面向数字孪生的区间二型T-S模糊建模方法研究

Research on IT2TSK Fuzzy Modeling Method for Digital Twin

为提高变负荷工况下火电机组动态过程模型精度以达到数字孪生模型要求,构建串联结构混合模型,采用量子粒子群算法(QPSO)构建全局模型,引入区间二型T-S(IT2TSK)模糊模型实现孪生体子模型及过渡过程的精准演化。针对IT2TSK模糊模型中模糊集和模糊规则较难确定的问题,设计了两步循环迭代法确定前后件参数,保证模糊集参数能够纵向体现不同状态集之间的不确定性,以及模糊规则能够横向实现子模型间的平稳过渡,保证全工况下孪生体模型高度逼近动态过程。结果表明:以空气预热器为例,在变负荷训练集上,较简化定工况模型和一型T-S模糊模型而言,基于IT2TSK模糊模型的参数在线自适应混合模型的输出能够更准确地跟踪实际值,整体精度更高。

In order to improve the accuracy of thermal power unit dynamic process model under variable power conditions to meet the requirements of digital twin model,a series structure hybrid model was constructed.Quantum particle swarm optimization(QPSO)algorithm was used to build global model and interval type-2 T-S fuzzy model(IT2TSK)was introduced to realize the precise evolution of the twin sub-model and the transition process.Aiming at the problem that the fuzzy sets and fuzzy rules are difficult to determine in IT2TSK model,a two-step iterative method was designed to determine the parameters of the front and rear parts,so as to ensure that the parameters of fuzzy set can reflect the uncertainty between different state sets vertically,and the fuzzy rules can realize the smooth transition between sub-models horizontally,which ensured that digital twin body model is highly approximate to the dynamic process under all working conditions.Results show that taking the air preheater as an example,the output of parameter online adaptive hybrid model based on IT2TSK fuzzy model can track the actual value more accurately and the overall precision is higher on the variable power training set,compared with the simplified fixed condition model and type-1 T-S fuzzy model.