Select

the top menu bar, Edit, Preferences, Limits.

Change

Maximum Number of Columns in Correlation: to 150.

Select

All ,

Right mouse-click

, and

select

Analyze and Correlation.

A Correlation is the linear correlation coefficient (e.g. maximum 1 for variables with themselves, which is the diagonal, i.e. what variables go up and down together) and after selecting all a correlation of each variable with each other variable in the dataset is displayed. Larger squares indicate a larger correlation. Blue rather than red (positive) correlations indicate reflective (negative) correlations. Put your cursor over any interesting square and see the variables (also listed on the axis) and the represented correlation coefficient. To see what is correlated with Unit 2 Inlet NOx observe the box sizes across the third row – until the diagonal (NOx versus NOx) and follow that third column down. The top correlations appear to be: Unit 2 Inlet O2 (-0.559) although there is also a high correlation with Unit 1 efficiency (that will be discussed later, there are few overlapping rows between the two units operations). Because we have completed TimeMerge rows and time are now the same.

Select

that larger correlation box of Unit 2 Inlet O2 and Inlet NOx. Follow the second column down and multi-select (ctrl-mouse-click) about 4 or 5 of the larger correlation boxes including Unit2, Duct Burner Fuel Gas Flow. With your cursor on one of the active, selected correlation boxes. select right mouse-click, Analyze Delay Correlation.

A Delay Correlation checks the correlation between selected variables (and a primary variable) comparing shifted rows of data (here every 5 seconds step from our time merge). When your cursor is on a bar the values are displayed as pop-ups and when selected shown on the plot (same information). The marked vertical plot indicates that the maximum correlation is 6 (5 sec.) steps where Inlet O2 shifts and then Inlet NOx shifts. Duct burner fuel gas flow appears more strongly correlated 2 steps before. You may shift focus window with right and left double arrows. This starts to indicate the process/system lags represented in time-series data.

Select

your Correlation tab.

Scroll

to the bottom right of the correlation chart and

locate

the peak (largest) correlation variable for Unit 2 Efficiency (bottom row).

Right mouse-click

and

select

Plot and xy plot.

The high correlation suggested and the xy plot displays a very nice relationship between efficiency and generated power. Increased efficiency increases with total power, but more and more gradually. Unfortunately this is an output and not actionable.

Go back to (

select

the tab) correlation plot and

repeat

for a few more xy plots on one to two other high correlation variables. Multi-select (

select & Ctrl-Select)

Unit2 Efficiency versus “Unit 2 CTG200 Fuel Gas Flow” and Efficiency versus “Unit 2 Average T5 Temp-2nd Stage Turbine Temp”.

Right mouse click

and

select

Plot and xy plot. You will also find these two measurements are highly correlated with each other (just move up vertically until you are two boxes from the diagonal 1.0). Another problem is now we have three different xy plots in two frames.

To provide a more descriptive (understandable) name we can label any plot we want to keep.

Double-click

on the plot label (“XY Plot(1)”)

or

select

the ‘Rename’ icon .

Enter

a useful, appropriate name such as “Efficiency2 vs Fuel and DeltaT”. Also

select

(activate) the Pin icon [ ] or Save icon ([ ] next to the rename icon), which marks any tab as one to save for later access.

Update

the tab names and ‘

pin

’ the first xy plots (e.g. something like “Efficiency2 vs Total KWh”). Now

select

the Project Cabinet icon in the upper left just below the Project taskbar option. The project cabinet also supports renaming or removing project items with a right mouse click.

Our ‘pinned’ tabs are retained as part of our project set-up and can be recovered if closed, by opening the project definition [ ] and selecting the desired, pinned tab (plots, datasets and analysis). Pin (save the configuration) of any reusable tab. To update this in the file system, close the project definition window by

selecting

anywhere outside it and then

select

the taskbar option “Project” and “Save”.