The Hindu Business Line : Artificial intelligence in ‘power’ prediction

“With the advent of computer power in the early 1990s, they have become a widely studied and applied electric load-forecasting technique,” writes Rafal Weron in Modeling and Forecasting Electricity Loads and Prices: A Statist ical Approach ( www.wiley.com).

AI-based methods tend to flexible and can handle complexity and non-linearity, he explains. In contrast, statistical methods — which forecast the current load value by using a mathematical combination of the previous loads, and values of exogenous factors such as weather and social variables — are ‘often criticised for their limited ability to model the (usually) non-linear behaviour of load and related fundamental variables.’

But, why forecast, you may wonder. Because, in a competitive power market, where electricity can be bought and sold at market prices like any other commodity, “load forecasting has gradually become the central and integral process in the planning and operation of electric utilities, energy suppliers, system operators and other market participants,” Weron informs.

“The financial penalties for forecast errors are so high that research is aimed at reducing them even by a fraction of a per cent.”

Among the many examples of AI-based models that the book discusses is an artificial neural network (ANN) based on multiple strategies to capture various trends in the data. “The architecture is that of a multiplayer perceptron trained with the error back propagation algorithm.” An improved, third generation system, has two ANN forecasters — one for base load, and the other, for the change in load.

You would also read about a study by Srinivasan et al (1999) that combined fuzzy logic, neural networks and expert systems, “in a highly automated hybrid STLF (short term load forecasting) approach with Kohonen’s self-organising feature map and unsupervised learning.”

A different study by Ho et al (1990), based on 11 ‘day types’ and weather parameters, used an algorithm that performed ‘better for Taiwan system loads than a classical ARIMA (autoregressive integrated moving average) model.’ SVM (support vector machine) is another tool, with roots in statistical learning theory. Unlike ANN that tries to define complex functions of the input space, SVM performs ‘a non-linear mapping of the data into a high dimensional space.’ Then, it uses simple linear functions to create linear decision boundaries in the new space, says Weron.

Applications of SVMs have been successful in OCR (optical character recognition), early medical diagnostics, text classification and corporate bankruptcy analysis.

“Chen et al (2004) proposed an SVM to predict daily load in eastern Slovakia for the 31 days of January 1999. Their program was the winning entry of the competition organised by the EUNITE (European Network of Excellence on Intelligent Technologies for Smart Adaptive Systems) network.”

Prescribed read for the ‘power’ hungry.

Science, technology, and magic

“People today not only expect but demand everything from technology and make no distinction between destructive and productive technology,” writes Umberto Eco in an essay titled ‘science, technology, and magic’, included in Turning Back the Clock ( www.landmarkonthenet.com).

“The youngster who plays ‘Star Wars’ games on the computer, uses a cell-phone as if it were a natural extension of the eustachian tub, and chats on the Internet, lives in technology and cannot conceive of the existence of a different world, one without computers or telephones.”

That everything scientific is technological is a confusion created by the mass media bemoans the author. “Technology gives you everything instantly; science proceeds slowly,” he distinguishes.

“We obviously live in the age of speed… We are so accustomed to speed that we get angry if we can’t open our e-mail immediately or if our plane is delayed.” This ‘addiction to technology’ has nothing to do with the practice of science, declares Eco. “It has to do, instead, with the eternal resort to magic.”

The author narrates an example, drawn from the sixteenth century, about the Benedictine abbot Trithemius, a forerunner of cryptology, who “worked out secret codes for rulers and army chiefs — but to make his discoveries and formulas (today easy to create on a computer but brilliant for those days) more appetising, he demonstrated that his technique was in fact a magical operation thanks to which you could have angels transmit messages in secrecy over a long distance in a single second.”

Magic is indifferent to the long chain of causes and effects, Eco reasons. “Above all, it does not trouble itself to establish by constant experiment that there is a replicable relation between a cause and its effect. Hence its appeal, from primitive cultures to the Renaissance to the myriad occult sects to be found all over the Internet.”

Since technology does everything possible, we lose sight of the chain of cause and effect, rues Eco. “The first computer operators compiled their programs in Basic (acronym for Beginner’s All-purpose Symbolic Instruction Code), which was not a machine language but gave us a glimpse of its mystery… Windows has obscured Basic programming… The user experiences computer technology as if it were magic.”

As a result, the science that emerges through the mass media is, unfortunately, “only its magical aspect. When science does appear, it appears because it promises a miracle of technology.” Alas, “it’s hard to convey to the public that research is a combination of hypotheses, control experiments, and proofs of falsification.” Not surprisingly, therefore, the society, deep into the magic mentality, ‘remains indifferent when big cuts are announced in research funding.’

Eco finds that everything boils down to ‘the ever-victorious short circuit between the cause and effect.’ For example, “antibiotic pills are a technological product; research into influenza causes and remedies is work for a university.” He is distressed that people in government (‘who sometimes consult magicians and astrologers’) think like the man in the street and not like the man in the laboratory. “It is pointless to ask the media to abandon the magic mentality,” frets Eco. “They must adopt it because the relationship they draw between cause and effect every day is of a magical nature.”

The only hope, according to him, is in the classroom. “It is up to schools, and to all initiatives that can educate, including reliable Internet sites, to ensure that young people gradually acquire a correct understanding of scientific procedure.”

Not an easy task though, Eco concedes, “because even knowledge transmitted by schools is often deposited in the memory like a sequence of miraculous episodes: Madame Curie who comes home one evening and discovers radioactivity thanks to a mark on a sheet of paper, Dr Fleming who glances absently at some mold and discovers penicillin, Galileo who sees a lamp swaying and suddenly discovers everything, even that the world rotates….”

Deep insights.

Tailpiece

“He is so illiterate…”

“That?”

“He thinks ‘thumb drive’ is something like a signature campaign!”

dmurali@thehindu.co.in