In power engineering cost management, data islands are serious, forecasting lags behind, and the whole life cycle cost is difficult to co-ordinate. This paper proposes an intelligent cost accounting model that integrates BIM (Building Information Model), IoT (Internet of Things) and ERP (Enterprise Resource Planning) multi-source data. Construct a layered data fusion framework incorporating attention mechanisms to achieve dynamic weighted integration of heterogeneous data, thereby enhancing data utilization efficiency. Leverage LSTM (Long Short-Term Memory Network) to capture temporal characteristics such as construction processes and market price fluctuations, and combine probabilistic forecasting to output cost confidence intervals, thereby strengthening uncertainty response capabilities. Further integrate the full life cycle net present value (NPV) model to holistically account for implicit costs during both construction and operation phases, shifting the cost perspective from "static calculation" to "dynamic prediction." Employ particle swarm optimization (PSO) for global optimization of model hyperparameters to enhance prediction accuracy. Taking two actual power transmission and transformation projects as examples, the results show that the deviation rate of the total cost prediction of the model is reduced to 1.8%, which is significantly better than the traditional static model (8.5%) and the single ERP-LSTM model (3.4%), and it shows good real-time response and interpretability in the monthly cost dynamic prediction. The research provides a feasible intelligent cost control technology path for the digital transformation of power engineering.
Zhang et al. (Sun,) studied this question.