โ Scribed by R. Hodgson; H. Oman
No coin nor oath required. For personal study only.
Expert systems, readily implemented by battery specialists on personal computers with programs such as "empty shells", can now perform complex comparisons of alternative battery choices. The applications being evaluated might be power supplies, solar energy storage, emergency power, or customer-side-of-the-meter load leveling. Each has unique requirements in voltage limits and delivered energy during discharge, over a range of temperatures. Batteries being compared need not be identical in performance.
For example, in cycling service, two sets of $2000 batteries lasting three years each can be cheaper than a $4000 set lasting six years because of interest earned in the first four years by the unspent $2000.
Standardized performance data are a key to validity of these comparisons. A simple example is temperature at which the battery voltage and ampere hours are measured. The battery can be tested at any reasonable temperature, as long as a formula is available for converting test performance to performance at operating temperature.
A more difficult but very important extrapolation is life in a charge-discharge service. Data from an accelerated test at a given temperature and depth of discharge might have to be extrapolated to another temperature with a varying depth of discharge. Today, we have an opportunity to initiate the creation of a commonly useful expert system. Each of us can continue developing his own battery bank and his own ways to manipulate these data. Our present difficulties and ambiguities in comparing data and conclusions with each other will continue. Alternatively, we can adopt a standard way of defining, storing, and manipulating banks of battery data. In the ultimate we could exchange complex battery data by mailing computer disks to each other.
In this paper we try to identify the steps and effort that would be needed to achieve standard formats for organizing and presenting battery data. First, we illustrate the data and procedures vocabulary, with examples from our Boeing computer-based battery testing, data acquisition, and data analysis system.
๐ SIMILAR VOLUMES
Reasoning modelling and efficient numerical computation seem a priori incompatible: on one hand, Expert Systems are usually designed in order to deal with symbolic data. They are able to solve problems in domains where computers failed before, but they are not able to perform efficiently sophisticat
Machine learning offers the possibifity of deriving knowledge from data automatically. In the application described, the results of corrosion tests are used to derive rules about the corrosion resistance of materials. These can be used in an expert system to develop new materials and to answer custo
In my paper "Use of the Gibbs sampler in expert systems", I attempted to survey statistics and genetics literature of interest to researchers in AII faded to stress, however, that the best reference for understanding Markov chain Monte Carlo (MCMC) methods is the paper by Hastings [3 ], which genera