针对矿井涌水量预测问题,提出一种新的既考虑模糊性又考虑随机性的云加权马尔可夫预测模型.其过程为:首先,利用云理论对矿井涌水量状态概念进行云划分,通过X条件云发生器确定训练样本各年份涌水量所属状态；由于云模型具有模糊和随机特性,同一样本属性值通过X条件云发生器时,输出不同的状态,从而形成不同情形下的马尔可夫状态空间,为此,根据各种情形出现的频数,确定各自的权重.然后,应用加权马尔可夫预测法,计算各种情形下预测样本所属状态的概率矩阵,加权求和得到最终的预测概率矩阵,并根据最大隶属度原则,确定预测样本所属状态.最后,以河南鹤壁四矿1982-1999年的矿井涌水量时间序列作为训练样本,采用所建立的云加权马尔可夫预测模型,对2000-2001年的矿井涌水量所属状态进行预测,研究结果表明:2000年和2001年的涌水量预测状态同为状态5,属涌水量较少年份,与其实际状态相一致.%A new model was proposed to predict the mine water inrush considering the randomness and fuzziness called cloud-weighed Markov model. Firstly, cloud model theory was used to classify the state concept of mine water inrush. Then the X-term cloud generation was used to gain the state of every training sample according to great determination method. Due to randomness and fuzziness of cloud model, output state of every training sample may be different, but belongs to no more than two ones, different state spaces of Markov chain were developed for every situation. After a certain number of simulations, different occurrence probabilities of different state spaces were regarded as the weight to calculate the final prediction probabilities. According to the final prediction probabilities, state of sample for prediction was determined by the principle of maximum membership. Finally, the mine water inrush of the 4th Mine in Hebi in the years from 1982 to 1999 was taken as training samples of time series and die water inrush in 2000-2001 was forecasted with the established model. The results show that states of prediction in 2000-2001 are both state S, which coincide with their real states, and the two years belongs to less water inrush years.
展开▼
