Liang Haibo

Neural network prediction model to achieve intelligent control of unbalanced drilling’s underpressure value

In underbalanced drilling, accidents like well leakage and overflow not only damage the reservoir but cause great safety risks to drilling operations. Therefore, it is of great engineering significance to maintain a reasonable under-pressure state by controlling a reasonable underpressure value. Data mining is an advanced method for retrieving and creating corresponding models in massive data. During the drilling process, there are a large number of real-time monitoring data and historical data. Therefore, a neural network prediction control model based on improved rolling optimization algorithm has been proposed. Combined with control principle of underpressure value, a set of online rolling optimization neural network control model for achieving underpressure value intelligent control of underbalanced drilling is formed. The control model optimizes the neural network prediction control model through the rolling optimization algorithm, realizing advanced prediction of reasonable underpressure value, and performs fast and stable self-feedback control of the output prediction results. By using field data for optimization analysis, the analysis results show that using the neural network prediction control model of online rolling optimization can effectively conduct accurate prediction and real-time control for the reasonable underpressure value.

Application of an intelligent early-warning method based on DBSCAN clustering for drilling overflow accident

Oil and gas are still the necessities of production in today’s society. However, the exploration and mining of them are extremely complex and dangerous. Overflow accidents are undoubtedly one of the biggest threats to safe drilling operations during the oil and gas exploration. Due to the complexity of geological information or lack of adjacent well data in drilling process, the problem of overflow warning model based on sample information can not be established. Data mining is the process of revealing meaningful new patterns, relationships and trends by analyzing data, therefore, based on the correlation between the occurrence of overflow accidents and the change trend of casing pressure, a method of intelligent warning based on improved DBSCAN clustering method for drilling overflow accidents is proposed. The early warning method uses time-series scanning and stratification to rule the idea of clustering, not only improve the efficiency of clustering, but also enhance the clustering effect. According to the results of clustering fitting and the sensitivity of overflow accident, output the warning result of overflow accident. The data analysis is made by using the field data. The experimental results show that the flood warning method based on improved DBSCAN clustering can effectively predict the overflow accidents.