Are you sure the OPF converged?

-- 
Ray Zimmerman
Senior Research Associate
B30 Warren Hall, Cornell University, Ithaca, NY 14853
phone: (607) 255-9645





On Jun 5, 2013, at 12:35 PM, Alexandra Kapetanaki 
<[email protected]> wrote:

> Dear Ray,
> 
> I run DC opf for the 24 bus IEEE network and I have negative virtual gen for 
> the representation of the load  and for an intact system with 2850 MW load 
> there is 2850 MW dispatchable load.
> However, in the case I add extra virtual gen to all the 24 buses with zero 
> output (also zero Pmin and Pmax) the dispatchable load is 1036MW instead of 
> 2850MW.
> Basically, the network is the same eg generation capacity, load, transmission 
> lines's capacity in both cases.  
> The only difference is the extra generators with zero output in the second 
> case.
> Can you please insight me why the load is not satisfied in the second case?
> 
> Thank you in advance,
> Alexandra
> 
> Alexandra Kapetanaki
> PhD Student
> Electrical Energy & Power Systems Group, School of Electrical & Electronic 
> Engineering
> Ferranti Building (B18), The University of Manchester, M13 9PL, United Kingdom
> Tel: +44 (0) 161 306 2263; Mobile: +44 (0) 7857 598179
> From: [email protected] 
> [[email protected]] on behalf of Alexandra Kapetanaki 
> [[email protected]]
> Sent: 04 June 2013 15:39
> To: MATPOWER discussion forum
> Subject: RE: transmission losses-DC opf
> 
> How can I exclude the Q losses in AC opf?
> Because the formula is applicable in a DC model without the effect of Q.
> 
> Thank you in advance!    
> 
> Alexandra Kapetanaki
> PhD Student
> Electrical Energy & Power Systems Group, School of Electrical & Electronic 
> Engineering
> Ferranti Building (B18), The University of Manchester, M13 9PL, United Kingdom
> Tel: +44 (0) 161 306 2263; Mobile: +44 (0) 7857 598179
> From: [email protected] 
> [[email protected]] on behalf of Ray Zimmerman 
> [[email protected]]
> Sent: 04 June 2013 15:05
> To: MATPOWER discussion forum
> Subject: Re: transmission losses-DC opf
> 
> I'm not sure, but you could try your formula to estimate the losses for the 
> AC model as well. That would give you an idea of how good your estimate is.
> 
> -- 
> Ray Zimmerman
> Senior Research Associate
> B30 Warren Hall, Cornell University, Ithaca, NY 14853
> phone: (607) 255-9645
> 
> 
> 
> 
> 
> On Jun 4, 2013, at 7:10 AM, Alexandra Kapetanaki 
> <[email protected]> wrote:
> 
>> Dear Ray,
>> 
>> Thank you for your support!
>> 
>> I run DC OPF in the 24 bus IEEE network and my aim is to include a 
>> transmission losses model in the system. 
>> I followed the procedure you insisted to me and for the losses per line I 
>> used the already proved equation Ploss = rij(Pk)^2 which is applicable in a 
>> p.u. system. 
>> So I divided the Pk of each line by baseMVA(=100) whereas the rij is already 
>> in p.u..
>> However, I compared the losses of the branches for an intact network(24bus) 
>> by using  AC opf  to the corresponding losses by using DC opf and they seem 
>> to be different as you can see below:
>> Losses for each branch_DC(real values)=
>> [4,32
>> 6,306
>> 2,4
>> 2,21
>> 2,114
>> 1,173
>> 1,318
>> 5,486
>> 2,31
>> 2,4
>> 5,67
>> 12,4
>> 6,18
>> 3,27
>> 6,791
>> 3,83
>> 14,3655
>> 4,268
>> 5,459
>> 3,761
>> 5,155
>> 5,11
>> 3,94
>> 4,13]
>> 
>> Losses for each branch_AC(real values)=
>> [0,0022
>> 0,13
>> 0,864
>> 0,644
>> 1,25
>> 0,156
>> 0,933
>> 0,245 
>> 0,059 
>> 0,94 
>> 1,63
>> 0,756 
>> 0,423 
>> 0,285 
>> 0,35 
>> 0,50 
>> 0,611 
>> 0,537 
>> 1,391
>> 0,309 
>> 5,80
>> 4,69
>> 6,129
>> 0,030 
>> 2,187
>> 2,187
>> 2,692
>> 4,1436
>> 0,230 
>> 0,833 
>> 2,697
>> 0,191 
>> 0,191 
>> 0,100
>> 0,100
>> 0,313
>> 0,31 
>> 1,829]
>> 
>> I would like to let you know that Plosses_AC_opf were obtained by the 
>> following command: 
>> results_2(N).Plossesbranch_AC(br)=results.branch(br,14)+results.branch(br,16);
>>  
>> 
>> Why do you think the losses are considerable different between the AC,DC 
>> models?
>> In DC the Q is not included so this is one reason of the deviation but is 
>> that enough? 
>> 
>> Thank you in advance,
>> 
>> Alexandra Kapetanaki
>> PhD Student
>> Electrical Energy & Power Systems Group, School of Electrical & Electronic 
>> Engineering
>> Ferranti Building (B18), The University of Manchester, M13 9PL, United 
>> Kingdom
>> Tel: +44 (0) 161 306 2263; Mobile: +44 (0) 7857 598179
>> From: [email protected] 
>> [[email protected]] on behalf of Ray Zimmerman 
>> [[email protected]]
>> Sent: 28 May 2013 17:58
>> To: MATPOWER discussion forum
