Feature Article
Electrical System Reliability Analysis on a PC
Powertechnic has announced a break-through in the reliability analysis of complex power systems. New software has been developed to analyse the reliability of a.c. and d.c. power systems on a desktop computer. Systems can include secondary batteries which have traditionally been difficult to model.
Following two and a half years of research and development, the break-through was achieved in September 1997, when a circuit containing 22 components, took around 10 minutes of model building and verification and a further 4 minutes to carry out the simulation on a Pentium 90 personal computer. This new technology is a major step forward in the reliability analysis of electrical systems at the engineering work-front. The new software allows electrical engineers to analyse systems without any specialised reliability skills or expertise in logical model building.
The technique places no limitations on the complexity of the power system. Complex switching arrangements can be set up and modelled. Systems with time dependent components such as batteries and generators are easily incorporated. Systems with repair can be modelled and each component can have its own failure, repair time and distribution.
Traditional Reliability Analysis
Engineers are continually faced with calls to
carry out reliability analysis. This may be to make A-B comparisons between design
options, or to quantify or justify expenditure on reliability grounds. For all but the
simpler systems, deterministic methods quickly become inadequate. Some methods rely on
reducing the systems to series-parallel equivalents and then applying block-diagram
techniques. The example of Fig1
shows a simple system that cannot be reduced to series-parallel.
For more complex systems logical models are built so that Monte-Carlo methods may be employed. In this latter approach, the limitations are two-fold. Considerable time and effort is needed to learn to build and test logical models. The quality of the model relies heavily on the skills of the modeller. Second, systems are commonly found that do not easily convert to a logical model. Systems with storage elements such as batteries and generators (with finite fuel reserves) quickly become impractical to model without many simplifying assumptions.
Break-through
The new technology does not rely on logical
models and does not require gross simplifying assumptions about the operational logic of
the electrical system. Building and verifying models remains within the domain of the
electrical designer.
Users work within a graphical environment drawing a one-line schematic of the system to be studied. The schematic is then checked for correct electrical function. The schematic provides dynamic information enabling users to quickly verify its electrical and logical operation. To facilitate this process tools are provided such as the SPOF (Single Point Of Failure) Analysis aid which systematically takes the user through the failure of every component in turn - see Fig2. Once the schematic has been drawn and verified, reliability analysis is carried out by introducing "System" components. As the name implies, a System component represents any group of electrical loads that equate to a real-world system of interest. For example consider a set of UPS units powering several computer loads. The computer loads are encapsulated in one or more System components, then, the failure/success logic is chosen within each System component. If a System component contains two computer loads and both must be available for mission success, then "parallel" logic applies. System components may contain other System components allowing complex logic combinations. The process of adding System components is quick and effective as the logic is represented visually on screen. The reliability analysis is then carried out.
Output
The software can output the number of system failures, system and
component MTBF & MTTR, system and component availability & unavailability. The
technique enhances other methods of analysis such as FMECA (Failure Mode Effects and
Criticality Analysis) which may be carried out from an intuitive graphical environment.
Performance
A Pentium 90 computer was used as the test bed
for performance assessment. A system of average complexity was set up which included a
standby generator and three lead acid batteries (Fig3). The inclusion of batteries provided for
a worst-case scenario due to the heavy calculation demands of a component with
continuously variable output. The software was able to calculate 92 gamed events per
second, completing the 24,000 calculations needed to generate a statistically significant
number of system failures in just 4 minutes and 21 seconds. In schematics without
batteries, calculation rates were much higher with figures of over 2,000 gamed events per
second. These figures suggest that calculation times for most systems will be well within
the useable range of desktop computers. Considering the advances now taking place in
desktop processing power, the new technology appears set for wide application in a range
of complex power system analysis problems on the desktop.
Development
The R&D was supported by Telstra's Product
Development Fund which provides investment capital for new ventures of special interest to
Telstra. The primary goal was to develop a commercially viable method of carrying out
complex reliability calculations. A variety of computer platforms and operating systems
were considered including high-end workstations, multi-processor systems, 16 and 32-bit
operating systems, etc. Ultimately due to commercial requirements high-end platforms were
ruled out as being too restrictive, 16 bit platforms were ruled out due to concerns about
speed limitations leaving 32 bit operating systems and single processor desktop pc's. C++
was chosen as the development language as it provides un-compromised speed performance in
an Object-Oriented (OO) design environment. Furthermore it is often the language of choice
amongst simulation practitioners having its earliest roots in simulation applications.
"With no prior development track record in OO, the learning curve proved to be very
manageable and it quickly became clear that as a software technology, OO has a well-earned
reputation for robustness." Special high-performance algorithms were developed to
carry out the calculations in the a.c. and d.c. domains. The algorithms were individually
optimised for each domain. The development was layered into two levels, the graphical user
interface and the calculation engine. This provided a natural means of devolving parts of
the project to separate companies. The layering also helped to maintain a rigorous
interface so that when eventually the layers were joined, very few problems were
experienced.
Data Security
It is vital that the data which feeds the
reliability analysis process is protected from corruption. Data management can quickly
become nightmarish as system complexity rises and users must manage hundreds or even
thousands of variables. A feature called CertLock (Fig4) was developed to provide a means for
certifying the data used in component models. The data is "locked" by encryption
to a unique, recognised certifying authority code. Any attempt to change the data results
in loss of certification. This means that users can have greater confidence that the data
they use has been tested and certified by a recognised authority.
Application
The reliability of a.c. and d.c. power systems
is of universal importance to the electrical industry. Companies that provide essential
public services such as mains power, telecommunications, transport and aviation, health
and emergency services, are just a few that rely critically on a reliable supply of power.
Electrical consultants are frequently called on to provide reliability analysis services.
Most systems however, are too complex to analyse. Typically the analysis is carried out by
academic or research institutes offering specialised services in reliability analysis. The
costs in doing this often exceed available budgets and engineers are left to manage these
tasks in-house.
Lead-Acid Battery Model
The lead-acid battery has been one of the main
stumbling blocks to creating an integrated a.c. and d.c. software design environment. Its
characteristics are highly non-linear making it difficult to model. Powertechnic has
developed a model that can be used to emulate both the discharge and re-charge
characteristics over a wide range of current. The model is accurate enough to replace
battery manufacturers discharge data sheets and can be used to estimate capacities in
situations beyond the constant current and power figures available from data sheets.
Commercial Release
The technology is now available in two product
forms. Power Designer provides a design environment for both a.c. and d.c systems and is
suited to companies that must maintain a design database for managing many power system
sites.
The second product, Analyst is a super-set of Power Designer and includes the reliability analysis technology described in this article.