At this level, the transmission network is assumed to be perfectly reliable (a "copper plate"). The solution focuses solely on whether the total generating capacity is sufficient to meet the total system load.
Before evaluating reliability, Billinton insists on a precise definition of the "solution." In his framework, an engineering system is reliable if it satisfies three conditions:
The "solution" to a reliability problem, therefore, is not a single number but a set of probabilistic indices that quantify the frequency, duration, and magnitude of failures. Billinton famously argued that a deterministic "margin" (e.g., 15% spare capacity) is a poor solution because it ignores the stochastic nature of component failure and load variation.
“For the average user, how many minutes per year is the system dead?” At this level, the transmission network is assumed
World-class: <10 minutes/year. Developing grid: >1,000 minutes/year.
Pro move: Model two states – up and down – as a Markov chain. Billinton showed that even a 2-state model catches 80% of real risk.
Problem: A factory has two parallel power feeds from different substations. Seems reliable. The "solution" to a reliability problem, therefore, is
Billinton’s calculation:
Result:
Naïve view = 0.01% annual outage.
Actual = Loss of both feeds simultaneously = 1/2000 chance per year, but when switch fails → 10-hour outage.
Fix: Not more generators – just a faster, redundant switch. World-class: <10 minutes/year
A classic mistake: treating all failures equally. Billinton’s genius was separating loss of load from inconvenience.
The Hierarchy of Failure (from his work): | Level | Event | Reliability Impact | |--------|--------|--------------------| | 1 | A light bulb burns out | Zero (system continues) | | 2 | One of two redundant pumps fails | Reduced margin, but no outage | | 3 | The single feed pump fails | System stops |
Your new mantra: “Redundancy without analysis is just expensive hope.”
Try this exercise:
Draw your system as a Reliability Block Diagram (RBD) – series vs. parallel.
You’ll immediately see where your real risk lives (hint: it’s always the single point of failure you forgot).