Solved Problems · Set 13

Medium and Long Lines Compared

Part 3 · Line Performance — three models of one line, the errors each incurs, and the single dimensionless number that decides between them. Chapters 10–12 reviewed.

Prof. Mithun Mondal 20 solved · 12 practice · 3 challenge GATE · ESE · University

Set 13 — Medium and Long Lines Compared

Twenty worked problems on choosing between the models of Sets 9, 11 and 12. They are not rivals but successive truncations of one series in the electrical length \(\gamma l\), and the choice between them is therefore a question about a single dimensionless number rather than about kilometres. This set quantifies the errors, locates the boundaries, and shows where the conventional length-based rules break down — for cables, at harmonic frequencies, and near resonance.

Textbook Chapters 1012 · 20 solved · 12 practice · 3 challenge · 12 MCQs

i Method Recap
  • One series, three truncations. Short line: \(A = 1\), \(B = Z\), \(C = 0\). Nominal-\(\pi\): \(A = 1 + \frac{ZY}{2}\), \(B = Z\), \(C = Y(1+\frac{ZY}{4})\). Exact: \(\cosh\gamma l\), \(Z_c\sinh\gamma l\), \(\sinh(\gamma l)/Z_c\).

  • The criterion is \(\gamma l\), not the length. Errors scale with powers of the electrical length, so a rule in kilometres is only a proxy — valid for overhead lines at 50 Hz because \(\beta\) is nearly the same for all of them, and invalid for cables and harmonics.

  • Error growth. The nominal-\(\pi\)'s error in \(B\) grows as \((\gamma l)^2\) and its error in \(A\) as \((\gamma l)^4\). A doubling of length therefore quadruples one and multiplies the other by sixteen.

  • The conventional boundaries. Short below about 80 km (\(\beta l < 5^\circ\)), medium to about 250 km (\(\beta l < 16^\circ\)), long beyond. They correspond to errors of roughly 0.4%, 1.4% and more.

  • Cables are electrically longer than their length suggests. A cable's velocity is about \(0.63c\) against \(0.94c\) overhead, so its wavelength is only two thirds as long and a given kilometre counts for half as much again.

  • Harmonics multiply the electrical length. \(\beta\) scales with frequency, so a 400 km line is 26° long at 50 Hz and 75° long at the third harmonic — close to quarter-wave resonance, with a voltage magnification of nearly four.

  • Cascaded sections converge quadratically. Splitting a line into \(n\) nominal-\(\pi\) sections reduces the error by \(n^2\). It is the route when hyperbolic functions are unavailable, and it is how transient programs represent a line.

VideoWalkthrough
Problem 1Exam levelThree Models

A line has \(z = 0.16 + j0.40\ \Omega\)/km and \(y = j3.0\times10^{-6}\) S/km. For a length of 300 km, compute the ABCD constants by all three models and set them side by side.

Solution

The two governing quantities, from Set 12:

\[ \gamma = 1.1369\times10^{-3}\angle79.10^\circ\ \text{/km}, \qquad Z_c = 378.95\angle-10.90^\circ\ \Omega \]
\[ \gamma l = 0.3411\angle79.10^\circ = 0.0645 + j0.3349, \qquad \beta l = 19.19^\circ \]

Total lumped parameters:

\[ \mathbf{Z} = 48 + j120\ \Omega = 129.24\angle68.20^\circ, \qquad \mathbf{Y} = j9.0\times10^{-4}\ \text{S} \]
\[ \mathbf{Z}\mathbf{Y} = -0.108 + j0.0432 \]

Short line: \(A = 1\), \(B = 129.24\angle68.20^\circ\), \(C = 0\).

Nominal-\(\pi\):

\[ A = 1 + \frac{\mathbf{Z}\mathbf{Y}}{2} = 0.946 + j0.0216 = 0.94625\angle1.31^\circ \]
\[ B = 129.24\angle68.20^\circ\ \Omega, \qquad C = 8.839\times10^{-4}\angle90.87^\circ\ \text{S} \]

Exact:

\[ A = \cosh\gamma l = 0.94665\angle1.31^\circ \]
\[ B = Z_c\sinh\gamma l = 126.93\angle68.55^\circ\ \Omega, \qquad C = 8.839\times10^{-4}\angle90.36^\circ\ \text{S} \]

Collecting:

\[ \begin{array}{lccc} & \text{Short} & \text{Nominal-}\pi & \text{Exact} \\ \hline |A| & 1.0000 & 0.94625 & 0.94665 \\ |B|\ (\Omega) & 129.24 & 129.24 & 126.93 \\ |C|\ (\times10^{-4}\ \text{S}) & 0 & 8.839 & 8.839 \end{array} \]
The short-line model is not slightly wrong here — it is wrong by 5.6% in \(A\) and infinitely wrong in \(C\). Setting \(C = 0\) is not an approximation of \(8.84\times10^{-4}\); it is the assertion that the line draws no charging current at all. The nominal-\(\pi\), by contrast, is within 2% of the exact model on every constant. That gap between "qualitatively wrong" and "quantitatively imperfect" is what this whole set is about.
AnswerShort: \(A = 1\), \(C = 0\). Nominal-\(\pi\): \(|A| = 0.94625\), \(|B| = 129.24\). Exact: \(0.94665\), \(126.93\)
Problem 2Exam levelError in A

Tabulate the nominal-\(\pi\) error in \(|A|\) against length for the line of Problem 1, and deduce the scaling law.

Solution

Computing \(|\cosh\gamma l|\) and \(|1 + \mathbf{Z}\mathbf{Y}/2|\) at each length:

\[ \begin{array}{cccccc} l\ (\text{km}) & \beta l\ (^\circ) & |A|_{\text{exact}} & |A|_{\pi} & \text{error (\%)} & \text{ratio} \\ \hline 100 & 6.40 & 0.99401 & 0.99400 & 0.001 & — \\ 200 & 12.79 & 0.97613 & 0.97605 & 0.011 & 11 \\ 400 & 25.58 & 0.90605 & 0.90482 & 0.195 & 18 \\ 600 & 38.38 & 0.79454 & 0.78875 & 1.119 & — \\ 800 & 51.17 & 0.65041 & 0.63486 & 4.272 & 22 \end{array} \]

Each doubling of length multiplies the error by roughly 16 to 22.

The scaling law. From Set 12 Problem 16, the neglected term is the next in the cosh series:

\[ \cosh\gamma l = 1 + \frac{(\gamma l)^{2}}{2} + \frac{(\gamma l)^{4}}{24} + \cdots \]
\[ \text{error} \approx \frac{(\gamma l)^{4}}{24} \]

A fourth-power law, so a doubling should multiply the error by 16 — which the table confirms, the excess coming from the next term again.

Checking at 400 km, where \(|\gamma l| = 0.4547\):

\[ \frac{(0.4547)^{4}}{24} = \frac{0.04275}{24} = 1.78\times10^{-3} = 0.178\% \]

against the observed 0.195% — agreeing to the accuracy of a one-term estimate.

A fourth-power law means the nominal-\(\pi\) is remarkably good in \(A\) until quite suddenly it is not. At 400 km the error is a fifth of a per cent; at 800 km it is four per cent. This is characteristic of truncated series: they are far better than one expects over most of their range and then deteriorate rapidly. It is also why a length-based rule can work at all — the transition is sharp enough to have a meaningful boundary.
AnswerError grows as \((\gamma l)^4/24\) — 0.011% at 200 km, 0.195% at 400, 4.27% at 800
Problem 3Exam levelError in B

Repeat Problem 2 for \(|B|\), and explain why the error is so much larger than in \(A\).

