Reduction of non-linear many objectives for coordinated operation of integrated energy systems

Authors：Qin, Y. J.; Zheng, J. H.^*; Li, Zhigang; Wu, Q. H.

Source：INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS, 2020, 117: 105657.

DOI：10.1016/j.ijepes.2019.105657

• Reduction of non-linear many objectives
• Principal component analysis
• Maximum variance unfolding
• Kernel matrix
• Coordinated operation
• Integrated energy systems

This paper establishes a model that considers different types of indices for the coordinated operation of integrated energy systems (IES), such as economic indices, environmental indices, security and stability indices. A many-objective optimization problem (MaOP) is developed to treat each index independently as objectives to participate in the optimization rather than constraints. In order to reduce the dimensionalities of objectives and improve the computing efficiency, a kernel matrix based principal component analysis (KM-PCA) method is proposed to achieve the reduction of non-linear many objectives. In the KM-PCA method, the initial correlation matrix of principal component analysis (PCA) is transformed to a kernel matrix by using the maximum variance unfolding (MVU) method. The KM-PCA aims to generate a smaller set of conflicting objectives with maximum information retention of the original objectives. Simulations studies are conducted on a test IES consisting of a modified IEEE 30-bus electricity network, a 15-node gas network and three distributed district heating and cooling units (DHCs) to verify the effectiveness of the KM-PCA method, and a final solution is selected for the coordinated operation of IES.