Protect Circuit Breakers from Close-In Faults

May 1, 2008 12:00 PM
By J.R. Ribeiro, Florida Power & Light Co.

A Close-In Short Circuit Can Cause Serious Out-Rush Currents from a station capacitor bank. Whether the short circuit is at the bus or at any line emanating from a station, the resulting out-rush currents have high-frequency and high-magnitude components, depending on the size of the bank.

An out-rush reactor, connected in series with the capacitor bank, has been a common method used to protect substation general-purpose breakers from failure caused by attempting to switch excessive fault currents. Without the out-rush reactor, out-rush current from a close-in fault can significantly exceed the general-purpose breaker capacity for the current-times-frequency (I×f) product, with potentially disastrous results.

THE PROBLEM WITH OUT-RUSH REACTORS

Typically, an out-rush reactor is specified to bring the current magnitude and frequency within the general-purpose substation breakers' capability. The device itself is not expensive, but the civil works to accommodate it can be a significant part of a capacitor-bank installation. The photo below shows an out-rush reactor installation associated with a 110-MVAR, 230-kV project.

The problem is that a short circuit between the capacitor bank and the out-rush reactor can effectively ground one end of the out-rush reactor. Breakers face significant stress when attempting to interrupt the short-circuit current associated with the small inductance of an out-rush reactor.

The process of interrupting the current through the out-rush reactor is highly complex, but fortunately can be understood in simple terms. Fundamentally, the successful interruption of a fault involves the dielectric strength of the breaker contacts being stronger than the voltage (stress) across the breaker contacts. Otherwise, the distance between the breaker contacts is arced over and no interruption occurs. Because the breaker contacts are in motion, this strength-versus-stress process is dynamic.

As the breaker attempts to interrupt the current, a race starts between the change in dielectric strength between the breaker contacts, or the rate of rise of dielectric strength (RRDS), and the increase in voltage difference between the breaker contacts, or the rate of rise of the recovery voltage (RRRV). If the RRRV is greater than the RRDS, the breaker may fail to interrupt the current.

Some investigators believe this is what happened in an incident reported by The Toronto Star on Jan. 31, 2007, where a fault at a capacitor bank resulted in a station fire and the “temporary interruption of about 1500 MW of capacity.” Breaker failures, particularly of oil breakers, can be quite damaging and dramatic events.

ANALYTICAL CONSIDERATIONS

The RRRV is largely a function of the capacitance to ground at the reactor side of the breaker and the inductance of the out-rush reactor. The smaller the capacitance to ground, the higher the RRRV.

In the case where there is no intentional additional capacitance at the location, a small stray capacitance to ground exists of the order of 400 pF for a typical 230-kV installation. This capacitance forms a resonant circuit with the out-rush reactor developing a high-frequency voltage on the reactor side of the breaker. The resulting RRRV can be several times higher than the capability of a standard special-purpose breaker. To bring the RRRV to within the breaker capability, the typical solution is the installation of fixed capacitors of nanoFarad size connected from the high side of the out-rush reactor to ground. For reference, the capacitance of a 110-MVAR, 230-kV bank is 5.5 µF.

Because no out-rush reactor is installed in series with the fixed capacitors, and the I×f product happens to be primarily a function of the inductance between the point of short circuit and the fixed capacitors, we are right back at the starting situation — a capacitor bank without an out-rush reactor. The I×f primary dependence on inductance can be readily seen from the defining equation:

This equation is a direct derivation from the expressions for peak current and frequency shown in IEEE Standard C37.012-1979. In this equation, V_LL is the bus nominal line-to-line voltage in volts, n is the number of capacitor banks in parallel, and L_EQ is the intrinsic inductance of the bank and the series combination of the inductances from the capacitor bank and the line and/or bus work to the point of the short circuit.