Solution

The nominal-\(\pi\) takes \(B = \mathbf{Z}\) flat, while the exact value is \(Z_c\sinh\gamma l\):

\[ \begin{array}{ccccc} l\ (\text{km}) & \beta l\ (^\circ) & |B|_{\text{exact}} & |B|_{\pi} & \text{error (\%)} \\ \hline 50 & 3.20 & 21.53 & 21.54 & 0.054 \\ 100 & 6.40 & 43.00 & 43.08 & 0.216 \\ 200 & 12.79 & 85.48 & 86.16 & 0.866 \\ 300 & 19.19 & 126.93 & 129.24 & 1.963 \\ 400 & 25.58 & 166.87 & 172.33 & 3.525 \\ 600 & 38.38 & 240.32 & 258.49 & 8.163 \\ 800 & 51.17 & 302.38 & 344.65 & 15.121 \end{array} \]

Each doubling multiplies the error by about four — a second-power law rather than a fourth.

Why. The exact \(B\) expands as

\[ B = Z_c\sinh\gamma l = \mathbf{Z}\left(1 + \frac{\mathbf{Z}\mathbf{Y}}{6} + \cdots\right) = \mathbf{Z}\left(1 + \frac{(\gamma l)^{2}}{6} + \cdots\right) \]

and the nominal-\(\pi\) keeps only the leading \(\mathbf{Z}\). It therefore discards a term of order \((\gamma l)^2\), where in \(A\) it discarded one of order \((\gamma l)^4\).

Checking at 400 km:

\[ \frac{|\gamma l|^{2}}{6} = \frac{0.2068}{6} = 0.0345 = 3.45\% \]

against the observed 3.53%.

The nominal-\(\pi\) is a second-order model in \(A\) and a zeroth-order model in \(B\), and that asymmetry decides everything. It is the error in \(B\) that sets the model's useful range, because \(B\) governs the voltage drop under load and the power transfer limit. At the conventional 250 km boundary the error in \(B\) is 1.4% while that in \(A\) is only 0.03% — a factor of fifty. Anyone judging the model by its \(A\) alone would use it far beyond its competence.
AnswerError grows as \((\gamma l)^2/6\) — 0.87% at 200 km, 3.53% at 400, 15.1% at 800
Problem 4Warm-upError in C

Tabulate the nominal-\(\pi\) error in \(|C|\), and explain why it is smaller than the error in \(B\) although both are second-order.

Solution

The values:

\[ \begin{array}{cccc} l\ (\text{km}) & |C|_{\text{exact}}\ (\times10^{-4}) & |C|_{\pi}\ (\times10^{-4}) & \text{error (\%)} \\ \hline 100 & 2.994 & 2.991 & 0.108 \\ 200 & 5.952 & 5.926 & 0.436 \\ 400 & 11.620 & 11.409 & 1.814 \\ 600 & 16.735 & 16.007 & 4.350 \\ 800 & 21.057 & 19.277 & 8.454 \end{array} \]

Also a second-power law — each doubling multiplies the error by four — but the coefficient is smaller.

Why smaller. The exact \(C\) expands as

\[ C = \mathbf{Y}\left(1 + \frac{\mathbf{Z}\mathbf{Y}}{6} + \cdots\right) \]

while the nominal-\(\pi\) takes \(\mathbf{Y}(1 + \mathbf{Z}\mathbf{Y}/4)\) — it does include a correction, but with the wrong coefficient.

The discrepancy is therefore the difference between the two coefficients:

\[ \text{error} \approx \left(\frac{1}{4} - \frac{1}{6}\right)|\gamma l|^{2} = \frac{|\gamma l|^{2}}{12} \]
\[ \text{At 400 km: } \frac{0.2068}{12} = 1.72\%\ \text{against the observed } 1.81\% \]

Exactly half the error in \(B\), as the coefficients \(1/12\) and \(1/6\) predict.

The nominal-\(\pi\)'s \(C\) correction is right in form and wrong by half in magnitude, which is oddly worse than having none at all would suggest. Had it simply taken \(C = \mathbf{Y}\) the error would have been \(|\gamma l|^2/6\) — twice as large. So the \(1/4\) factor does help, just not as much as it appears to. The nominal-T makes the mirror-image mistake in \(B\), which is why Set 11 Problem 11 found the two models erring in opposite directions.
AnswerError grows as \(|\gamma l|^2/12\) — exactly half that in \(B\): 0.44% at 200 km, 1.81% at 400
Problem 5Challenge-liteThe Criterion

Express the conventional length-based model boundaries in terms of electrical length, and state the conditions under which the length-based rule is valid.

Solution

For an overhead line at 50 Hz, \(\beta \approx 1.1\times10^{-3}\) rad/km, so:

\[ \begin{array}{lccl} \text{Model} & \text{Length} & \beta l & \text{Error in } B \\ \hline \text{Short} & < 80\ \text{km} & < 5^\circ & < 0.14\% \\ \text{Medium} & 80\text{–}250\ \text{km} & 5^\circ\text{–}16^\circ & < 1.4\% \\ \text{Long} & > 250\ \text{km} & > 16^\circ & > 1.4\% \end{array} \]

Why a length rule works at all. The phase constant is

\[ \beta \approx \omega\sqrt{LC} = \frac{\omega}{v} \]

and Set 6 Problem 16 showed that \(v \approx c\) for every overhead line regardless of its construction, because \(LC \approx \mu_0\varepsilon_0\). So \(\beta\) is nearly a universal constant at a given frequency, and kilometres translate directly into degrees.

The three conditions for validity:

\[ \begin{array}{ll} \text{Overhead line} & \text{Cables have } v \approx 0.63c,\ \text{so } \beta \text{ is 50\% larger} \\ \text{Power frequency} & \beta \propto f,\ \text{so harmonics scale the electrical length} \\ \text{50 Hz} & \text{At 60 Hz every boundary shrinks by 5/6} \end{array} \]

At 60 Hz, for instance, the medium-line boundary falls from 250 to about 208 km for the same electrical length.

The kilometre rule is a coincidence of overhead construction, and it is worth knowing that it is one. A 200 km cable is electrically a 300 km overhead line (Problem 15); a 400 km line carrying third-harmonic current is electrically a 1200 km line (Problem 17). In both cases the familiar rule gives the wrong model, and the only reliable criterion is \(\gamma l\) itself — which costs one square root to compute and removes the guesswork entirely.
AnswerShort \(\beta l < 5^\circ\), medium \(5^\circ\)\(16^\circ\), long above; valid only for overhead lines at power frequency
Problem 6Exam levelRegulation Compared

The 300 km line of Problem 1 delivers 100 MW at 220 kV, 0.9 power factor lagging. Find the voltage regulation predicted by each model.

Solution

The load, per phase:

\[ \mathbf{V}_R = 127\,017\angle0^\circ\ \text{V}, \qquad \mathbf{I}_R = 291.6\angle-25.84^\circ\ \text{A} \]

Computing \(V_S = |A\mathbf{V}_R + B\mathbf{I}_R|\) and then \(\%\text{reg} = (V_S/|A| - V_R)/V_R\) for each:

\[ \begin{array}{lccc} & V_S\ (\text{kV}) & |A| & \%\text{reg} \\ \hline \text{Short line} & 271.82 & 1.00000 & 23.55 \\ \text{Nominal-}\pi & 260.95 & 0.94625 & 25.35 \\ \text{Exact} & 259.78 & 0.94665 & 24.74 \end{array} \]

Two errors that partly cancel. The short-line model overstates \(V_S\) by 4.6% — it has no shunt admittance to divert current from \(B\). But it also takes \(|A| = 1\), missing the 5.6% Ferranti rise. The two mistakes work in opposite directions and its regulation comes out only 1.2 points low.

The nominal-\(\pi\) gets \(|A|\) almost exactly right but overstates \(|B|\) by 1.8%, so it overstates the regulation by 0.6 points.

