Abstract: In order to establish the mapping relationship between the hydraulic support column load and its dynamic response characteristics, and to realize dynamic prediction of the column load, a hydraulic support column load prediction method based on extended response characteristic parameters was proposed. The high-frequency switching of the sliding mode surface of the sliding mode differentiator was used to filter out the measurement noise from the displacement and pressure signals of the column. Through derivative extraction, the dynamic response parameters were extended into five response characteristic parameters: displacement, velocity, acceleration, pressure, and pressure change rate. These five parameters were input into a BP neural network-based hydraulic support column load prediction model to predict the column load. An AMESim–Simulink co-simulation model of the hydraulic support column was built to analyze the dynamic response characteristics of the column under different levels of impact dynamic loads, the correspondence between the column load and its extended characteristic parameters, and the performance of the BP neural network prediction model based on extended multi-parameters. The simulation results showed that the estimation accuracy of the sliding mode differentiator for the measured displacement and pressure signals of the column was 98.73% and 92.95%, respectively. In five load prediction tests, the BP neural network prediction model based on extended multi-parameters had an average prediction accuracy of 97.2% and an average prediction error of 34.39 kN, which was an improvement of 31.97% over the BP neural network prediction model based on measured signals.