A ground system with radials and a perimeter ground that is well-made, and cables that are routed in a way that minimizes tower take-off voltages and coupling, can direct about 90% of the lightning's energy directly to ground. That percentage is somewhat of an upper limit; it is better even than most professional jobs will achieve. In fact, a common assumption in the pro-lightning-arrestor world is that 50% of the surge will go directly to ground. No matter how well the ground system is, it still leaves plenty of energy going down your power cables and into the electronics. Something will have to deal with this so it does no damage, and that is where lightning and surge arrestors come in.
Surge arrestors work by shorting excess voltage from a signal or power carrying conductor to ground. This diverts the surge energy to ground, instead of going through sensitive electronics. If the device is meant to handle the large voltages and currents of a direct lightning strike it goes by the name 'lightning arrestor'. The name 'surge suppressor' is also used, normally in the context of devices that can handle small surges, such as power bars and the like (Advertising will have us believe that all these devices are lightning arrestors, but that is wishful thinking on the advertiser's part). Surge arrestors need a good ground to work properly. In the absence of a good ground a surge arrestor may be able to keep the voltage differentials between the device's conductors within limits, but the device and ground potential as a whole may get raised by several million Volt with respect to ground potential some distance away. This will likely result in a spark to the nearest ground, usually taking an unintended and unexpected route. Bad Things Will Happen™, and very, very likely the magic smoke will be let out of the electronic devices...
There are roughly three types of surge arrestors: The first type employs some type of spark gap. These days the spark gap is usually enclosed by a capsule and some type of gas is employed so it has predictable properties. These devices are often called gas discharge tubes, or GDTs. The good thing about a GTD is that can handle very large currents, in fact the better ones can handle a direct lightning strike. The bad news is that they are relatively slow to respond, allowing a relatively large voltage surge to pass through before kicking in and shorting out the surge. That voltage is generally larger than electronics can handle. The second type of arrestor uses a metal oxide that becomes conductive in the presence of a strong electrical field. These metal oxide varistors (MOVs) act as a resistor that conducts better and better as the current though it increases. The let-through voltage of an MOV is much lower than that of a GTD, but their current handling capacity is smaller and the let-through voltage can still get quite large for large currents. Another bad property of MOVs is that they tend to fail after a number of high-current conduction cycles, and when they fail they do so as a short-circuit. That means they should be fused in some way to avoid overheating or overloading of the circuit that they are protecting. The third type of surge arrestor are semi-conductors such as zener diodes. They have the lowest let-through voltage and the best controlled clamping behaviour of the various arrestor types. Their weak point is that their current handling capabilities are the smallest of all.
Surge arrestors are rated through various parameters. The important ones are their rated voltage (the normal operating voltage of the device, where it does not affect normal operation), their nominal discharge current (the normal current they can handle repeatedly when clamping a surge), their let-through voltage (the maximum voltage left over by the device when it is clamping, usually at their nominal discharge current), and their response time (how long it takes for the arrestor to respond to a surge). The let-through voltage is sometimes described as 'clamp voltage' or 'residual voltage'.
What is still missing from this is the type of surge the device is tested with and rated for. A slow surge of 40 kA causes the device to absorb a great deal more energy than a fast surge with the same current. In North America surge arrestors are almost always rated using an "8/20" waveform. This means the current ramps up to the device's nominal discharge current in 8 μs, and then goes down to 50% in 20 μs. This waveform is a good approximation of an indirect of secondary strike; where the lines are not struck directly. For example, a nearby ground strike, an overhead cloud-to-cloud strike, or a strike half a mile down the road on the power lines would cause this type of waveform. There is a second waveform that is used mostly in Europe and describes a direct lightning strike, a "10/350" waveform. You can figure out for yourself what those numbers mean (and if it is not clear, take look at the figure below).
The curve labelled "1" is typical of a lightning arrestor's capability, while the curve labelled "2" is typical for a surge arrestor. The green (and red) shaded area is a measure of the energy the device needs to absorb, the graph illustrates how different the two devices are. How much current handling is needed? It depends. Say you have a good ground system, wired things up with lightning in mind, and 80% of lightning current goes directly to ground. We know that 95% of lightning strikes have currents of 100 kA or less. Our lightning arrestor will therefore have to deal with a current of up to 20 kA to cover 95% of strikes.
Very much in general, an arrestor that is placed in a location that is expected to see direct lightning strikes should be able to handle at least 12.5 kA per phase tested with a 10/350 waveform, an arrestor that is placed where induced (indirect) surges are expected should be able to handle at least 10 kA many times and at least 25 kA peak current handling, tested with an 8/20 waveform. For grid voltage systems, the residual voltage should be kept below approximately 1 kV. For other lines, it is generally desirable to keep the residual voltage at or below about 4x the normal line voltage. Most electronic devices can handle this without damage. Most arrestors on the North American market are rated with an 8/20 waveform. Very few are available that are rated and tested with a 10/350 waveform, one brand that makes them is Dehn from Germany. In most cases a 10/350 device needs to be combined with the properly coordinated 8/20 follow-up device to make the residual voltage low enough.
