OPF doesn't only apply for loss minimization. It has other objectives too, e.g. voltage deviation, reliability, cost etc. For example in your existing system, how you can minimize total load consumption cost etc.. For this you may need optimal placement of generators, capacitors etc. These types of problem also comes under OPF.

Hi Bishal,

While Saran is right in what he says about OPF's, a prerequisite of any optimisation problem is the slackness of variables. Needless to say this is true about OPF's as well.

So if what you say is correct, that is: all generators are anyway loaded to their full capacity, then there will be no scope of any optimisation.

Your apprehension (that the OPF's serve no purpose !) is therefore very true for the case that you have described. It only shows your thorough understanding of the basic concept of optimisation !!

-Sanjay

Hi Bishal,

I am not sure if you are simply considering a single snapshot in your optimization (I mean evaluating your system only for the peak demand scenario or what is often known as "the worst case scenario"). If that is the case and given the conditions you mentioned, perhaps, an OPF won't help. But I would like to make the following comments:

1. Demand is naturally variable, ranging from shallow to peak consumption. (I presume it is no different in Nepal). If you take this into account, perhaps, you will gain something from the optimization. Note that this works if the total generation capacity in the system is at least greater than the total shallow demand in the system. If not, OPF may still be of no use.

2. You may also want to include network switching in your optimization (i.e. allow a line to be ON/OFF, if deemed economical). Sometimes, disconnecting a line may improve the performance (efficiency) of the system as a whole. In your system, such improvement could come as a result of reduced losses.

-Desta-

I understand your problem. All the OPF and all other advanced techniques are for countries with surplus power. A few  years back our state is in similar situation like yours. Now the officals say we have surplus power and so many solar and other projects are coming. One day your countryalso may be having surplus power. Please remember, new theories and technologies are for the entire mankind not for just one country.

Dear Sanjay sir and Desta Sir,

U are right in every sense. After your comments, i followed the load curve in detail and found a loophole. In fact the power deficiency condition is encountered during the Peak demand period ranging for about 6-8 hours of day and for remaining period of day the average demand is pretty much around 65% of max peak and below the total generation capacity. So, i realized i might still have some ground to play on.

Please note that  my test grid is ongoing through multiple upgrading project so i am just trying to slip in some low cost or no-cost solution for its performance improvement (in my case, Tx line minimization via optimal power flow). I am just not considering the FACTS implementation and all for now. Am i right to do so?

Dear Krishnarayalu sir,

Thanks. I am now considering the periods during which the generation exceed the demands rather than on the peak demand itself.

Will keep up all updated.

Regards and thanks again.

Bishal

Hi Bishal,

The fact that you are trying to find a quick fix to the system is commendable. Apart from the OPF and network switching, one relatively low-cost method you may want to explore is optimally allocating capacitor banks. Doing so can fairly reduce losses (even if the original system does not seem bad based on your description - 8% losses). In developing countries like mine, losses are rather high although I admit that some of this is non-technical.

Another fix worth exploring is the flexibility of demand. Reducing (shifting) the peak demand dramatically reduces losses.

-Desta-

Yes, Bishal.

Your approach should be fine provided the capital investment required for the FACTS devices is justified in the first place ! Recall our earlier discussion on the subject ??

-Sanjay 

I agree with Sanjoy. You may leave the decision of whether or not installing FACTS devices is justifiable (economical) to your optimization model. If the net savings exceed the net cost of installing such devices, why not.

Dear Sanjay sir,

Yes i do remember your previous suggestion regarding the justification for extensive use of FACTS devices throughout the power grid (while implementing my optimization solution). I will definitely keep that in mind.

Meanwhile, i am about to extract new solutions for the Tx line optimization problem in my test Grid. I am working on the refinements of the system modeling to make the solution as accurate as possible. I am expecting help from you all for deriving the conclusion and strengthening the rationale of the study. 

I will keep you updated via next feed.

Thanks and regards,

Bishal

Sure, Bishal.

Any time ! You know how to locate me on RG...

-Sanjay

Hi Bishal,

Below are some pertinent references to the problem you posed.  They support what is described in the following paragraphs.