The short-line model gave a nearly right answer for entirely wrong reasons, and that is the most dangerous kind of agreement. Its two large errors happened to oppose each other at this loading and this power factor. Change either — go to a leading load, or to no load — and the cancellation vanishes: at no load the short-line model predicts zero regulation where the truth is 5.6%. An approximation that works by cancellation cannot be trusted outside the case where the cancellation was verified.
AnswerShort 23.55%, nominal-\(\pi\) 25.35%, exact 24.74%
Problem 7Warm-upFerranti Compared

Find the Ferranti rise predicted by each model for the line of Problem 1 at 150, 300 and 600 km.

Solution

The rise is \(1/|A| - 1\) in every case:

\[ \begin{array}{cccc} l\ (\text{km}) & \text{Short} & \text{Nominal-}\pi & \text{Exact} \\ \hline 150 & 0\% & 1.37\% & 1.36\% \\ 300 & 0\% & 5.68\% & 5.64\% \\ 600 & 0\% & 26.78\% & 25.86\% \end{array} \]

The short-line model predicts zero at every length, because \(A = 1\) exactly. This is not an error that shrinks as the line shortens — it is the wrong answer at any length whatever.

The nominal-\(\pi\) tracks the exact model closely, and — worth noting — it overstates the rise slightly at every length:

\[ \text{At 600 km: } 26.78\%\ \text{against } 25.86\%,\ \text{an overstatement of } 3.6\% \]

This is the safe direction for design, though at much greater lengths it will reverse, since the lumped model has no pole.

This is the sharpest illustration of the difference between an error of magnitude and an error of kind. At 150 km the short-line model's \(V_S\) is within half a per cent of the truth, and one might reasonably use it. Its Ferranti prediction at that same length is not within half a per cent — it is zero against 1.36%, an error of one hundred per cent. Which quantity is being asked for decides whether the model is excellent or useless, and the length alone does not.
AnswerShort predicts 0% always; nominal-\(\pi\) 1.37/5.68/26.78%; exact 1.36/5.64/25.86%
Problem 8Exam levelPower Limit Compared

Find the maximum power transfer predicted by each model for the line of Problem 1 at 150, 300 and 600 km, with both terminals at 220 kV.

Solution

From Set 11 Problem 19, the maximum receiving-end power is

\[ P_{\max} = \frac{3V_R^{2}}{|B|}\left[1 - |A|\cos(\beta - \alpha)\right] \]

where \(\alpha\) and \(\beta\) are here the angles of \(A\) and \(B\).

Evaluating:

\[ \begin{array}{cccc} l\ (\text{km}) & \text{Short} & \text{Nominal-}\pi & \text{Exact} \\ \hline 150 & 470.8 & 470.8 & 474.2 \\ 300 & 235.4 & 235.4 & 242.2 \\ 600 & 117.7 & 117.7 & 131.5 \end{array} \]

in MW.

The short line and nominal-\(\pi\) agree exactly, because both take \(B = \mathbf{Z}\) and the power limit depends chiefly on \(|B|\). Their agreement is not evidence of correctness — both are wrong together.

Both understate the limit, by 1.4% at 300 km and 10.5% at 600 km:

\[ \frac{131.5 - 117.7}{131.5} = 10.5\% \]

The understatement follows directly from their overstatement of \(|B|\), since \(P_{\max} \propto 1/|B|\).

Two models agreeing is worth nothing when they share the same assumption. Both take \(B = \mathbf{Z}\), so both inherit the same 8% error in \(|B|\) at 600 km and the same 10.5% understatement of the power limit. Agreement between models is evidence only when the models differ in the assumption being tested — and here the nominal-\(\pi\)'s improvement over the short line lies entirely in \(A\) and \(C\), neither of which dominates the power limit.
AnswerShort and nominal-\(\pi\) both give 470.8/235.4/117.7 MW; exact 474.2/242.2/131.5 MW
Problem 9Warm-up150 km

Assess the three models for a 150 km line and recommend one.

Solution

Electrical length:

\[ \beta l = 9.59^\circ = 0.167\ \text{rad} \]

Errors of the nominal-\(\pi\) against exact:

\[ \begin{array}{lccc} & \text{Exact} & \text{Nominal-}\pi & \text{Error} \\ \hline |A| & 0.98654 & 0.98651 & 0.004\% \\ |B|\ (\Omega) & 64.33 & 64.62 & 0.49\% \\ |C|\ (\times10^{-4}) & 4.480 & 4.469 & 0.24\% \end{array} \]

Under load at 100 MW, 0.9 lagging:

\[ \begin{array}{lccc} & V_S\ (\text{kV}) & \%\text{reg} & \text{Ferranti} \\ \hline \text{Short} & 245.11 & 11.41 & 0\% \\ \text{Nominal-}\pi & 242.26 & 11.62 & 1.37\% \\ \text{Exact} & 242.11 & 11.55 & 1.36\% \end{array} \]

Recommendation: nominal-\(\pi\). Its errors are under half a per cent in every constant, its regulation is within 0.07 points of the exact value, and its Ferranti prediction is within 0.01 points.

The short-line model would give \(V_S\) 1.2% high and no Ferranti effect at all — acceptable only if the Ferranti rise is irrelevant to the study, which for a line that will be energised before being loaded it is not.

150 km sits comfortably inside the medium band, and the nominal-\(\pi\) costs nothing to use. Both models require the same data; the \(\pi\) requires one extra multiplication. Given that, there is no reason to use the short-line model on anything but the shortest feeders — its only genuine advantage is that it can be computed on paper without complex arithmetic, which mattered in 1920 and does not now.
AnswerNominal-\(\pi\) — errors under 0.5%; the short line misses the 1.36% Ferranti rise entirely
Problem 10Exam level300 km

Repeat Problem 9 for a 300 km line — the borderline case.

Solution

Electrical length:

\[ \beta l = 19.19^\circ = 0.335\ \text{rad} \]

Beyond the conventional 16° boundary, so formally a long line.

Errors of the nominal-\(\pi\):

\[ \begin{array}{lccc} & \text{Exact} & \text{Nominal-}\pi & \text{Error} \\ \hline |A| & 0.94665 & 0.94625 & 0.06\% \\ |B|\ (\Omega) & 126.93 & 129.24 & 1.96\% \\ |C|\ (\times10^{-4}) & 8.839 & 8.751 & 1.00\% \end{array} \]

Under the same 100 MW load:

\[ \begin{array}{lccc} & \%\text{reg} & \text{Ferranti} & P_{\max}\ (\text{MW}) \\ \hline \text{Short} & 23.55 & 0\% & 235.4 \\ \text{Nominal-}\pi & 25.35 & 5.68\% & 235.4 \\ \text{Exact} & 24.74 & 5.64\% & 242.2 \end{array} \]

The verdict depends on the study. For regulation and Ferranti the nominal-\(\pi\) is within 0.6 points and 0.04 points — entirely adequate. For the power limit it is 2.8% low, which for a stability margin is marginal.

Recommendation: nominal-\(\pi\) for load-flow and voltage studies; exact constants for stability and power-limit work. The extra cost is one evaluation of two hyperbolic functions.

The boundary between models is not a property of the line — it is a property of the line and the question. At 300 km the same model is excellent for one purpose and marginal for another, which is why the conventional single boundary at 250 km is a simplification. The honest procedure is to compute the error in the constant the study actually depends on, which Problem 19 sets out formally.
AnswerNominal-\(\pi\) adequate for voltage work (0.6 points in regulation), marginal for power limits (2.8% low)
Problem 11Exam level600 km

Repeat Problems 9 and 10 for a 600 km line, and state what the lumped models get badly wrong.

