Formatting data using an if function

Hello everybody, how can I adjust the data, so that the purple curve will always be connected? The next purple section should start at the end of the section before.... I've no clue how to do that. I really need help, because of this thing I'm not able to finish my bachelor thesis. Thanks guys.

(Ignore the legend)

C=cumsum(diffE,'omitnan'); % needed Capacity 
Cbmax= 1000; % in MWH, max. capacity of the battery storage
Cbmin= 100; % in MWh, deep charge border 

imax= C>=Cbmax-Cbmin; % find where C is larger than the Capacity
imin= C<=0; % find where C is =0
C(imax) = Cbmax-Cbmin; % if C is larger than Cbmax: C=Cbmax
C(imin) = Cbmin-Cbmin; % if C is smaller than Cbmin: C=Cbmin 

Cres=cumsum(diffE,'omitnan');

for k=1:length(Cres)-1
    if Cres(k)>(Cbmax-Cbmin) && Cres(k)>Cres(k+1) 
           C(k+1)=C(k)-Cres(k)+Cres(k+1); 
    end
    if Cres(k)>(Cbmax-Cbmin) && C(k)<(Cbmax-Cbmin)
         C(k+1)=C(k)-Cres(k)+Cres(k+1); 
    end
end

Answer 1

[Stephen23]

"how can I adjust the data, so that the purple curve will always be connected?"

• loop over the length of the data vector.
• detect any discontinuities (e.g. difference/offset, or some other metric).
• shift the remaining data by the detected offset (use indexing).

Something like this:

for k = 2:numel(vec)
    d = diff(vec(k-1:k))
    if abs(d)>tol  % pick a tolerance, or some metric
        vec(k:end) = vec(k:end)-d;
    end
end

I just tried this on a fake data vector:

tol = 2;
vec = [0,1,2,3,4,10,11,12,13];

giving:

>> vec
vec =
   0   1   2   3   4   4   5   6   7

You could fine-tune the algorithm so that it includes an offset of the same sign as d, but with magnitude tol. Also it would pay to do some reading into detecting discontinuities:

https://www.mathworks.com/help/wavelet/examples/detecting-discontinuities-and-breakdown-points.html

https://en.wikipedia.org/wiki/Step_detection

[Benedikt Wessel]

That's exactly what I need! Thank you. But unfortunately the diff functions is always in an endless loop. How can it be that you don'T have this issue with your vector? Sorry I'm quite bad at matlab....

[Stephen23]

"But unfortunately the diff functions is always in an endless loop"

Check if you have any other functions named diff:

which diff -all

And if so, rename them. Also ensure that you do not have any variables name diff. Note that diff is not strictly required, you could do this:

d = vec(k)-vec(k-1)

[Benedikt Wessel]

Oh nice, now it's working. I although tried to use indexing to find the gaps and to adjust them, but my code is a mess I think. Sorry to bother you, but can you find my mistakes? I also added the needed data if it's helpful. Here's my code:

C=cumsum(diffE,'omitnan'); % needed Capacity, will be adjusted by Cbmax anc Cbmin
Cres=C;
Cbmax= 1000; % in MWH, max. capacity of the battery storage
Cbmin= 100; % in MWh, deep charge border 

imax= C>=Cbmax-Cbmin; % find where C is larger than the Capacity
imin= C<=0; % find where C is =0
C(imax) = Cbmax-Cbmin; % if C is larger than Cbmax: C=Cbmax
C(imin) = Cbmin-Cbmin; % if C is smaller than Cbmin: C=Cbmin

Cres=cumsum(diffE,'omitnan');
change=zeros(length(diffE),1);

for k=1:length(Cres)-1
    if Cres(k)>(Cbmax-Cbmin) && Cres(k)>Cres(k+1) 
           C(k+1)=C(k)-Cres(k)+Cres(k+1); 
    end
    if Cres(k)>(Cbmax-Cbmin) && C(k)<(Cbmax-Cbmin)
         C(k+1)=C(k)-Cres(k)+Cres(k+1); 
    end

      change(k)= Cres(k)>(Cbmax-Cbmin) && Cres(k+1)<(Cbmax-Cbmin);

end

cha=find(change==1); % find the positions of the gaps between the sections

for k = 1:length(C)-1
   for n=cha
       if C(k)==C(n)  % find the data with the "crack"
            C(k:n+1) = C(k:n+1)-C(n)-C(n-1); % subtract every value in the actual section with the difference between the sections
       end
   end
end

Answer 2

[Walter Roberson]

Use https://www.mathworks.com/matlabcentral/fileexchange/27994-inpaint-over-missing-data-in-1-d-2-d-3-d-nd-arrays

[Benedikt Wessel]

Thanks. But I don't need to interpolate the values. I need all values to be lower to fit the section before.