1.  Looking at your single line diagram, it appears that you have 8 generation units.  None of these is very large.  Operating them at the most efficient point is important.  However, it is very likely that these units have efficiency curves that may not be accurate.  (I am assuming they are running on some sort of fossil fuel – natural gas, coal or oil.)  Additionally, as previously indicated, the load over the course of the day is changing.  Hence, understanding the nature (shape) of these efficiency curves is important.  Typically, they do not match well with what was determined when the unit was shipped.  They also vary significantly with season.  Making sure you are operating as effectively as possible is important.  If you feed in the “correct” curves into an optimal control algorithm (OPF) it will give you a correct answer, for a given load point.  However, if your efficiency curves are wrong, you won’t be helping yourself.  You have at least 8 degrees of freedom when the units are not at maximum.  Additionally, the manner in which you approach maximums of each unit can also affect the overall efficiency of the system.

2. You should evaluate optimal power flow at various set points.  This will help you determine if you should let “all” your units vary with load, or else keep several of them at a base load and just vary a few.

3. I suspect between assuring the correct operational characteristics of your units, and also determining how to best let them vary during the day you may have more reserve capacity than you think.

4. Once you have more certainty about how to operate your units, then you can investigate placement of capacitors or else adjusting the taps on transformers to improve losses.

5. Ultimately, you will have to add generation.  However, your OPF results and careful analysis will help in making sure that proper sizing and interconnection of a new unit will ideally augment your present situation.  By evaluating inclusion of a new unit (before it is put on line), it will also help you to understand what you don’t need to do in the short term.

Good Luck

Gary Viviani

A.H. Noyola, W.M. Grady, G.L. Viviani, "An Optimized Procedure for Determining Incremental Heat Rate Characteristics," IEEE Transactions on Power Systems, Vol 5., No. 2, May 1990.

G. L. Viviani, "Practical Optimization", IEEE Transactions on PAS, July 1985, PAS-104, No. 7, pp. 1691-1695.

G. L. Viviani, C. E. Lin, and M. Webb, ”A Real Time Approach for Revising Generation Unit Performance Characteristics", IEEE Transactions on PAS, Feb. 1985, pp. 280-285.

G. L. Viviani, G. T. Heydt, "Stochastic Optimal Energy Dispatch", IEEE Transactions on Power Apparatus and Systems, pp. 3221-3228, V. PAS-100, July 1981.

Dear all,

please go through this update regarding the study. Your feedback is going to be crucial.

Basic Algorithm of Study (Transmission loss minimization by optimal power flow)

1.       Power System Profile study :

Test Network: Integrated Nepal Power System, NEPAL

(Note: I am not considering IEEE standard bus system for this study because I have already done that for the educational purpose and now I want to develop something useful relevant with the real world problem itself)

Total Generation capacity- around 825 MW (Mostly all Hydro Power Plants)

Total no of generators- 13 Hydro power plants (largest one of 144 MW) and 2 Diesel Plants (total of 55 MW)

Peak load Demand during Test condition (June- August)- around 1300 MW(during 18:00 hr to 22:00 hr)

Average load demand for same period- around 800 MW (Below 750 MW for off peak hours)

(Note: during the peak hours, the demand is higher than available generation so, this study only considers the period when the demand is at least 50 MW lower than the available generation as some slack power is essentially required to implement the optimization concept.)

2.       Formation of YBUS and modeling of power TX line network for load flow.

3.       Load flow analysis of existing system by Newton Rapson Method. The transmission line loss itself was calculated to be around 55 MW (around 7 %)

4.       Formulation of Genetic Algorithm code for optimization.

-Optimization function- Transmission line Loss

- Changing variables- Active power generation (Pg) of 15 generators.

- Other GA parameters (population size, termination criteria, etc.) set accordingly

5.       Results:

Genetic Algorithm was implemented in tandem with same Newton Rapson Method in the same power Network under same condition.

-          The Loss was re-calculated (around 5%) 40 MW (against 55 MW under previous Condition)

-          The respective Pg values for each of 15 generators were obtained as solution.

6.       Advantages/ rationales as I derived:

-          A difference of 15 MW of power in a network of capacity of around 800 MW

-          Available system reserve (Spinning Reserve in this case) of the system increases.

-          This concept could be quite cheap to implement as it is not considering any additional FACTS or compensating devices but just the control or re-shuffling of Active power (Pg) generation of all generators to meet the load demand.

-          Also as an added advantage, burden in power grid is reduced even if by lower margin.

7.       Limitations:

-          For this above described power Grid, Peak demand Exceeds the Total Generating capacity, so, this optimization concept can only be valid during the off peak hour where the slack power available is at least around 50-75 MW.

-          Electrically, low loss in the system is always an added advantage. However, from non-technical aspect or non-electrical aspect, can this study be justified?