Solution

Electrical length:

\[ \beta l = 38.38^\circ = 0.670\ \text{rad} \]

The constants:

\[ \begin{array}{lccc} & \text{Exact} & \text{Nominal-}\pi & \text{Error} \\ \hline |A| & 0.79454 & 0.78875 & 1.12\% \\ |B|\ (\Omega) & 240.32 & 258.49 & 8.16\% \\ |C|\ (\times10^{-4}) & 16.735 & 16.007 & 4.35\% \end{array} \]

Under the same 100 MW, 0.9 lagging load at 220 kV:

\[ \begin{array}{lcccc} & V_S\ (\text{kV}) & \%\text{reg} & \text{Ferranti} & P_{\max}\ (\text{MW}) \\ \hline \text{Short} & 328.47 & 49.30 & 0\% & 117.7 \\ \text{Nominal-}\pi & 289.44 & 66.80 & 26.78\% & 117.7 \\ \text{Exact} & 280.37 & 60.39 & 25.86\% & 131.5 \end{array} \]

What goes badly wrong. Three failures, in order of severity:

\[ \begin{array}{ll} \text{Short-line } V_S & 17\%\ \text{high; regulation 11 points low; no Ferranti at all} \\ \text{Nominal-}\pi\ V_S & 3.2\%\ \text{high; regulation 6.4 points high} \\ \text{Both } P_{\max} & 10.5\%\ \text{low, from the 8\% error in } |B| \end{array} \]

Recommendation: exact constants only. At \(\beta l = 38^\circ\) neither lumped model is defensible for any quantity.

Notice that the short-line model's regulation error has now reversed sign. At 300 km it was 1.2 points low; here it is 11 points low, and the two large opposing errors that cancelled so conveniently in Problem 6 no longer come close to balancing. Cancellation is a coincidence of a particular electrical length, and it cannot be relied upon at any other. This is the general hazard of validating an approximation at one operating point.
AnswerErrors of 1.1% in \(A\), 8.2% in \(B\); regulation 60.4% exact against 66.8% and 49.3% — exact constants required
Problem 12Exam levelCascaded Sections

Represent the 600 km line as two, then four, nominal-\(\pi\) sections in cascade, and find the error in \(|B|\) each time.

Solution

Each of \(n\) sections has \(\mathbf{Z}/n\) and \(\mathbf{Y}/n\), and the overall constants are the \(n\)-fold matrix product of the section matrices.

For \(n = 2\), each section covers 300 km with \(\mathbf{Z}_1 = 48 + j120\) and \(\mathbf{Y}_1 = j9.0\times10^{-4}\):

\[ A_1 = 0.94625\angle1.31^\circ, \quad B_1 = 129.24\angle68.20^\circ, \quad C_1 = 8.751\times10^{-4}\angle90.87^\circ \]
\[ B = 2A_1B_1 = 244.59\angle69.51^\circ\ \Omega \]

Against the exact \(|B| = 240.32\ \Omega\):

\[ \text{error} = \frac{244.59 - 240.32}{240.32} = 1.78\% \]

Down from 8.16% with a single section — a reduction of 4.6 times.

Continuing:

\[ \begin{array}{ccc} n & |B|\ (\Omega) & \text{error (\%)} \\ \hline 1 & 258.49 & 8.163 \\ 2 & 244.59 & 1.908 \\ 3 & 242.20 & 0.838 \\ 4 & 241.37 & 0.469 \\ 6 & 240.79 & 0.208 \\ 8 & 240.58 & 0.117 \end{array} \]

The error falls almost exactly as \(1/n^2\): from \(n = 2\) to \(n = 4\) it drops by a factor of 4.07, and from 4 to 8 by 4.01.

Quadratic convergence, and the reason is that each section's own error is quadratic in its electrical length. Halving the section length quarters each section's error, and there are twice as many sections — net effect a factor of four. This is exactly how an electromagnetic transient program represents a distributed line when the hyperbolic solution is unavailable in the time domain: a chain of short lumped sections, each short enough that its own lumping error is negligible.
AnswerError falls 8.16% → 1.91% → 0.47% for \(n = 1, 2, 4\) — a \(1/n^2\) law
Problem 13Warm-upHow Many Sections

How many nominal-\(\pi\) sections are needed to represent the 600 km line with an error below 1% in \(|B|\)? Derive a general rule.

Solution

From the table of Problem 12, the error passes below 1% between \(n = 2\) (1.91%) and \(n = 3\) (0.84%):

\[ n = 3\ \text{sections of 200 km each} \]

The general rule. Each section has electrical length \(\gamma l/n\), and from Problem 3 its error in \(B\) is

\[ \varepsilon \approx \frac{1}{6}\left|\frac{\gamma l}{n}\right|^{2} \]

Setting this equal to the target \(\varepsilon_{\text{target}}\) and solving:

\[ n \ge |\gamma l|\sqrt{\frac{1}{6\varepsilon_{\text{target}}}} \]

For the 600 km line, \(|\gamma l| = 0.6821\) and \(\varepsilon_{\text{target}} = 0.01\):

\[ n \ge 0.6821\sqrt{\frac{1}{0.06}} = 0.6821(4.082) = 2.78 \quad\Rightarrow\quad n = 3\ \checkmark \]

Equivalently, and more memorably: each section should be electrically shorter than

\[ \left|\frac{\gamma l}{n}\right| \le \sqrt{6\varepsilon_{\text{target}}} = 0.245\ \text{rad} = 14^\circ\ \text{for 1\%} \]
The rule is about the section, not the line — each piece must be electrically short, and how many pieces that takes is arithmetic. A 14° section for 1% accuracy is essentially the medium-line boundary of Problem 5, which is no coincidence: cascading is simply the statement that a long line is a chain of medium ones. For a transient program working at surge frequencies, where \(\beta\) is thousands of times larger, the same rule demands sections of a few hundred metres — which is precisely what such programs use.
AnswerThree sections of 200 km; generally \(n \ge |\gamma l|/\sqrt{6\varepsilon}\), or each section under 14° for 1%
Problem 14Exam levelA Cable

A cable has \(R = 0.03\ \Omega\)/km, \(L = 0.15\) mH/km and \(C = 0.17\ \mu\text{F}\)/km. Find its \(\gamma\), \(Z_c\), wavelength and velocity, and compare with the overhead line.

Solution

Per-km parameters at 50 Hz:

\[ x = 2\pi(50)(0.15\times10^{-3}) = 0.04712\ \Omega/\text{km}, \qquad b = 2\pi(50)(0.17\times10^{-6}) = 5.341\times10^{-5}\ \text{S/km} \]

Note the striking contrast with the overhead line: the reactance is eight times smaller and the susceptance eighteen times larger.

Propagation constant and characteristic impedance:

\[ \gamma = \sqrt{zy} = 1.7273\times10^{-3}\angle73.76^\circ\ \text{/km} \]
\[ \alpha = 4.831\times10^{-4}\ \text{Np/km}, \qquad \beta = 1.6584\times10^{-3}\ \text{rad/km} \]
\[ Z_c = \sqrt{z/y} = 32.34\angle-16.24^\circ\ \Omega \]

Wavelength and velocity:

\[ \lambda = \frac{2\pi}{\beta} = 3789\ \text{km}, \qquad v = f\lambda = 1.894\times10^{5}\ \text{km/s} = 0.63c \]

Comparing directly:

\[ \begin{array}{lcc} & \text{Overhead} & \text{Cable} \\ \hline Z_c\ (\Omega) & 379 & 32 \\ \beta\ (\text{rad/km}) & 1.116\times10^{-3} & 1.658\times10^{-3} \\ \lambda\ (\text{km}) & 5628 & 3789 \\ v/c & 0.94 & 0.63 \\ \alpha\ (\text{Np/km}) & 2.15\times10^{-4} & 4.83\times10^{-4} \end{array} \]
A cable's surge impedance is a twelfth of an overhead line's, and that single fact drives everything. It follows directly from the geometry of Sets 5 and 6: millimetres of dielectric instead of metres of air give a large \(C\) and a small \(L\), so \(\sqrt{L/C}\) collapses while \(\sqrt{LC}\) — and hence \(\beta\) — rises. The consequence for the surge impedance loading is dramatic: a 132 kV cable has an SIL of \(132^2/32 = 545\) MW against an overhead line's 46 MW, which is why cables are almost never operated anywhere near it.
Answer\(\gamma = 1.727\times10^{-3}\angle73.76^\circ\), \(Z_c = 32.3\angle-16.2^\circ\ \Omega\), \(\lambda = 3789\) km, \(v = 0.63c\)
Problem 15Challenge-liteCable Boundaries

Find the model boundaries for the cable of Problem 14, and compute its Ferranti rise at 50, 100 and 200 km.

