By Kumar David –

Prof. Kumar David

Part I of III for layman; if you are an expert, Go Away!

Planning future additions to the electricity generation system is a long-term exercise extending over 20 to 30 years because if the order in which different types (technologies) and sizes of plant are added is incorrect (not optimal) it can be very costly and the system may be difficult to operate as load fluctuates or when something unforeseen happens. These are hundreds of billions of rupee (billions not millions) decisions not to be messed around with by amateurs nor an opportunity for amateurs with scant experience to fool politicians. Nevertheless it is possible to explain the basics to the concerned public – which will have to pay for it anyway. There is a story, apocryphal no doubt, that Einstein was wont to remark that Relativity could be explained to anyone in a way apposite to his familiarity with physics if one had mastered the theory oneself.

There is a storm brewing about the government’s hush-hush cancellation of the proposed Sampur coal-power joint venture project with India and fear that Lanka will be high and dry facing power shortages in a few years. Therefore ‘generation expansion planning’ is of public interest. I will set out the principles of planning as simply as possible and lay groundwork for dissection, in the next two thrilling instalments (sic!) of recent power sector conundrums and cock-ups.

Since the plan must stretch 20 to 30 years into the future it is divided into many stages for computational convenience; as an example let us assume a 20 year plan and divide it into five 4-year stages. Of course the planner is not committed for all 20 years since it is a rolling plan and only the immediate ‘next’ decision is important. A year or two later the planner redoes it all again, using the latest information then available, and again makes the best decision at that stage. A very critical but unsettled parameter is the anticipated load over this 20 year term; economic conditions, government policy, and global factors can play havoc with forecasts. Another surprise is technology as with the recent remarkable decline in solar electricity costs and rise in photovoltaic (PV) efficiencies. A rolling plan allows updating and correction and facilitates best decisions at each stage.

Generation Expansion Planning (GES)

A big difficulty in GES is what is called the curse of dimensionality as there is a whole gamut of different aspects or dimensions to bear in mind all at the same time. There are several future time stages to plan over. Many different fuel options (hydro if not already exhausted, nuclear, coal, natural gas and oil) and technology options (one big unit or a series of smaller ones, types of turbines, simple or combined cycle, reciprocating engines – gigantic cousins of the diesel engine under your car bonnet – and of course solar and wind whose dynamic behaviour is very different) are available. There are even more ‘dimensions’ to consider; ‘nested’ into expansion decisions are system operating issues corresponding to the state of the system at each intermediate stage; more on this ‘nested’ stuff in a moment. Finally, one does not wish to optimise only capital costs but capital, fuel and in the case of nuclear retirement costs; and of course one wishes to minimise not immediate costs but the sum of costs considered over the whole long period discounted to present values.

Let me illustrate the ‘curse of dimensionality’. Say we have 15 fuel-cum-technology options to choose from (fuels, locations, technologies, plant sizes and environmental options). Then think of it like this. We have 15 possible alternatives to choose now, that is in stage 1 – assume all are feasible – so there are 15 potential end points of the system (system states) at the end of stage 1. If in stage 2 all 15 are feasible again we have 15 more choices from each of the end of stage-1 states, or 225 different states at which the system may be at the end of stage-2. It’s like a tree branching out. If you go on with this game for 5 stages you end up with 759, 375 possible states at any one of which the system may end up (mathematically) after 20 years. Of course it never gets this crazy. Hundreds of intermediate options are ruled out at each step along the way for financial, technical (‘nested’), environmental or political reasons. Thus the multiplicative expansion is heavily curbed; still the problem is big and complicated enough to demand powerful computational packages.

One of the things that make it complicated is what I called the ‘nested’ issues. It is not enough to go stage by stage in economic analysis; that’s only the start of a headache. One has to pause and see if every state along the way is operable in the technical sense (power dispatch, reserve margins, stable, secure). If not this intermediate state is discarded; but what I am (the computer will) grumble about is that one has to pause at every step, go off and do a whole lot of checks about the operability of every one of the hundreds (may be thousands) of intermediate optional states.

And all this just to get one ‘best’ path progression through the 20 year five stages! The ‘best’ (optimal in the jargon) is the cheapest path from start to finish that is also feasible. This then is the “optimal” among all feasible long term options. (Actually the best few, called suboptimals, are worth bearing in mind). And all this just to choose the first step on this path because, as I alerted you, next year the planner will make the computer do all this work again because data would have changed in the interim – the rolling plan concept. Actually it’s not quite as bad as one can do sensitivity studies and prepare oneself for eventualities.

