基于模糊网络和粒子群优化的油田指标预测

Forecast of Oilfield Indexes Based on Fuzzy Neural Networks and Particle Swarm Optimization

针对油田开发指标预测问题,提出一种模糊神经网络模型,该模型包括输入层、模糊化层、规则层和输出层。模糊化层采用高斯隶属函数,规则层每个节点对应一条模糊逻辑规则。网络可调参数为模糊集参数和输出层权值。提出了基于改进量子粒子群优化的网络训练方法。以油田开发指标中含水率预测为例,结果表明该方法是有效的可行的。

Aiming at the forecast of oilfield development indexes, a fuzzy neural networks model is proposed that includes input layer, fuzzification layer, rules layer, and output layer. The Gauss function is applied in fuzzification layer, and each node in rules layer corresponds to a fuzzy logic rule. The adjustable parameters of proposed model include the fuzzy set parameters and the weight value of output layer. For determining these parameters, an improved quantum particle swarm optimization is presented. With forecast of moisture content as an example, the experimental results show that this method is effective and feasible,