Solution

The boundaries in electrical length are the same as for any line — \(5^\circ\) and \(16^\circ\). Converting to kilometres with \(\beta = 1.6584\times10^{-3}\) rad/km:

\[ \text{Short: } l < \frac{0.0873}{1.6584\times10^{-3}} = 53\ \text{km} \]
\[ \text{Medium: } l < \frac{0.2793}{1.6584\times10^{-3}} = 168\ \text{km} \]

Against the overhead values of 80 and 250 km:

\[ \frac{53}{80} = 0.66, \qquad \frac{168}{250} = 0.67 \]

Both ratios equal the wavelength ratio \(3789/5628 = 0.673\), as they must.

The Ferranti rise at each length:

\[ \begin{array}{cccc} l\ (\text{km}) & \beta l\ (^\circ) & |A| & \text{rise} \\ \hline 50 & 4.75 & 0.9969 & 0.32\% \\ 100 & 9.50 & 0.9875 & 1.27\% \\ 200 & 19.00 & 0.9504 & 5.21\% \end{array} \]

Comparing with the overhead line, where 200 km gave 2.45%: the cable's rise at 200 km is that of a 300 km overhead line.

A cable is electrically half again as long as its length suggests, and the kilometre-based rules are therefore wrong for it by a third. This matters because the charging-current problem of Set 6 already limits a.c. cables to about 80 km — so in practice the model question rarely arises, since no such cable is long enough to be a "long line". Where it does arise is in HVDC cable schemes during a.c. fault studies, and in submarine links where the a.c. sections at each end feed cables of unusual length.
AnswerShort below 53 km, medium to 168 km — two thirds the overhead values; Ferranti 0.32/1.27/5.21% at 50/100/200 km
Problem 16Exam levelAt Harmonics

The 400 km overhead line carries harmonic currents. Find its electrical length and \(|A|\) at the fundamental and at the 3rd, 5th, 7th and 11th harmonics.

Solution

At harmonic order \(h\) both the reactance and the susceptance scale with frequency:

\[ x_h = hx_1, \qquad b_h = hb_1 \quad\Rightarrow\quad \beta_h \approx h\beta_1 \]

The resistance is taken as unchanged, neglecting skin effect — which understates the damping at high orders.

Evaluating \(\gamma_h l\) and \(A = \cosh\gamma_h l\) at each order:

\[ \begin{array}{cccc} h & \beta_h l\ (^\circ) & |A| & 1/|A| \\ \hline 1 & 25.6 & 0.9060 & 1.10 \\ 3 & 75.5 & 0.2655 & 3.77 \\ 5 & 125.6 & 0.5891 & 1.70 \\ 7 & 175.8 & 1.0012 & 1.00 \\ 11 & 276.2 & 0.1393 & 7.18 \end{array} \]

The pattern is not monotonic. \(|A| = |\cosh\gamma l|\) behaves as \(|\cos\beta l|\) damped by the losses, so it passes through minima near \(\beta l = 90^\circ, 270^\circ\) and maxima near \(0^\circ, 180^\circ\).

The 7th harmonic, at \(\beta l = 176^\circ\), is almost exactly a half wavelength — and there \(|A| \approx 1\), so the line is transparent to it.

A 400 km line is a fourteenth of a wavelength at 50 Hz and three quarters of a quarter-wave at the third harmonic. Every length-based intuition built at power frequency fails immediately at harmonics, because \(\beta\) scales with \(h\) while the line does not change. This is why harmonic studies never use the short-line or nominal-\(\pi\) models on transmission lines — the exact constants must be recomputed at every order of interest.
Answer\(\beta l\) = 25.6°, 75.5°, 125.6°, 175.8°, 276.2°; \(|A|\) = 0.906, 0.266, 0.589, 1.001, 0.139
Problem 17Challenge-liteHarmonic Resonance

Find the harmonic order at which the 400 km line reaches quarter-wave resonance, the voltage magnification at the third harmonic, and the practical consequences.

Solution

Quarter-wave resonance occurs at \(\beta_h l = 90^\circ\). Since \(\beta_1 l = 25.58^\circ\):

\[ h_{\text{res}} = \frac{90}{25.58} = 3.52 \]

Between the third and fifth harmonics — and since only odd harmonics are present in a balanced three-phase system, the third is the nearest.

Magnification at the third harmonic. From Problem 16, \(|A| = 0.2655\):

\[ \frac{V_R}{V_S} = \frac{1}{|A|} = 3.77 \]

A third-harmonic voltage injected at the sending end appears at the open receiving end nearly four times larger.

Why it does not diverge. Two things limit it: the line is not exactly at resonance (75.5° rather than 90°), and the resistance damps it. Even so, a 2% third-harmonic distortion at the source becomes 7.5% at the far end.

The practical consequences:

\[ \begin{array}{ll} \text{Voltage distortion} & \text{Amplified towards the open or lightly loaded end} \\ \text{Capacitor banks} & \text{Can resonate with the line at a nearby order} \\ \text{Filter design} & \text{Must account for the line, not just the source} \\ \text{Measurement} & \text{Distortion measured at one end misrepresents the other} \end{array} \]

The problem is worst on a lightly loaded long line, since load damps the resonance — exactly the condition that also produces the worst Ferranti effect.

Third-harmonic resonance on a 400 km line is not a hypothetical: it is why harmonic studies for long radial lines are done at all. In a balanced system third harmonics are zero-sequence and largely trapped by delta windings, so the fifth and seventh usually dominate — and a line resonant near the fifth would be more troublesome still. The remedy is the same as for the Ferranti effect: shunt reactors, which shift the resonant order upward by reducing the effective \(Y\), and are one more reason they are fitted.
AnswerResonance at \(h = 3.52\); the third harmonic is magnified 3.77 times at the open end
Problem 18Warm-upEquivalent-π

Compute the equivalent-\(\pi\) parameters of the 600 km line and compare with its nominal-\(\pi\), showing that the equivalent form removes the need to choose a model at all.

Solution

From Set 12 Problem 10, the exactly equivalent lumped parameters are

\[ Z' = Z_c\sinh\gamma l = B, \qquad \frac{Y'}{2} = \frac{1}{Z_c}\tanh\frac{\gamma l}{2} \]

With \(\gamma l = 0.1290 + j0.6699\) for 600 km:

\[ Z' = 240.32\angle68.86^\circ\ \Omega \]
\[ \tanh\frac{\gamma l}{2} = 0.3403\angle77.61^\circ \quad\Rightarrow\quad \frac{Y'}{2} = 8.981\times10^{-4}\angle88.51^\circ\ \text{S} \]

Against the nominal-\(\pi\), which uses \(\mathbf{Z} = 258.49\angle68.20^\circ\) and \(\mathbf{Y}/2 = 9.0\times10^{-4}\angle90^\circ\):

\[ \begin{array}{lcc} & \text{Nominal-}\pi & \text{Equivalent-}\pi \\ \hline Z\ (\Omega) & 258.49\angle68.20^\circ & 240.32\angle68.86^\circ \\ Y/2\ (\text{S}) & 9.000\times10^{-4}\angle90^\circ & 8.981\times10^{-4}\angle88.51^\circ \end{array} \]

The series impedance differs by 7.6% and the shunt admittance by only 0.2% — consistent with the errors found in Problems 3 and 4.

