The dynamic liquid level (DLL) of an oil well is traditionally measured onsite by using the acoustic method. This method

however

has its limitation in determining real-time dynamic liquid level. A new ensemble soft-sensor approach of DLL based on the multi-source information feature fusion was proposed. The polish rod load and vibration signal in the time domain was transformed into the frequency domain using fast Fourier transform (FFT). The kernel principal component analysis (KPCA) was used to extract the nonlinear feature of the load and vibration spectral signal and the power

casing head pressure and tubing head pressure time signal. The improved fuzzy interactive self-organizing data analysis technique algorithm (ISODATA) and Gaussian process regression (GPR) were used to fuse time/frequency information feature and establish multiple sub-models. Then

the final DLL prediction model was obtained through the ensemble of the sub-models based on the weight factor calculated by optimized-weighted Dempster-Shafer (D-S) theory. The oil field application showed the validity of the proposed method.