基于动态广义Takagi-Sugeno型模糊关联规则挖掘的火电厂烟气光谱在线定量分析技术研究

Real-time monitoring the flue gas of thermal power plant has great significance for energy saving. Since the current methods could not be suitable for the change of working condition, this project would study the generalized Takagi-Sugeno (T-S) type fuzzy association rules mining algorithm and the consequent expression of the generalized T-S type fuzzy rule is not fixed. The proposed algorithm is performed on the real recorded data to mine the relationships between the spectral data and the concentration of the components of the flue gas, and the online quantitative analysis of flue gas is realized. First, the clustering analysis algorithm for flue gas spectral data is studied, and the spectral data are classified according to the different wording conditions to ensure the generalization of the model. Second, the generalized T-S type fuzzy association rule mining algorithm is studied and the mined results are used for building the spectra online quantitative analysis model of the flue gas to ensure the flexibility and accuracy of the model. Third, the incrementally updating algorithm for the generalized T-S type fuzzy association rule mining is studied and the self-updating of the rules base is implemented to ensure the autonomy and adaptability of the model. The laboratory and field experiments results would verify the effectiveness and practicability of the spectra online quantitative analysis technology based on dynamic generalized T-S type fuzzy association rules mining for flue gas of thermal power plant. This study would expand the scope of association rules mining in theory, and achieve the online real-time prediction for the concentration of each component of the flue gas, which could not only provide a reliable basis for pollution control and combustion optimization, but also open new avenues for the quantitative measurement of the emission of complex industrial systems.

火电厂烟气实时监测对节能减排具有十分重要的意义。针对现有方法难以适应现场工况变化，本项目拟开展规则后件表达式不固定的动态广义Takagi-Sugeno(T-S)型模糊关联规则挖掘算法的研究，从实际数据中找出烟气各成分光谱数据和浓度之间的关系，实现火电厂烟气在线定量分析。研究烟气光谱数据聚类分析算法，将数据按不同工况归类，确保模型的泛化性；研究广义T-S型模糊关联规则挖掘算法，以挖掘结果为本体构建烟气光谱在线定量分析模型，确保模型的灵活性和正确性。研究广义T-S型模糊关联规则增量式算法，实现规则库自我更新，确保模型的自主性和适应性。通过实验室和现场实验，验证基于动态广义T-S型模糊关联规则挖掘的火电厂烟气光谱在线定量分析技术的效果和实用性。本研究将在理论上拓展关联规则挖掘的适用范围，且实现烟气各成分浓度在线预测，为污染治理和燃烧优化提供可靠依据，也为复杂工业系统排放物定量测量开辟新的途径。

火电厂烟气实时监测对节能减排具有十分重要的意义。针对现有方法难以适应现场工况变化，本项目围绕着火电厂烟气组分光谱定量分析技术开展了相关的研究工作。提出了基于核近邻密度改变因子的孤立点检测算法，实现了对烟气光谱数据的清洗；提出了融合约束发现及代表性对象生成的分割聚类算法，完成了对光谱数据的划分；提出了基于随机决策吸引子的粒子群波长选择算法，准确地选择波长信息使后续建立的模型具有更强的预测能力和更好的稳健性；提出了基于非线性Takagi-Sugeno（T-S）型规则的广义的模糊语言值模型，可以根据不同的工况来调整自身结构，很好地解决了光谱数据中的非线性问题；提出基于动态关联分析的模型剪枝技术，降低了模型复杂度，使其具有更好的适应性。此外，还对光谱定量分析方法并行化设计的相关内容进行了研究。通过实验验证了火电厂烟气光谱定量分析技术的效果和实用性。本研究在理论上拓展了关联规则挖掘的适用范围，且实现烟气各成分浓度预测，为污染治理和燃烧优化提供可靠依据，也为复杂工业系统排放物定量测量开辟新的途径。