The point. The equivalent-\(\pi\) is a lumped circuit of three elements, exactly like the nominal one, and slots into a \(Y\)-bus identically. But it reproduces the hyperbolic constants exactly at any length.

The equivalent-\(\pi\) makes the whole model-selection question disappear, and that is how modern practice sidesteps this set entirely. A load-flow program computes \(Z'\) and \(Y'\) once from the hyperbolic functions when the line data is prepared, and thereafter treats every line — 5 km or 500 — as three lumped elements with no approximation whatever. The short and nominal-\(\pi\) models survive as hand-calculation tools and as the conceptual ladder that makes the exact solution intelligible.
Answer\(Z' = 240.32\angle68.86^\circ\ \Omega\), \(Y'/2 = 8.981\times10^{-4}\angle88.51^\circ\) S — exact at any length
Problem 19Challenge-liteWhich Error Matters

Different studies depend on different constants. Determine which constant dominates each of the common studies, and hence the model each requires.

Solution

Mapping studies to constants:

\[ \begin{array}{lll} \text{Study} & \text{Depends chiefly on} & \text{Relation} \\ \hline \text{No-load voltage, Ferranti} & A & V_R = V_S/|A| \\ \text{Voltage regulation} & A \text{ and } B & \text{both terms} \\ \text{Load-flow voltage drop} & B & \mathbf{V}_S - A\mathbf{V}_R = B\mathbf{I}_R \\ \text{Power transfer limit} & B & P_{\max} \propto 1/|B| \\ \text{Charging current, reactors} & C & I_S = |C|V_R \\ \text{Losses, efficiency} & \text{Re}(B) & \text{resistive part} \\ \text{Harmonic studies} & \text{all, at each } h & \beta \propto h \end{array} \]

Combining with the error laws of Problems 2 to 4, where the nominal-\(\pi\) errs as \((\gamma l)^4/24\) in \(A\), \((\gamma l)^2/6\) in \(B\) and \((\gamma l)^2/12\) in \(C\):

\[ \begin{array}{lccc} \text{Study} & \text{300 km} & \text{600 km} & \text{Nominal-}\pi\ \text{usable to} \\ \hline \text{Ferranti} & 0.06\% & 1.12\% & \sim700\ \text{km} \\ \text{Regulation} & 0.6\ \text{pts} & 6.4\ \text{pts} & \sim350\ \text{km} \\ \text{Power limit} & 2.8\% & 10.5\% & \sim250\ \text{km} \\ \text{Charging current} & 1.0\% & 4.4\% & \sim300\ \text{km} \end{array} \]

So the boundary depends on the study by nearly a factor of three — from about 250 km for power-limit work to 700 km for Ferranti estimates.

The conventional single boundary at 250 km is therefore the most conservative of these, chosen so that one rule serves every purpose.

The single 250 km boundary is a safe simplification, not a physical fact, and knowing which constant a study depends on can extend or contract it substantially. The one case where no relaxation is permitted is any study involving \(A\) at zero load — because there the short-line model's error is not a percentage but a category error, giving zero where the truth is 26%. Percentage error is a meaningful measure only when the model gets the phenomenon qualitatively right in the first place.
AnswerNominal-\(\pi\) serves Ferranti work to ~700 km but power-limit work only to ~250 km — the conventional boundary is the most conservative
Problem 20Exam levelSelection Procedure

Set out a procedure for choosing a line model, and apply it to four cases: a 40 km 33 kV feeder; a 180 km 220 kV line; a 500 km 400 kV line; and a 120 km 132 kV cable.

Solution

The procedure:

\[ \begin{array}{ll} \text{1.} & \text{Compute } \gamma = \sqrt{zy} \text{ at the frequency of interest} \\ \text{2.} & \text{Form } |\gamma l|;\ \text{if harmonics matter, repeat at each order} \\ \text{3.} & \text{Identify which constant the study depends on (Problem 19)} \\ \text{4.} & \text{Estimate the error: } (\gamma l)^4/24 \text{ for } A,\ (\gamma l)^2/6 \text{ for } B \\ \text{5.} & \text{If the error exceeds tolerance, use exact or equivalent-}\pi \\ \text{6.} & \text{If the study involves no-load behaviour, never use the short line} \end{array} \]

Case 1 — 40 km, 33 kV feeder. \(\beta l = 2.6^\circ\), \(|\gamma l| = 0.045\):

\[ \text{error in } B \approx \frac{(0.045)^{2}}{6} = 0.03\% \]

Short line. The charging current is negligible and the feeder is never operated unloaded for long.

Case 2 — 180 km, 220 kV line. \(\beta l = 11.5^\circ\), \(|\gamma l| = 0.205\):

\[ \text{error in } B \approx \frac{(0.205)^{2}}{6} = 0.70\% \]

Nominal-\(\pi\). Comfortably inside the medium band for every study.

Case 3 — 500 km, 400 kV line. \(\beta l = 32.0^\circ\), \(|\gamma l| = 0.568\):

\[ \text{error in } B \approx \frac{(0.568)^{2}}{6} = 5.4\% \]

Exact or equivalent-\(\pi\). And the line needs compensation regardless, which the model must represent.

Case 4 — 120 km, 132 kV cable. Using the cable's \(\beta = 1.658\times10^{-3}\): \(\beta l = 11.4^\circ\), \(|\gamma l| = 0.207\):

\[ \text{error in } B \approx 0.71\% \]

Nominal-\(\pi\), though the length alone would have suggested a short line by the overhead rule. The charging current is the dominant concern here in any case — at 120 km this cable is beyond its critical length.

Two of the four cases would have been decided differently by a kilometre rule, and one of those wrongly. The 120 km cable is electrically a 180 km overhead line, and a 40 km feeder and a 120 km cable are not in the same class despite both sounding "short". Computing \(\gamma l\) costs one square root and removes the ambiguity entirely — which is why every serious analysis does it rather than consulting a table of lengths.
AnswerShort / nominal-\(\pi\) / exact / nominal-\(\pi\) — the cable being electrically 50% longer than its length suggests
Practice

Practice Problems

Work each on paper before opening the answer. The answer is given so you can check yourself; the method is deliberately not.

  1. P1. A line has \(|\gamma l| = 0.25\). Estimate the nominal-\(\pi\) error in \(B\).

    Show answer
    \((0.25)^2/6 = 0.0104 = \mathbf{1.04\%}\).
  2. P2. For the same line, estimate the error in \(A\).

    Show answer
    \((0.25)^4/24 = 1.63\times10^{-4} = \mathbf{0.016\%}\) — sixty times smaller.
  3. P3. A 250 km line has \(\beta = 1.1\times10^{-3}\) rad/km. Which model?

    Show answer
    \(\beta l = 0.275\) rad \(= 15.8^\circ\) — just inside the medium band. Nominal-\(\pi\), with about 1.3% error in \(B\).
  4. P4. Why do the short-line and nominal-\(\pi\) models give identical power limits?

    Show answer
    Both take \(B = \mathbf{Z}\), and \(P_{\max} \propto 1/|B|\). Their agreement is no evidence of correctness — Problem 8.
  5. P5. How many nominal-\(\pi\) sections give under 0.5% error in \(B\) for a line with \(|\gamma l| = 0.68\)?

