Abstract: Artificial neural network is a technique that can analyze and control the inherent regularity of corresponding input and output data pre-available, and obtain the estimating output results from the new input data. On the basis of this technique, this paper presents the prediction model and method for deliverability of reservoirs following hydraulic fracturing, mainly studies the relative five parameters to fracturing modification of reservoirs in normal conditions, such as fracture length, fracture conductivity, fracture height, extent of fracture communicating with aquifer, degree of damage of fracturing fluid on formation. Also, based on the comparison of more accurate prediction result of reservoir productivity with current program and fracturing effect, the reservoir modification prospect evaluation technique and corresponding software package are developed and widely used in explorative blocks in the northwestern margin of Junggar basin. The overall coincidence rate has reached 83.33%.

Key words: Junggar basin, neural network, reservoir modification, post-fracturing analysis, response evaluation, comprehensive evaluation