There are powerful computer programs. In my view the best algorithm for problems of this stage-by-stage nature is Dynamic Programming (DP). The CEB uses WASP (Wein Automatic System Planning), a proven package which utilises DP and checks many nested concerns. The US Electric Power Research Institute put out, with much fanfare, a package using an optimisation routine called gradient climbing nonlinear programming, but there were no takers; it seems to have been withdrawn. Another factor that needs pushing in Lanka but has not received much attention from the government is demand side management.

My intention today is not to teach you power system planning in one easy lesson but rather to give you a flavour of how serious a responsibility it is and that when politicos, on their own steam or under the influence of quacks, throw out carefully planned decisions with the aplomb of calling out to the chef “Change my order from omelette to scrambled egg” they are in all likelihood making billion dollar blunders.

Fuels and Technologies

Let’s keep it brief. Lanka has exploited nearly all its hydro and the bit left will be used soon. Nuclear like hydro is capital intensive but cheap to run but it will be resisted by the public for fear of accidents, radiation and hazardous waste. (It is probably no longer true that nuclear in sizes absorbable by Lanka are not available; the French are offering 150 MW plant at competitive prices). Mini hydro and firewood will make small contributions but must no longer be accepted at prices above avoided coal or LNG (liquefied natural gas) prices. Onshore wind potential is small unless you swallow the Grimm’s Fairy Tale accounts circulating in some quarters. Last month two offers of wind power at Rs 12 per kWh were received by the CEB. This is attractive but how much more is there available? Oil fired electricity penetrated Lanka due to the stupidity and cupidity of political leaders in the1990s and early 2000s. It is going to happen again; but more on that next week.

That leaves two base load bearing options – coal and LNG. The long term future may belong to LNG because coal is more environmentally damaging. This is not to say environmentally decent high-tech coal plant cannot be absorbed at all by Lanka, but more on that next week. Coal prices are fairly stable while LNG prices are volatile though attractively low right now. The only honest statement that can be made is that there is no reason to expect, at this moment in time, that one of these fuels will have a cost advantage over the other in the long-run. In that case LNG wins long-term from environmental and health considerations – the short-term is a cock-up reserved for next week. There is a point of concern about LNG; not even Lanka’s daft politicos will stick an LNG power plant in Trinco, the LNG harbour must be near load centres on the West coast to be financially viable. The $500 million land based harbour needed for LNG can only be justified if other usages (transport, industry, etc.) are involved. A floating harbour facility is cheaper than a land based harbour but entails an annual rental of about $50 million.

Solar electricity

The surprise in the last year has been a surge of interest and investment in solar electricity; PV (photo voltaic) technology, not concentrated solar electricity generation (CSG), is becoming a champion. CSG is where in a square kilometre or so of desert, curved mirrors are arranged around a central tower. The mirrors track the sun and focus radiation high on the tower where a potassium or sodium salt is liquefied. This primary fluid passes through ground level heat-exchangers where water is boiled, steam generated, turbines rotated and electricity generated in the usual way. The technique is irrelevant to Lanka.

PV is breaking new ground. In places high on the Andes and in desserts (Arabia, Middle East, US, India and North Western China) they are doing brisk business and expanding. In the best locations the new prices are competitive with coal and LNG; in Lanka solar insolation is less intense or reliable, and given the island’s demographics, accessible locations are limited. Nevertheless 1000 MW of utility (CEB) level PV is economically viable and can be installed within 10 to 15 years if a concerted effort is made. At 30% plant factor this is 2.6 terra-watt-hours of electricity a year or 20% of current generation – say 10% of 2030 output. Average prices at fairly good sites may fall as low as Rs 10 to 12 per kWh in the light of global technological trends.

There are three technical hitches with PV if used on a large scale. It is stochastic, meaning the sun may shine or clouds and rain may win at a moment’s notice, so normal plant must be built (capital cost duplication) and be available at all times. It is non-dispatchable, meaning the operator can’t get out of bed in the morning and say “I am going to dispatch so much PV today; take that Vicoria-3 out for maintenance”. And thirdly PV plant does not add “inertia” to the system, meaning it does not help to smooth out frequency fluctuations or to ride through disturbances to system stability.