This insights article is by Dr Josh Macabuag OBE FREng, Co-Founder and CEO, Renew Risk

On 8 August, a 14-megawatt wind turbine at Ørsted’s Greater Changhua 4 offshore wind farm in Taiwan tripped after detecting an abnormal condition, and a fire was then seen on the turbine. Less than two months later, a turbine blade at the same site was damaged. The causes are still under investigation, but they point to a challenge far beyond one project: underwriters are pricing these machines with little loss history.
Offshore wind turbines are becoming larger and at a pace claims data cannot keep up with, and by the time one generation of technology is understood, the industry has often moved on to the next. The Greater Changhua 2b and 4 sites illustrate this issue well, as the first to use Siemens Gamesa’s 14-megawatt SG 14-236 turbine and the first large-scale offshore wind farm in Asia-Pacific built on suction bucket jacket foundations. Any underwriter assessing the project would be pricing a new turbine variant with no operating history of its own.
No loss history to price risk
Insurers price risk by looking at how similar assets have performed in the past and extrapolating forward. The problem is that turbine designs don’t stay in service long enough to build a record insurers can rely on. Offshore developers want to generate as much power as possible from each site and each turbine, which means bigger machines.
The first offshore wind farm, in 1991, used 0.45-megawatt turbines with 17-metre blades. The Changhua turbines have 115-metre blades and a rotor wider than London’s Leadenhall Building is tall, and 20-megawatt machines are already at sea. Each new generation of turbines is developed and built before the last generation has had time to generate meaningful loss data.
Engineers often talk about the bathtub curve: failure rates are high when equipment is first installed, then settle for most of its operational lifespan, then climb back up as it ages. All new turbine models start on the steep slope, and each time the industry introduces a new turbine model, it returns to the start of the curve. A turbine that has been operating for a year tells insurers very little about how it’ll cope with typhoons ten years into its lifespan.
When underwriters can’t clearly assess risk, the market sets the price. With capacity abundant and rates falling, today’s danger is writing untested technology too cheaply.

Fewer, larger turbines concentrate risk
Larger turbines also concentrate risk. Greater Changhua 4 has 583 megawatts of capacity from 42 turbines. Reaching the same capacity with the 3.6-megawatt turbines common at many earlier offshore sites would take around 160 turbines. So one turbine out of action removes nearly four times the share of site capacity.
Repair costs add to this. Offshore, the bill is driven by repair vessels rather than parts: we estimate a minor blade repair costing around $30,000 can become a bill of $500,000 or more once vessel mobilisation is included.
Pricing what history can’t show
Common market practice for assessing the risk of offshore assets uses generic damage curves, with no wave hazard. A wind farm’s exposure is aggregated to a single point, shifted onshore and run through generic damage curves that ignore the impact of waves entirely.
When we compared this common industry method to asset-level modelling for a 1 gigawatt North Sea wind farm, the proxy produced losses up to 90 per cent lower. Keeping the same aggregated exposure offshore produced losses up to 50 per cent higher. The direction of the difference depends on those assumptions, so an underwriter cannot simply adjust for it.
Modelling won’t predict a component fault. But it can stand in for missing claims history when pricing storms and earthquakes. Risk has to be viewed from an engineering standpoint. This means modelling a wind farm as a collection of assets: turbines with their own foundation type, capacity, water depth, and age. This also includes the cables and substations they connect to, each with its own failure modes and repair logistics.
For turbines, these estimate the likelihood of six states of damage, from a remote restart to unrecoverable structural failure. Where claims data doesn’t yet exist, damage can be estimated from analytical data and expert elicitation, using simulation and fragility curves for each asset type.
Greater Changhua 4 wasn’t the first wind farm to have a difficult year, and it won’t be the last. The offshore wind industry will continue building bigger turbines, and insurers will keep trying to underwrite assets that nobody has insured before. Your risk understanding has to move faster than the claims can tell you. Modelling each asset does that, and lets accurate pricing reward the most resilient designs.

Be the first to comment