>> Subject: Re: transmission losses-piece wise linear approach
>> 
>> - Solve a DC OPF with no losses.
>> - Compute the loss for each branch, Ploss = rij(Pk)^2.
>> - Add half of the loss to the load at the buses connected by the branch.
>> - Re-solve the DC OPF.
>> - Recompute the loss for each branch based on the new flow.
>> - Adjust the loads at each end of the branch to reflect the change in losses.
>> - Repeat, until the change from one iteration to the next is smaller than 
>> some threshold.
>> 
>> -- 
>> Ray Zimmerman
>> Senior Research Associate
>> 419A Warren Hall, Cornell University, Ithaca, NY 14853
>> phone: (607) 255-9645
>> 
>> 
>> 
>> 
>> On May 28, 2013, at 12:25 PM, Alexandra Kapetanaki 
>> <[email protected]> wrote:
>> 
>>> Dear Ray,
>>> 
>>> Thank you for your help!
>>> Can you please explain me more what do you mean with the  "adjusting the 
>>> set of dummy loads used to represent losses based on flows in the  previous 
>>> iteration".
>>> For example, I want to represent the losses as dummy loads according to the 
>>> following figure:
>>> 
>>> 
>>> <losses.png>
>>> 
>>> 
>>> 
>>> 
>>> and  express the losses by using the quadratic equation : Ploss=rij(Pk)^2 .
>>> How can I practically implement that within "few iterations", as you 
>>> recommend?
>>> 
>>> Thank you once again for your support,
>>> 
>>> 
>>> Alexandra Kapetanaki
>>> PhD Student
>>> Electrical Energy & Power Systems Group, School of Electrical & Electronic 
>>> Engineering
>>> Ferranti Building (B18), The University of Manchester, M13 9PL, United 
>>> Kingdom
>>> Tel: +44 (0) 161 306 2263; Mobile: +44 (0) 7857 598179
>>> From: [email protected] 
>>> [[email protected]] on behalf of Ray Zimmerman 
>>> [[email protected]]
>>> Sent: 28 May 2013 16:55
>>> To: MATPOWER discussion forum
>>> Subject: Re: transmission losses-piece wise linear approach
>>> 
>>> Dear Alexandra,
>>> 
>>> Since the network equations in MATPOWER's DC OPF assume no losses, it seems 
>>> you would have to introduce the losses as dummy (dispatchable) loads at the 
>>> downstream end of each branch. You would have to add an additional set of 
>>> constraints for each of these dummy loads constraining the consumption lie 
>>> above the piecewise linear constraints you propose. These constraints could 
>>> be added using the mechanism for user-defined constraints described in 
>>> section 5.3.2 and chapter 6 of the User's Manual.
>>> 
>>> Another approach to using MATPOWER to solve a "DC OPF with losses" is to 
>>> simply run the DC OPF iteratively, each time adjusting the set of dummy 
>>> loads used to represent losses based on flows in the previous iteration. 
>>> I'm guessing it wouldn't require more than a few iterations to converge to 
>>> a pretty good solution. This approach is a bit brute-force, but may be 
>>> simpler to implement and allows you to use whatever function you like (e.g. 
>>> a quadratic) to compute the losses. Just another idea.
>>> 
>>> -- 
>>> Ray Zimmerman
>>> Senior Research Associate
>>> 419A Warren Hall, Cornell University, Ithaca, NY 14853
>>> phone: (607) 255-9645
>>> 
>>> 
>>> 
>>> 
>>> On May 27, 2013, at 3:16 PM, Alexandra Kapetanaki 
>>> <[email protected]> wrote:
>>> 
>>>> Dear Dr Ray,
>>>> 
>>>> As the AC power flow requires high computation time for the losses to be 
>>>> calculated, a DC opf can be used in conjuction with a linear model for 
>>>> transmission losses.
>>>> My aim is to adjust the piece wise linear approach of Matpower in order to 
>>>> accommodate the losses of a transmission line.
>>>> More particularly, the losses can be expressed through the following 
>>>> equation: Ploss=rij(Pk)^2 [where rij the resistance of the line and Pk the 
>>>> power flow in the transmission line].
>>>> However, the above equation is a quadratic function but can be expressed 
>>>> with a piecewise linear model. The figure below shows a linear model 
>>>> consists of N line pieces 
>>>>  <piecewiselinear.png>
>>>> 
>>>> 
>>>> 
>>>> the equation of the nth line piece is:
>>>> 
>>>> <bbbbbbbbbbbbbbb.png>
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> -How can I modify the piece wise linear approach of Matpower with the view 
>>>> to include the losses model?
>>>> 
>>>> 
>>>> Thank you in advance,
>>>> 
>>>> Alexandra Kapetanaki
>>>> PhD Student
>>>> Electrical Energy & Power Systems Group, School of Electrical & Electronic 
>>>> Engineering
>>>> Ferranti Building (B18), The University of Manchester, M13 9PL, United 
>>>> Kingdom
>>>> Tel: +44 (0) 161 306 2263; Mobile: +44 (0) 7857 598179
>>> 
>>> 
>> 
>> 
> 
> 

Reply via email to