The Study And Application Of Reservoir Parameters’ Difference Inversion Using Time-lapse Seismic Data

After decades of exploration, large structural reservoirs on land have basically been discovered and developed. In recent years, a lot of companies did a secondary exploration to an existing oil field for the purpose of directly characterizing the variation of reservoirs’ properties by comparing the seismic responses obtained in different exploration time periods which could also be called time-lapse seismic data. Seismic inversion is one of the most effective ways to connect seismic responses and reservoir parameters.Currently, most researchers do the inversion of seismic data from different periods separately, and get parameters changing information by analyzing the differences between two inversion results. However, this could be time-consuming. Inversion based on time-lapse seismic data which means acquiring the difference between two-period seismic data first can help to improve computational efficiency and simplify the inversion process.In this paper, I mainly conducted researches on two different inversion methods including AVO (Amplitude versus Offsets) inversion method and FWI (Full Waveform Inversion) method using time-lapse data. I was focused on solving the following problems:1) There’s big uncertainty of the relations between elastic properties and reservoir parameters in field data, and the consistency of time-lapse seismic data is poor, which is unable to provide an accurate time-lapse difference data2) Existing AVO inversion methods perform poor stability, and also cause serious multi-solution problems3) No one has made the use of FWI method to process time-lapse seismic data so far.To address these problems, I first constructed the relationship between elastic propertiesand reservoir parameters by doing wave-field numerical simulations of time-lapse seismic data based on rock physics experiment of cores. Afterwards, I improved the AVO inversion theory and developed FWI algorithm to dealing with time-lapse seismic data, which is successfully providing theoretical basis for field data processing. According to the above work, I have the following achievements:1)I did rock physical experiments on typical cores from Zhao’ao oil field. Results show that porosity plays an important roles on elastic properties. When porosity is over 15%, resolution of seismic data would be advantageous under water driving condition.2)I did numerical simulation and illustrated the influence of reservoir parameters on seismic responses. Moreover, I discussed how the layer’s thickness affects seismic reflection characteristics and quantitatively analyzed seismic response differences aroused by effective pressure and water saturation changes.3)I proposed principles and philosophy of seismic data consistency and completed the consistency processing of non-recurring acquisition-lapse seismic data.4)I took least squares inversion algorithm to estimate the variation of elastic parameters with time-lapse, and studied AVA (Amplitude versus Incidence Angle) sensitivity to elastic parameters during AVO inversion process, based on which, I deduced a method that using AVO difference directly inverses reservoir parameters. The errors of inversion results using AVO difference inversion method would decrease with incidence angle increasing when incidence angle is less than 30 degrees. Otherwise, when incidence angle exceeds 30 degrees, the errors would keep growing. When the difference between approximate PP reflectivity and the real values is more than 7%, errors increase obviously and inversion results are of bad quality.5)I conducted FWI method using Marmousi model. Results show that this method requires good quality of data acquisition, model building and inversion parameters’ settings and it is heavily relying on the initial model. Furthermore, it has low accuracy if the offsets is too long, which would also affect the convergence speed. It has also been demonstrated that increasing the number of frequency groups especially when there is abundant low-frequency data can help to obtain a better inversion result.In this dissertation, I have two main innovations.1)I improved the algorithm of time-lapse AVO difference inversion method and took the algorithm of using time-lapse data estimating parameters’difference into force. Furthermore, I built a nonlinear relation between reflectivity and reservoirs properties including water saturation and effective pressure and constrained the inversion process with Gardner’s equation which makes the result more stable. In addition, I quantitatively analyzed the sensitivity of estimated parameters.2)I successfully conducted complicated model simulation using FWI method in frequency domain, and qualitatively studied the factors affecting the inversion results.