To appear as a paper in ACM E-Energy. We study how an adversary might attack forecasting techniques in power systems and analyze the potential impact on system operations, such as load shedding and increased dispatch costs. These types attack shows that forecasting techniques can be vulnerable to cyber security threats even when the adversary has no information about the system under attack and very little information about the forecasting algorithms themselves (compared to standard data injection attacks which require full knowledge of the system topology). Code