We can use the equation to calculate the I×f product for a short circuit at the bus 50 ft (15 m) away from a capacitor bank. Assuming 10 µH for the capacitor inherent inductance, 0.285-µH per ft of bus for a typical 230-kV substation and operating the bus at 245 kV, we have an I×f product of 1312-kA kHz. The I×f product capability of general-purpose breakers is 20-kA kHz, according to C37.06-2000, and 110-kA kHz, according to the draft of C37.06-2007.

In summary, adding the out-rush reactor to mitigate the high I×f product of the capacitor bank requires the addition of fixed capacitors to mitigate the high RRRV. However, the fixed capacitors generate similar I×f products as the original situation, and these are in excess of the capability of the general-purpose breakers.

RISK ANALYSIS

The most relevant parameter of the I×f product is the equivalent inductance between the capacitor bank and the point of fault. Assuming 0.6 mile (1 km) of reactance for the transmission line, the I×f product will be within the general-purpose breakers' capability on a typical 230-kV substation, provided that the short circuit is farther than 0.16 miles (0.26 km) from the substation and the total installed capacitance is less than about 660 MVAR. This is more than most existing installations, especially for Florida Power & Light Co. (FPL; Miami, Florida, U.S.), where the largest installation totals 440 MVAR.

The risk occurs only when attempting to reclose the general-purpose breaker on a permanent fault, which is one that remains on the system after the breaker opens for the first time. The risk does not occur at the first attempt to clear a fault, because at the moment of clearing, typically about three cycles after fault inception, the out-rush transient has already abated.

For a 660-MVAR installation, the breaker peak current rating is more limiting than the I×f, requiring the minimum distance from the substation to be about 0.29 miles (0.47 km). The probability of a fault closer than 0.29 miles from a substation is very small.

For 19 years, the FPL Outage Data Bank has covered about 104 substations (230 kV and 500 kV) and roughly 5600 miles (9012 km) of lines. It shows only 18 faults within 0.29 miles of a substation. The only permanent fault was on a line emanating from a generation station, where the policy is always to block automatic reclosing, so FPL did not attempt to reclose it. These statistics help explain why there have been so few reports of breaker failures caused by capacitive current out rush, even though FPL has some capacitor bank installations without out-rush reactors.

A SIMPLE SOLUTION

With the facts noted previously, and the consideration that if the I×f product limitation for breakers specified in Standard C37.06 is correct, then in the event of close-in faults, the out-rush currents not only from capacitor banks but also from CCVTs, breaker-bushing capacitors or stray-bus capacitance exceed the capability of general-purpose breakers.

Therefore, the simplest solution is twofold: eliminate out-rush reactors to remove the risk of excessive RRRV and block automatic reclosing on faults close to the substation. This solution is reliable, robust and relatively easy to implement.

It would be prudent to suggest setting relays for 0.5 miles (0.8 km) or even 1 mile (1.6 km). The table on this page shows the number of faults within 0.29 miles, 0.5 miles and 1 mile along with the number of faults over the 19-year period. The farther from the substation the relay is set to block reclosing, the safer it is, in the sense that it will block reclosing for more faults. The tradeoff is that each blocked reclosing event will require patrolling of the line before re-energization. To assess the impact of additional patrolling, the table also shows that preventing automatic reclosing for faults closer than 0.5 miles results in an average of 3.2 more such patrols a year. This should not break the operations budget.

Except for possible replacement of some obsolete relays, this solution has negligible cost. It is important to note that the fault-location function cannot rely on current magnitudes because these can vary. It is necessary that the fault location be determined by modern relays capable of accurate and robust fault-location identification.

J. “Joe” R. Ribeiro is staff engineer in the Transmission Services and Planning department of Florida Power & Light Co. (FPL). He joined FPL in 1984 after seven years with Niagara Mohawk (Syracuse, New York, U.S.), three with PTI (Schenectady, New York) and three with American Electric Power Service Corp. (New York, New York), where he started his career as a power-system planner back in 1970. Ribeiro holds a BSEE degree from New York University and an MSEE degree from Union College in Schenectady. He is a professional engineer in the states of Florida and New York.
J_R_Ribeiro@FPL.com