基于深度网络的钢铁工业能源系统隐性调度知识提取与可释研究及应用

The big data of energy system of iron and steel industry contains tacit scheduling knowledge. Effectively extracting and explaining the tacit scheduling knowledge and applying them to practical production scheduling process will promote the level of scheduling intelligence and automation. Facing this practical demand, this project will study the theory and method of extracting and explaining of the tacit scheduling knowledge based on deep network, and apply them to practical production scheduling process of large-scale iron and steel enterprise in China. In the theory study field, extracting methods of the tacit scheduling knowledge contained in big data of energy system of iron and steel industry will be mainly studied. The new theory of extracting tacit scheduling knowledge based on deep network will be first proposed. After that the graph expression and explaining patterns of the tacit scheduling knowledge based on knowledge scheme will be proposed. In the application study field, using the proposed extracting methods and explaining patterns of the tacit scheduling knowledge builds scheduling knowledge platform of energy system of iron and steel industry. The developed scheduling scheme based on the tacit scheduling knowledge is applied to energy system of iron and steel industry as a demonstration, and therefore make a contribution for promoting the level of intelligence and automation of energy system scheduling of iron and steel industry of China and decreasing energy consumption of iron and steel production process.

钢铁工业能源系统大数据中蕴含着隐性调度知识，有效地提取和解释隐性调度知识，并应用到实际生产调度过程将会提升调度的智能化和自动化水平。本项目面对这一实际需求，开展基于深度网络的钢铁工业能源系统隐性调度知识提取与可释的理论及方法研究，并在我国大型钢铁企业进行示范应用。在理论研究层面，主要研究钢铁工业能源数据中隐含的隐性调度知识的提取方法，形成基于深度网络的隐性调度知识提取的新理论和基于知识图的隐性调度知识的图形化表达和可释方式。在应用研究层面，使用提出的基于深度网络的隐性调度知识提取与可释的方法建立钢铁工业能源系统调度知识平台，并将产生的基于隐性调度知识的调度方案在典型钢铁工业能源系统进行示范应用，从而为提升我国钢铁工业能源系统调度的智能化和自动化水平，为钢铁生产过程的节能降耗作贡献。

钢铁工业能源系统大数据中蕴含着隐性调度知识，有效地提取和解释隐性调度知识，并应用到实际生产调度过程将会提升调度的智能化和自动化水平。本项目首先针对钢铁工业能源系统调度知识的提取问题，提出一种基于分层粒度对比网络的知识建模方法，为后续知识处理更新提供基础；随后针对钢铁工业能源系统调度知识的更新问题，提出一种基于迁移学习的调度知识泛化方法；最后，研究钢铁工业能源系统调度知识的关系发现与解释等问题，提出基于粒度因果关系网络的钢铁能源调度知识关系可释化方法。本项目研究成果可为钢铁企业节能减排提供帮助，也对相关领域智能化水平提升大有裨益。