    Show answer
    \(n \ge 0.68/\sqrt{6(0.005)} = 0.68/0.173 = 3.9\), so \(n = \mathbf{4}\).
  6. P6. A cable has \(v = 0.63c\). What is its wavelength at 50 Hz?

    Show answer
    \(\lambda = v/f = 0.63(3\times10^5)/50 = \mathbf{3780}\) km, against about 5600 km overhead.
  7. P7. A 200 km cable and a 200 km overhead line — which is electrically longer, and by how much?

    Show answer
    The cable, by the inverse wavelength ratio \(5628/3789 = \mathbf{1.49}\) — it behaves like a 300 km overhead line.
  8. P8. A 300 km line has \(\beta_1 l = 19^\circ\). At what harmonic does it reach quarter-wave resonance?

    Show answer
    \(h = 90/19 = \mathbf{4.7}\) — dangerously close to the 5th harmonic, which is usually the dominant one.
  9. P9. Which constant governs the Ferranti effect, and which governs the power limit?

    Show answer
    Ferranti: \(\mathbf{A}\), through \(V_R = V_S/|A|\). Power limit: \(\mathbf{B}\), through \(P_{\max} \propto 1/|B|\) — Problem 19.
  10. P10. Why is the short-line model never acceptable for a no-load study, at any length?

    Show answer
    It sets \(A = 1\), predicting zero Ferranti rise. That is a category error, not a percentage error, and it does not shrink with length — Problem 7.
  11. P11. Write the equivalent-\(\pi\) parameters in terms of \(\gamma l\) and \(Z_c\).

    Show answer
    \(Z' = \mathbf{Z_c\sinh\gamma l}\) and \(Y'/2 = \mathbf{\tanh(\gamma l/2)/Z_c}\) — exact at any length.
  12. P12. A line at 60 Hz — how do the model boundaries in kilometres change?

    Show answer
    \(\beta \propto f\), so every boundary shrinks by \(50/60 = \mathbf{5/6}\): the medium limit falls from about 250 to 208 km.
Challenge

Challenge Problems

Each of these needs an idea rather than a formula. Decide what the governing principle is before opening the answer.

  1. C1. Problem 6 found the short-line model giving a nearly correct regulation through the cancellation of two large errors. Analyse when such cancellation occurs, why it is dangerous, and how to detect it.

    Show answer
    The two errors. The short-line model makes exactly two mistakes:
    — It sets \(A = 1\) instead of \(|A| < 1\), which understates the no-load voltage and hence the regulation.
    — It omits the shunt branch that would divert current from \(B\), which overstates \(V_S\) under load and hence overstates the regulation.

    When they cancel. Writing the regulation as \((V_S/|A| - V_R)/V_R\), the short-line model has \(V_S\) too large by roughly \(|\gamma l|^2/2\) relative and \(|A|\) too large by the same order. The two effects are of similar magnitude and opposite sign, so cancellation is not accidental — it is structural, at moderate lagging loads.

    It is exact when the two happen to balance, which for the 300 km line of Problem 6 occurred near 0.9 lagging. It fails when either term is suppressed:
    At no load, the second error vanishes (no current) and only the first remains: the model predicts 0% against 5.6%.
    At leading power factor, the impedance drop reverses sign and the errors add rather than cancel.
    At long lengths, the two errors grow at different rates — \(A\)'s as \((\gamma l)^4\), \(B\)'s effect as \((\gamma l)^2\) — so the balance is destroyed, as Problem 11 showed at 600 km.

    Why it is dangerous. A model validated at one operating point appears trustworthy and is not. The agreement carries no information about the model's fidelity, because it arises from a coincidence rather than from the model capturing the physics.

    How to detect it. Test the approximation at the extremes of the operating range, not at the nominal point — no load, full load, leading and lagging. If the error changes sign anywhere in that range, cancellation is occurring and the agreement at the nominal point is worthless. This is a general principle of approximation validation, not a fact about transmission lines.
  2. C2. The kilometre-based model boundaries work for overhead lines at power frequency. Establish exactly why, and enumerate every circumstance in which they fail.

    Show answer
    Why they work. The true criterion is \(\beta l\), and
    \[ \beta = \omega\sqrt{LC} = \frac{\omega}{v}, \qquad v = \frac{1}{\sqrt{LC}} \]
    Set 6 Problem 16 showed that \(LC \approx \mu_0\varepsilon_0\) for every overhead line, because the logarithms cancel in the product. So \(v \approx c\) universally and \(\beta\) is nearly a constant of nature at a given frequency — about \(1.05\times10^{-3}\) rad/km at 50 Hz. Kilometres therefore translate into degrees at a fixed rate, independent of voltage class, conductor size, bundling or spacing.

    Where they fail:
    Cables. The dielectric has \(\varepsilon_r \approx 2.3\), so \(v \approx 0.63c\) and \(\beta\) is 50% larger. Boundaries shrink to two thirds — Problem 15.
    Other frequencies. \(\beta \propto f\). At 60 Hz boundaries shrink by 5/6; at the fifth harmonic by a factor of five — Problem 16.
    Transients. At surge frequencies of tens of kilohertz, \(\beta l\) for even a single span exceeds the boundaries, which is why travelling-wave methods are mandatory there.
    Series-compensated lines. A capacitor cancels part of \(x\), reducing the effective \(\beta\) and making the line electrically shorter than its length — the deliberate purpose of the compensation.
    Shunt-compensated lines. Reactors reduce the effective \(b\), again shortening the line electrically. A 600 km line with 60% shunt compensation behaves like a 380 km one.
    Gas-insulated lines. Different \(L\) and \(C\) again.
    Zero-sequence networks. Different \(z\) and \(y\) entirely, and hence a different \(\beta\) — the earth-return path changes both.

    The general statement: the length rule is valid whenever the line's \(LC\) product is that of air at power frequency. Every failure above is a case where it is not. Computing \(\gamma\) directly costs one square root and is never wrong.
  3. C3. Modern practice uses the equivalent-\(\pi\) universally, which appears to make this entire set obsolete. Assess that claim.

    Show answer
    The claim has real force. A load-flow program computes \(Z' = Z_c\sinh\gamma l\) and \(Y'/2 = \tanh(\gamma l/2)/Z_c\) once at data preparation, then treats every line identically. No model selection occurs, no approximation is made, and the computational cost is two hyperbolic functions per line — negligible. In that sense the short and nominal-\(\pi\) models are indeed obsolete as computational tools.

    What survives, and why it matters:
    Judgement about results. An engineer who knows that \(|A| = 0.79\) means a 26% Ferranti rise can recognise an implausible load-flow output. One who has only ever seen \(Z'\) and \(Y'\) as numbers in a data file cannot.
    Hand estimation. Feasibility studies, examination work, and sanity checks on a program's output all need a calculation that can be done without software. \(V_S \approx V_R + IR\cos\phi + IX\sin\phi\) is still the fastest route to a number.
    Understanding the exact solution. The hyperbolic constants are opaque; the series expansion of Problem 16 makes them intelligible by connecting them to the lumped models. Nobody develops intuition about \(\cosh\gamma l\) directly.
    Knowing what the equivalent-\(\pi\) still cannot do. It is exact for terminal behaviour at one frequency in the steady state — and says nothing about the interior voltage profile (Set 12 Challenge C2), about transients, or about other frequencies. An engineer who believes the equivalent-\(\pi\) is simply "exact" will use it where none of those conditions hold.
    Transient programs. These cannot use the frequency-domain equivalent at all and fall back on cascaded lumped sections — Problems 12 and 13 — so the cascade analysis remains live.