Devices rated for 10/350 surges are generally some type of spark gap device, usually employing gas discharge tubes. As mentioned, these devices excel at eating the very high currents from a direct lightning strike, making them good devices for this purpose. Their downside is that they leave a high let-through voltage behind, generally too high for electronics to survive. That means a 10/350 arrestor has to be teamed up with another surge arrestor, an 8/20 rated type, that then treats the left-over so electronics can handle what remains of the surge. The tricky part is that it is not a good idea to put multiple surge arrestors on the same line. What usually happens is that one of the devices (the one that is fastest to respond, or that has the lowest response voltage) ends up handling the entire surge. Usually that is the weaker of the devices, killing it in the process and leaving the electronics vulnerable again. There are several ways around this; the first method is to have enough wire between the 10/350 device and the 8/20 rated arrestor that the two are electrically decoupled. When there is at least 50 feet of wire it will have a high enough impedance for surges that the 8/20 device will work independently of the 10/350 arrestor. If there is not enough distance available to do this, the second method is to employ a surge arrestor that has the needed impedance to decouple the arrestors build into it, in the form of a few coils. These are combination devices that have a 10/350 arrestor and an 8/20 arrestor in one package. Then there are families of 10/350 and 8/20 devices that are designed to work together without any additional decoupling. They work well, though the good ones are not cheap.
It should be clear from the above that there is no single ideal surge arrestor device. That means a combination of devices needs to be employed to provide effective lighting arresting. A divide and conquer strategy. By dividing responsibilities we can also put a little redundancy into the system; when one device fails it does not immediately leave the entire system vulnerable. For a typical wind turbine install this means there should be a lightning arrestor close to the wind turbine alternator, to protect the alternator and slip-rings, and to take as much energy as possible off the turbine wires before they go to the inverter or charge controller (or batteries). Since the turbine is very much exposed to the elements and can get hit by a direct strike this requires a 10/350 rated device. Hopefully the turbine wiring goes underground for some distance from there (50 feet or more), adding enough inductance to the system to decouple from the direct lightning current and turn the left-over surge into something that resembles more of an 8/20 waveform. Therefore, directly before the electronics (rectifier, inverter, charge controller, or batteries) an 8/20 rated surge arrestor is a good choice. The other side of the electronics, the grid side, needs protecting too. If this is not exposed to direct strikes an 8/20 device is all it takes here. For locations with overhead grid lines prone to direct strikes it takes another 10/350 device, and a subsequent 8/20 arrestor down the line (or a combination device) to properly protect this. In case there are any additional wires coming or going to the electronics, such as Internet, phone, or DC from photovoltaics, they will need surge protection too. It only takes a single unprotected line to bring in a surge that will do thousands of dollars in damage.
A good surge arrestor is one that is fast enough to react to a surge, so the surge gets shorted to ground before it rises to the point of doing damage. The pass-through voltage should also be low enough for the application, so the equipment behind it survives what is left over. The surge arrestor should be rated for a sufficiently high current, and waveform, to do the job that it is intended for. Very important, the surge arrestor should be able to handle that current not just once, but many times. The bad news is that many manufacturers of arrestors, even a few respected ones, publish the one-time surge rating for their device, making it seem as if it can handle a great deal of current. While this has its place, we want to make sure the device is rated for multiple surges of the current that it needs to shunt to ground. It does not much good to have a surge arrestor that we believe was rated for the job, but that after a single strike is no longer working. Standards differ, but look for the surge current capacity of the device that it can handle at least several thousand times, and look for the waveform that goes with this current. For example, a very respected manufacturer advertises what is essentially a 10 kA 8/20 device as a 90 kA device, just because it can handle a single 90 kA surge. Do not fall for this trap! A good surge arrestor should make it easy to see when the device is no longer protecting the lines, for example because its surge rating has been exceeded one too many times. This can mean there is a flag on the device, or an indicator light, that shows the device needs to be replaced. Finally, for devices that fail in a state that shorts out the line to ground, it should have a fuse or other disconnect mechanism incorporated in it. To protect the system from overload and possibly overheating.
As described in the wiring section, we want to hook up surge arrestors so there is a 'protected side' and an 'unprotected side'. The unprotected wires carry the large surge currents, including the ground wire. By keeping the protected wires away from the unprotected wires the induced surge voltages and currents are kept to a minimum.
By placing all the renewable energy related electronics in close proximity it is possible to create an 'island of protection': All incoming and outgoing wires are protected by surge arrestors, and all the devices can be grounded by short, direct ground straps to the same ground bar. This will keep any voltage difference between devices due to surges to a minimum, and that will hopefully keep the magic smoke inside. While this can protect the electronics in one spot, it will do little to help protect the rest of the house. It is certainly no luxury to add surge protectors to the various (other) lines that come into house. The service entrance (breaker panel) is a prime candidate, even if you already have a surge arrestor on the grid side of the inverter. Other candidates are the phone, Internet, cable-TV, and the well-pump wires. Especially the latter are often overlooked when it comes to the surge damage potential from a nearby strike. Hopefully the information on these web pages will give you an idea of how lightning does its dirty work, and help you decide for yourself what protection makes sense, and where to place it.
Even with good surge arrestors at key points, there is still a need for (smaller) surge suppressors locally near sensitive equipment. This is because a nearby lightning strike can generate large voltage surges due to inductive coupling into the house wiring. So do not discount that power bar with build-in arrestor just yet. It too has its place.
Once again, to be effective surge arrestors need to be connected with short and straight wires. In case of multiple wires, tie them together (so they run closely in parallel), but keep the ground wires away from the power or signal wires. Every foot of wire from a surge arrestor (or its ground lead) adds roughly 300 Volt to the let-through voltage due to its inductance, for an 8/20 surge. That underlines the importance of keeping the arrestor wires short.