摘要
为准确预测卡钻事故的发生,利用一种基于时间序列的神经网络卡钻预测方法,将时间序列ARMA建模与神经网络非线性建模相结合。选取与卡钻事故相关性较大的参数作为神经网络的输入项,运用现场数据对神经网络进行训练,再利用神经网路的强非线性和自适应学习能力来建立卡钻事故预测模型;通过时间序列对历史数据的挖掘功能,揭示实际钻井过程中对卡钻事故影响较大的各参数的隐含规律,建立时序ARMA模型,求出卡钻时刻钻井相关参数的预测值;将预测值放入神经网络模型进行测试训练,从而达到预测卡钻事故的效果。运用延安地区实际现场数据证实该方法具有精确的卡钻预测能力及较好的泛化能力。
In order to accurately predict the accident of drilling rig sticking,a neural network prediction method based on time series for drilling rig sticking was used to combine the time series ARMA modeling with neural network non-linear mode-ling. First of all,the drilling parameters closely related with sticking accident is selected as input item of a neural network to train the neural network by field data,the strongly nonlinearity and adaptive learning ability of the neural network are used to es-tablish the prediction models of sticking accident,and then through the mining function of the time series to the historical data, the implicit rule of each parameter closely related with the sticking accident in the actual drilling process is revealed time-series ARMA model establish and the predicted values of relative parameters at sticking point are deduced. Finally,the predicted va-lues are put into the neural network model for testing training,so as to achieve the effect of predicting the sticking accident. The predictive capability and generalization ability of the method for drilling rig sticking were confirmed with yan’an area actual field data.
出处
《现代电子技术》
2013年第22期17-19,23,共4页
Modern Electronics Technique
基金
陕西省自然科学基础研究计划:油气田钻井卡钻的预测与诊断技术研究(2010JM8022)
关键词
卡钻
预测
时间序列
ARMA建模
BP神经网络
sticking of drilling rig
prediction
time series
ARMA modeling
BP neural network
