To address the issue that limited edge resources of the existing Edge Server Providers （ESPs） reduce the Quality of Service （QoS）of hierarchical federated learning edge nodes， a dynamic Edge Federated Framework （EFF）was proposed by considering the potential edge federation probability among edge servers. In the proposed framework， different ESPs cooperated to provide additional edge resources for hierarchical federated learning， which suffered from reduced model training efficiency due to client heterogeneity or Non-Independent and Identically Distributed （Non-IID）data. Firstly， decisions were offloaded by quantifying the communication model， and offloading tasks were assigned to the edge servers of other ESPs within the framework， so as to meet the elastic demand of edge resources. Secondly， the Multi-round Iterative EFF Participation Strategy （MIEPS）algorithm was used to solve the evolutionary game equilibrium solution among ESPs， thereby finding an appropriate resource allocation strategy. Finally， the existence， uniqueness， and stability of the equilibrium point were validated through theoretical and simulation experiments. Experimental results show that compared to non-federation and pairwise federation strategies， the tripartite EFF constructed using MIEPS algorithm improves the prediction accuracy of the global model trained on Independent and Identically Distributed （IID） datasets by 1.5 and 1.0 percentage points， respectively， and the prediction accuracy based on Non-IID datasets by 2.1 and 0.7 percentage points， respectively. Additionally， by changing the resource allocation method of ESP， it is validated that EFF can distribute the rewards of ESP fairly， encouraging more ESPs to participate and forming a positive cooperation environment.

To solve the problem of poor performance in segmenting a large number of proprietary words in Chinese text in electric power domain， an improved Chinese Word Segmentation （CWS） method in electric power domain based on improved BERT （Bidirectional Encoder Representations from Transformer） was proposed. Firstly， two lexicons were built covering general words and domain words respectively， and a dual-lexicon matching and integration mechanism was designed to directly integrate the word features into BERT model， enabling more effective utilization of external knowledge by the model. Then， DEEPNORM method was introduced to improve the model’s ability to extract features， and the optimal depth of the model was determined by Bayesian Information Criterion （BIC）， which made BERT model stable up to 40 layers. Finally， the classical self-attention layer in BERT model was replaced by the ProbSparse self-attention layer， and the best value of sampling factor was determined by using Particle Swarm Optimization （PSO） algorithm to reduce the model complexity while ensuring the model performance. The test of word segmentation was carried out on a hand-labeled patent text dataset in electric power domain. Experimental results show that the proposed method achieves the F1 score of 92.87%， which is 14.70， 9.89 and 3.60 percentage points higher than those of the methods to be compared such as Hidden Markov Model （HMM）， multi-standard word segmentation model METASEG （pre-training model with META learning for Chinese word SEGmentation） and Lexicon Enhanced BERT （LEBERT） model， verifying that the proposed method effectively improves the quality of Chinese text word segmentation in electric power domain.