Innovation Series: Advanced Science (ISSN 2938-9933)

Research and Analysis of Quantitative Methods for the Feasibility Study and Evaluation of Power Grid Digital Projects Based on Floating Impact Factors

Fengxi Gao, Mingze Sun

State Grid Liaoning Electric Power Company Limited Economic Research Institute, Shenyang, Liaoning, 110015, China

Abstract: In view of the current problem that the impact factor approval relies on expert experience to achieve strong subjectivity and insufficient accuracy, and the manual review mode is difficult to adapt to the surging demand of power grid digital projects, this paper proposes a quantitative method for impact factors that integrate multiple technologies and constructs a digital auxiliary review system. Firstly, the method uses natural language understanding technology to automatically process the historical WBS workload evaluation form, feasibility study report and other project data, and extract key features. Based on expert experience, discriminant strategies and conflict resolution mechanisms (such as hierarchical judgment method and weighted voting method) are formulated, and a fixed impact factor prediction model is constructed to achieve the regularization and objectivity of basic quantification. Secondly, a machine learning model (including stacked integration, gradient lifting tree and other optimization methods) is constructed based on the preprocessed historical data to calculate the first floating influence factor. The second floating influence factor is obtained by using the semantic similarity weighted sum historical influence factor of the large language model. The workload is adjusted by automatic code generation of the generative large model, and the evaluation model is constructed in combination with artificially set factors, and the third floating impact factor is output. Finally, the fixed impact factor is used as the benchmark to set the judgment interval to eliminate the outliers, the weight is determined according to the characteristic correlation degree, and the three floating impact factors are fused to form a digital auxiliary review system. Practice shows that the proposed method can effectively reduce manual intervention and subjective bias, fully mine the value of historical data, improve the evaluation efficiency and quantitative accuracy of impact factors, optimize project resource allocation, and adapt to the review needs of the rapid development of power grid digital projects.

Keywords: Power grid digitization project; Feasibility study; Floating impact factor
 

References