-          This load flow and optimization gives static solution only. How to make it applicable in dynamic real world case? Probably running the algorithm in a regular interval in loop and setting the generators output as per the obtained solution?

Please do share your opinions and expertise.

Regards,

Bishal

Dear Bishal,

 I just want to share my opinion in some points. However I do not familiar in a power system  of Nepal so most of my discussions are based on the previous questions and answers of all experts and top up with my idea.

1) Almost countries plan generation reserve capacity of conventional power plants about 5-15%. In the country of Nepal it seem less than this value in a period of peak demand. Thus, security operation (SCOPF) must be considered during this period. As a result, transmission upgrading and a proper generation planning must be considered as a basis.

2) From the characteristics of generations that rich of hydro, a unit commitment may be relax. Additionally, the OPF of minimizing fuel cost is not a crucial point. Reducing transmission loss (5%, very high) is a key issue. The successive is rely on a planning stage of upgrading transmission system and build new power plants at the proper location.

3) To reduce loss in an operation stage, a sub-problem of optimal reactive power flow may be considered if reactive power inadequate in some area or in cases of improper control of AVR coordination.

Dear Bishal,

Based on the fact that you are 95% hydro based, my previous answer needs some revision.

1.  The primary reason for OPF solutions is to reduce “fuel” costs.  As such, hydro, nuclear, wind, solar etc. are not so much concerned with optimization with regard to fuel costs.

2. For a typical hydro plant, it is best operated (least wear and tear) by running in a fixed, continuous fashion.  Apparently, you must be cycling your hydro plants, which is more expensive (due to wear and tear).

3. Hence, you don’t really have a fuel cost problem.  You are left with a demand problem as well as a transmission loss problem.  You don’t really have a conventional OPF cost problem.

4. Again, given your location, I am going to assume that you don’t have a constraint on availability of hydro resources.  If you do, then you have a so called integral-constraint-problem, which is a typical hydro generation scheduling problem that is often addressed by “dynamic programming” algorithms.  If you do have significant constraints on hydro reserves, then you would not appear to have any good options.

5. My suggestion is as follows:

• a.  You need a less expensive – faster way – to meet demand.  New conventional hydro plants already appear to be in progress (which suggests that you have still more hydro available).  That is good.  But in the interim, you might consider operating your hydro plants at a constant output that is higher than your minimum demand load and installing energy storage as follows:
•  (I) Energy storage derived from re-used electric vehicle batteries is a growing trend.   See http://www.computerworld.com/article/3005757/sustainable-it/daimler-to-recycle-electric-car-batteries-for-massive-energy-storage-systems.html     for example.  This can also be accomplished with lead acid batteries.

b. I did some investigation regarding this topic and found the following thesis from MIT.  Maharjan, Sushila, “ Electric Vehicle Technology in Kathmandu, Nepal:  A Closer Look at its Development, Master in City Planning, MIT, September 2002.  Apparently, Kathmandu has an air pollution problem and electric vehicles were vigorously investigated and implemented in the 1990’s and early 2000’s.    Also, it looks like there was a fossil fuel lobby that was more effective.  It seems electric vehicle technology in Kathmandu is in place – hence there may exist a supply of recyclable batteries.  (If not, you could probably get them elsewhere.)

c. You could propose building a well-placed recycled  battery storage facility that is “charged up” by your hydro plants and then discharged during periods of peak demand.  This would also include AC/DC/AC  converters.  If these were suitably placed,  they would also solve your transmission loss problems at the same time.

d.  You are fortunate to have so much hydro available.  The proposed solution will also help your air pollution problem and may be a useful argument for eliminating fossil fuels.  In the referenced study (above), one of the negative concerns cited was that there was no good way to recycle batteries.  With your dependency on hydro, you can always make use of more and more recycled battery energy storage devices.  This is likely to be the fastest and cheapest way to increase your ability to meet demand – assuming you have hydro reserves and assuming you have access to recycled batteries.

Regards,

Gary Viviani, PhD, PE

Dear Komson Sir and GARY sir,

Thank a lot for the feedback. 

I would like to quote your 2nd point 

 " 2) From the characteristics of generations that rich of hydro, a unit commitment may be relax. Additionally, the OPF of minimizing fuel cost is not a crucial point. Reducing transmission loss (5%, very high) is a key issue. The successive is rely on a planning stage of upgrading transmission system and build new power plants at the proper location "

Actually, that is what my study is all about. Focusing on reducing the transmission line loss rather than minimizing the operating cost since the operating cost is almost constant and insignificant in case of Hydro Power Plants.

