Something in a recent discussion with a NeuCo colleague made me dig up an ancient memory about how far we’ve come with respect to computing power.  I was recalling the early days when I worked at CRA International (formerly Charles Rivers Associates). It was the early 1990’s, and I had developed a computer model that performed stochastic simulations of large thermal-hydro regional electric power systems. The model determined the optimal combination of power plants and demand-side management that would be required over a 20-year time horizon in the face of uncertain fuel prices, demand growth, and rainfall (i.e. hydro-electric availability).
Another CRA Principal was depending on my model for a big project with an IPP. The project was facing a major deadline with unexpected client demands. And while my model was very accurate, it was incredibly difficult to use. So even though I was starting my vacation and driving with my young son to Florida, I agreed to perform the simulations during the drive.

My model required a 486 CPU to cope with the computational demands it imposed. Back then, heat and space constraints limited laptops to 386 CPUs, and also did not provide enough memory.  So I carried my 486 desktop computer from the 43rd floor of the Hancock Tower in Boston, put it in the car trunk, picked up my son, and set off for Santa Rosa Beach, Florida.

Pay phones and early optimization
My colleague at CRA needed a new simulation about every two hours, and each one took roughly 30 minutes to run. Cell phones had been recently invented and I didn’t have one. So I pulled off the interstate every two hours, found a restaurant or gas station that would help me out (the one thing that got easier as I traveled further south), pulled the desktop out of the car’s trunk, plugged it in to a payphone or other landline, ran the simulation, and then called my colleague back in Boston to read him the detailed results from the model’s balance sheets, income statements, and sources and uses of funds statements. 

The project was a success and I enjoyed the rest of my vacation in Florida. But this memory reminded me of how far we’ve come in the last 15 years in our ability to optimize power plant processes in real-time. The multi-gigahertz dual-core, dual CPU servers used by NeuCo’s ProcessLink® optimizers cost substantially less than my old 486 and yet have literally hundreds of times the processing power, memory, and storage capacity. If Moore’s Law continues to prove true, we’ve only scratched the surface of optimization’s potential. And as power plants evolve, the developers and engineers at NeuCo will continue to think creatively about how we can harness increasing computational power and extend our technology to address increasingly complex power plant issues.

www.theoptimizationblog.com

www.neuco.net