Algorithms generate answer to complex problem
Generating the right amount of power to meet the varying needs of an entire country is a delicate and demanding task.
In the UK, the National Grid is responsible for managing electricity generation and transmission. Its object is to balance, from moment to moment, the electricity produced by dozens of different sources with the demand from homes and industry.
Among the factors is it has to take into account are the endless variations in demand, the price charged by each generator for their power along with how much they can produce, transmission constraints, and the ever-present need to reduce emissions of greenhouse gases.
Since 1993, the algorithms National Grid uses to solve this complex problem have come from the Brunel Institute of Power Systems, under a series of research contracts awarded to Professor Malcolm Irving and Professor Gareth Taylor.
In its most recent form, the National Grid’s balancing mechanism uses a Sparse Dual Revised Simplex (SDRS) model to determine the optimum pattern for starting up and shutting down generators over a designated scheduling period, and for allocating generating targets to the power providers so that the predicted demand is satisfied at minimum cost. Under this system, generators are given revised target outputs every five minutes.
The sums involved are enormous. Annually, the UK’s electricity market is valued at around £5 billion. The balancing mechanism operates on a margin of 2-3% of the total transmission capacity, meaning that if it works effectively, it can save £100-£150 million each year.
The most recent set of software, SDRS2, has been shown to be an order of magnitude more efficient than commercially available optimisation packages. Its developers believe that this is due to its focus on power dispatch problems, and the refinements available from more than 20 years of continuous development and experience.