Regards,

Bishal

Hi Bishal,

If I understood it correctly, you are disregarding the peak hours where the total demand exceeds the total available generation capacity. Losses are unfortunately significant at peak hours because the flows in the transmission lines are higher. You said that the total system losses are 55 and 40 MW before and after optimization, respectively (which by the way is almost the size of the combined capacity from the diesel generators). Are these figures average values and represent for the off-peak hours? If so, you must have used the average demand for the off-peak hours. This is fine but I would personally optimize for each hour (including the peak hours) in a day. I would like to make the following points:

1. While I understand the requirement of slackness for an optimization problem to work, I thought of a way to trick this for the peak hours. In the balance equations, I would add a slack variable (which will collect the unserved power at each node). This variable can be understood as a fictitious generator whose marginal cost is equal to the estimated load shedding cost. I would add a penalty term in the objective function (in addition to the losses), which is to be minimized. This concept would help you to deal with all demand hours (peak and off-peak ones alike) and simultaneously minimize the load being shed and losses ("Too birds with one stone").

2. In my opinion, a 30% reduction of losses in such a system is impressive by any measure. This is not only technically but also economically sound. For most of the hours in a day, I presume the diesel generators are peaking units (according to the merit-order concept). So, reducing losses by such a margin means reducing the power that would otherwise be generated by the diesel generators to cover those losses by the same margin. Since such generators are expensive, the economic benefit of the losses reduction can be considerable.

3. System operators (load dispatch centers) are supposed to carry out such optimizations dynamically. Based on a predicted hourly load, the dispatch is carried out in advance (one day ahead, for example). The hourly productions of each generator are often known apriori. This is not often done in developing nations, though.

Regards,

Desta

Dear Desta Sir,

Thank you very much for such a wonderful review. 

I would like to quote some part of your content here.

1.

"Are these figures average values and represent for the off-peak hours? If so, you must have used the average demand for the off-peak hours. This is fine but I would personally optimize for each hour (including the peak hours) in a day. I would like to make the following points:" 

ANS: 

These loads mentioned in the summary, although roughly average, represent only a value of the load demand for a particular instant of time. I would have to consider such multiple load demands over various period of time to explore to what extent the transmission loss could be minimized under those particular conditions. Although, the load demand is very dynamic in nature and couldn't be replicated perfectly, i will try to select the load subset as extensively as possible. And yes, i agree that the load consideration shall be fairly frequent, as mentioned by you, at least once a hour.

2.

" While I understand the requirement of slackness for an optimization problem to work, I thought of a way to trick this for the peak hours. In the balance equations, I would add a slack variable (which will collect the unserved power at each node). This variable can be understood as a fictitious generator whose marginal cost is equal to the estimated load shedding cost. I would add a penalty term in the objective function (in addition to the losses), which is to be minimized. This concept would help you to deal with all demand hours (peak and off-peak ones alike) and simultaneously minimize the load being shed and losses ("Too birds with one stone")"

Response: 

I haven't considered this algorithm under the over demand i.e peak demand condition. So, you mean to say that it is possible to run the optimization concept even in the power deficit condition and somehow could be made to work in tandem with load shedding algorithm ?

Could you please mention some reference or some literature in this topic.

This in fact could be a huge value addition in my study.

Thanks and regards,

Bishal

Hi Bishal,

That's right. I meant you can run the optimization for all load points. When you introduce a positive load shedding variable (on the generation side of the load balance equations) which varies with bus index and time (hour), it means any load curtailed due to technical reasons (congestion, generation capacity deficits, outages, etc.) would be met by a "fictitious" generators that burn "Chanel No 5" (one of the most expensive perfumes).

Generally, if you formulate your problem taking this into account, the total generation capacity in the system will be equal to the total (demand + losses) in the system. This approach is pretty common in power systems optimization (either in operation or planning). The following relevant works can be useful:

Conference Paper Transmission network restoration considering AC power flow constraints

Article New Multi-Stage and Stochastic Mathematical Model for Maximi...

Hi bishal

i an also an student of power system. I am in Bangalore and i am also planning to do system studies of power system network of Nepal. so can u please help me i need data for load flow analysis can u please provide me.

please mail me at [email protected]

Dear Arun,  

Good to see u r interested  in the mentioned topic.  

Do contact me on my email [email protected],  specifically what  u need. Data of INPS are quite vague and scattered so,  Lets see if i could assist  in some way.  

Regards, 

Bishal