Well drilling abnormal state intelligent identification method based on multiple working conditions of well drilling

The invention discloses a well drilling abnormal state intelligent identification method based on multiple well drilling working conditions, and the method comprises the following steps: splitting a well drilling working condition into a plurality of well drilling states; automatically cleaning data, and automatically smoothing abnormal values through a filtering algorithm; constructing a drilling working condition-state recognition model, and adjusting the model into a big data driven adaptive threshold value; screening key logging parameters; establishing different drilling working condition-state model libraries; fusing a time sequence attention network; and establishing an exception explanation module. According to the method, on the basis of working condition identification in the whole drilling process and capture of risk characterization by the intelligent model, an abnormal state identification result has interpretability, drilling risks possibly occurring in the future can be effectively predicted, a