Oil and Gas Production Prediction Based on SVM and Improved Particle Swarm Optimization

Oil and Gas Production Prediction Based on SVM and Improved Particle Swarm Optimization

Jun Peng (School of Electronic Information Engineering, Chongqing University of Science and Technology, China), Yudeng Qiao (Chongqing Chuanyi Analyzer Co., Ltd, China), Shangzhu Jin (School of Electronic Information Engineering, Chongqing University of Science and Technology, China), Dedong Tang (Chongqing University of Science and Technology, China), Lan Ge (Sinopec Chongqing Fuling Shale Gas Exploration and Development Co., Ltd., China), Qinfeng Xia (Sinopec Chongqing Fuling Shale Gas Exploration and Development Co., Ltd., China) and Shaoning Pang (Federation University Australia, Australia)
DOI: 10.4018/978-1-7998-3038-2.ch012
OnDemand PDF Download:
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Cognitive information is widely used in the field of oil and gas, where production forecasts are of great importance to companies. In this chapter, combining support vector machine and improved particle swarm optimization algorithm, a gas field production prediction model is established, and the model is validated by the actual production data of an enterprise over the years. The results show that the model has good convergence, high prediction accuracy, and training speed and can predict its output more accurately. The method adopted in this chapter is the development of cognitive information technology. The authors have reason to believe that with the continuous development of cognitive information technology, it will have a far-reaching impact on social progress.
Chapter Preview
Top

Introduction

In the development and operation of oil and gas fields, the accurate prediction of oil and gas production is one of the important indexes which are necessary to realize oilfield production scheduling, project planning, economic regulation and personnel management. In order to manage, plan and operate more rationally, making the economic budget and staffing arrangements optimization. Scientific and reasonable forecasting methods must be used to forecast the actual oil and gas production (Zhang, 2016). There are some usual prediction methods that are neural network, yield reduction method, gray prediction, Weng’s model method, differential simulation method. There are also scholars using curve fitting methods or complex and time consuming Reservoir simulation method for large data analysis and so on (Pang, 2013). The neural network prediction method is to use the historical data of the production for training, and constantly adjust the weight between the connected neurons to achieve yield prediction (Wang, 2005).The yield reduction method uses the historical output data and output decreasing law equation to realize the production forecast (Chen, 2016; Bai, 2016). Gray prediction theory is to use a small amount of data to do the differential equation and establish the forecast model (Wu, 2013). Weng’s model method utilize the non-linear regression analysis of historical oil and gas production data to obtain the model parameters. Then the oil and gas production can be predicted after the parameters are obtained reasonably (Zhang, 2014). Based on the study of the dynamic time series of oilfield development, the differential simulation method utilizes the differential dynamic simulation principle to predict.

As demonstrated, the functional module of neural network prediction method exist definite limitations and the precision is bad (Fang, 2010). The yield reduction method is applicable only to the yield prediction which accords with the law of decreasing production (Liu, 2009). The gray prediction theory is just suitable for the forecast of the trend of exponential growth (Dan, 2014). As for the non-exponential growth data, the gray scale is larger and as the dispersion degree of data is bigger, the prediction accuracy is getting worse accordingly. The Weng’s model method has some limitations of prediction because of many factors affecting the yield cannot be taken into account. In the prediction of different states, the differential simulation method can not achieve effective yield prediction if the future information is unclear (Fu, 2010).

Complete Chapter List

Search this Book:
Reset