    Verdict: obsolete as arithmetic, essential as understanding. The situation is exactly that of long division after the calculator: nobody performs it, and everybody must know what it does. The genuine risk of the equivalent-\(\pi\)'s universality is that it conceals the electrical length entirely, so a user never learns that a 120 km cable and a 40 km feeder are in different regimes — which is the specific error Problem 20 was constructed to expose.
Self-Test

Multiple-Choice Questions

  1. MCQ 1. The correct criterion for choosing a line model is:
    (a) the length in km   (b) the voltage class   (c) the electrical length \(\gamma l\)   (d) the conductor size

    Show answer
    (c). The length rule is a proxy valid only for overhead lines at power frequency — Problem 5.
  2. MCQ 2. The nominal-\(\pi\) error in \(B\) grows as:
    (a) \(\gamma l\)   (b) \((\gamma l)^2\)   (c) \((\gamma l)^3\)   (d) \((\gamma l)^4\)

    Show answer
    (b), as \((\gamma l)^2/6\). The error in \(A\) is fourth-power — Problems 2 and 3.
  3. MCQ 3. Which constant does the short-line model get qualitatively wrong?
    (a) \(A\)   (b) \(B\)   (c) \(C\)   (d) none

    Show answer
    (c). Setting \(C = 0\) denies the charging current entirely — a category error rather than a percentage one.
  4. MCQ 4. The short-line and nominal-\(\pi\) models give the same power limit because:
    (a) both are exact   (b) both take \(B = Z\)   (c) both take \(A = 1\)   (d) coincidence

    Show answer
    (b). \(P_{\max} \propto 1/|B|\), so they inherit the same error — Problem 8.
  5. MCQ 5. Splitting a line into \(n\) nominal-\(\pi\) sections reduces the error by:
    (a) \(n\)   (b) \(n^2\)   (c) \(\sqrt n\)   (d) \(2^n\)

    Show answer
    (b) \(n^2\) — each section's error is quadratic in its own electrical length — Problem 12.
  6. MCQ 6. A cable's wavelength compared with an overhead line's is about:
    (a) the same   (b) two thirds   (c) half   (d) twice

    Show answer
    (b) two thirds, since \(v \approx 0.63c\) against \(0.94c\) — Problem 14.
  7. MCQ 7. At the \(h\)-th harmonic, a line's electrical length is:
    (a) unchanged   (b) \(h\) times larger   (c) \(h^2\) times larger   (d) \(\sqrt h\) times larger

    Show answer
    (b). \(\beta \propto f\), so a 400 km line is 26° at 50 Hz and 75° at the third harmonic — Problem 16.
  8. MCQ 8. A 400 km line reaches quarter-wave resonance at about which harmonic?
    (a) 1.5   (b) 3.5   (c) 7   (d) 11

    Show answer
    (b) 3.52, giving a magnification of 3.77 at the third harmonic — Problem 17.
  9. MCQ 9. The equivalent-\(\pi\) circuit is:
    (a) an approximation better than nominal-\(\pi\)   (b) exact at any length   (c) valid only for short lines   (d) the same as nominal-\(\pi\)

    Show answer
    (b) exact, since it is constructed to reproduce the hyperbolic constants — Problem 18.
  10. MCQ 10. The nominal-\(\pi\) serves Ferranti studies to a greater length than power-limit studies because:
    (a) \(A\) converges faster than \(B\)   (b) \(B\) converges faster   (c) they are equal   (d) it does not

    Show answer
    (a). \(A\)'s error is fourth-order and \(B\)'s second-order, so the usable length differs by nearly threefold — Problem 19.
  11. MCQ 11. At 60 Hz the model boundaries in kilometres:
    (a) are unchanged   (b) increase by 6/5   (c) decrease by 5/6   (d) decrease by 25/36

    Show answer
    (c). \(\beta \propto f\), so a given electrical length is reached at a shorter physical length.
  12. MCQ 12. Two models agreeing on a result is evidence of correctness when:
    (a) always   (b) never   (c) they differ in the assumption being tested   (d) they use the same data

    Show answer
    (c). The short line and nominal-\(\pi\) agree on the power limit only because both share \(B = Z\) — Problem 8.
Reference

Key Formulas

QuantityRelationNotes
Short line\(A = 1\), \(B = Z\), \(C = 0\)Correct to \((\gamma l)^0\)
Nominal-π\(A = 1 + \frac{ZY}{2}\), \(B = Z\), \(C = Y(1+\frac{ZY}{4})\)\(A\) correct to \((\gamma l)^2\)
Exact\(\cosh\gamma l\), \(Z_c\sinh\gamma l\), \(\sinh\gamma l/Z_c\)All orders
Equivalent-π\(Z' = Z_c\sinh\gamma l\), \(Y'/2 = \tanh(\gamma l/2)/Z_c\)Exact, lumped form
Error in \(A\)\(\approx |\gamma l|^4/24\)Fourth power
Error in \(B\)\(\approx |\gamma l|^2/6\)Second power — sets the boundary
Error in \(C\)\(\approx |\gamma l|^2/12\)Half that in \(B\)
Boundariesshort \(\beta l < 5^\circ\), medium \(< 16^\circ\)80 and 250 km overhead at 50 Hz
Cascade convergenceerror \(\propto 1/n^2\)\(n \ge |\gamma l|/\sqrt{6\varepsilon}\)
Section length rule\(|\gamma l/n| \le \sqrt{6\varepsilon}\)14° per section for 1%
Cable velocity\(v = c/\sqrt{\varepsilon_r} \approx 0.63c\)Boundaries shrink to two thirds
Harmonic scaling\(\beta_h = h\beta_1\)Resonance at \(h = 90^\circ/\beta_1l\)
Frequency scalingboundaries \(\propto 1/f\)60 Hz gives 5/6 of the 50 Hz lengths
Study dependenceFerranti \(\to A\); power limit \(\to B\); charging \(\to C\)Decides the usable length
Diagnostics

Common Mistakes

  1. Choosing a model by length alone. Valid only for overhead lines at power frequency — Problem 5 and Challenge C2.

  2. Applying overhead boundaries to a cable. A cable is electrically 50% longer per kilometre — Problem 15.

  3. Using power-frequency constants at a harmonic. \(\beta\) scales with order, and a 400 km line is near resonance at the third — Problems 16 and 17.

  4. Judging a model by its error in \(A\). \(A\) converges fourth-order and flatters the model; \(B\) sets the real boundary — Problem 3.

  5. Treating agreement between two models as confirmation. Not when they share the assumption in question — Problem 8.

  6. Trusting an approximation validated at one operating point. Cancellation of errors is common and does not survive a change of load — Problem 6 and Challenge C1.

  7. Using the short-line model for any no-load study. It predicts zero Ferranti rise at every length — Problem 7.

  8. Confusing the equivalent-\(\pi\) with the nominal-\(\pi\). One is exact, the other a two-term truncation — Problem 18.

  9. Cascading sections that are individually too long. Each section must satisfy the medium-line criterion in its own right — Problem 13.

  10. Ignoring compensation when assessing electrical length. Series and shunt compensation both shorten a line electrically.

  11. Assuming the equivalent-\(\pi\) is exact for everything. It is exact for terminal behaviour at one frequency in the steady state, and nothing else — Challenge C3.

  12. Applying positive-sequence \(\gamma\) to a zero-sequence study. The earth return changes both \(z\) and \(y\), and hence \(\beta\).

Looking Ahead

The three line models have been placed on one axis — the electrical length — and the choice between them reduced to an error estimate against the constant a given study depends on. The equivalent-\(\pi\) removes the choice entirely for computation, leaving the ladder as the route to understanding rather than to arithmetic.

Set 14 takes the phenomena this analysis exposed and treats them as design problems: the Ferranti effect and its compensation, surge impedance loading as an operating target, and the tuned line that Set 12's quarter-wave resonance suggested. Set 15 closes Part 3 with the complex power relations and the receiving-end circle diagram. From Set 16 the line becomes one entry in an admittance matrix and is not examined again.