基于免疫神经网络模型的瓦斯浓度智能预测

Forecast of coalmine gas concentration based on the immune neural network model

将免疫算法与神经网络理论相结合,提出免疫神经网络预测模型以预测采煤工作面瓦斯浓度,并对如何处理时间序列的数据模式问题进行研究.引入延迟单元,将原始输入样本转换为具有延迟特征的新样本,采用延迟算子的输出样本施加到网络预测模型,可以获得浓度时段变幅的信息,这对于提高网络对瓦斯扩散过程的拟合精度和预测精度十分有效.结合某矿井瓦斯预报实例,经过与现场实测值相比较,最大预测误差为6.86%,最小预测误差为2.36%,平均误差为4.61%,所建模型精度的拟合值与预测值都与实际数据吻合得较好,各测点的误差值均在许可的范围内.结果表明,基于免疫神经网络的瓦斯浓度预测模型,能够较好地识别采煤工作面瓦斯扩散的演进规律,对瓦斯浓度能进行合理预报,且该方法具有预报时间快、节省费用的特点.

Using immune neural network prediction model to predict gas concentration was put forward by combining the immune algorithm with neural network theory, meanwhile the data mode of how to deal with the time sequence was studied, delay unit was introduced, and the new samples what have characteristics of delay were converted from the original input samples, by adopting the output sample of delay operator applied to network prediction model, information of the concentration changes in the deferent periods were obtained, it was very effective to improve fitting precision and prediction accuracy of network to gas pervasion. Combined with an example of gas prediction in a mine, compared with the measured values in the spot, the largest forecast error was 6. 86%, the least forecast error was 2.36% , the average error was 4.61% , the fitted value and predictive value of the model accuracy were tallied fine with the actual data, the error of measuring points were within the allowable scopes. The result shows that the developing rules of gas pervasion can be identified well and the gas concentration is predicted reasonably by the gas concentration prediction model based on immune neural networks. And the characteristics of this method are faster forecasting and cost-saving.