It depends. Had a similar challenge using "SCADA-type" data to extend the performance of a network of systems used to test a fuel system. All of your SCADA data will only be useful, if they can be used to synthesize the optimum operating conditions JUST-IN-TIME for your controller (?) that oversees the wind turbine operation. So the first question is whether the optimization is in real-time or over a period of time (i.e. optimize offline, then deploy reconfiguration later)? If the former, then understanding the time that computation must be performed and it's output fedback to the controller is critical. In particular getting all the data you need at half the time the optimization params need to be fedback. Otherwise, performance will lag considerably.