A method for generating an abnormal state detection model, that includes: receiving first and second normal state signals detected, for a predetermined time by first and second sensors, respectively; generating a normal state data set based on the first and the second normal state signals; recognizing first and second variable values; generating a first abnormal state signal by applying the first variable value to an impulse function and a second abnormal state signal by applying the second variable value to a Gaussian distribution function; generating an abnormal state data set based on the first and second abnormal state signals; generating a detection model learning set for generating an abnormal state detection model by combining the normal state data set and the abnormal state data set; and generating an abnormal state detection model based on the detection model learning set.