Abstract: The monitoring of surface settlement in foundation pits plays a crucial role in ensuring the safe construction of foundation pit projects. However, traditional methods for predicting surface settlement in foundation pits have obvious limitations in parameter determination and complex calculations, and it is urgent to establish an efficient and accurate settlement prediction model. This paper utilizes the surrounding surface settlement monitoring data of the foundation pit project of Beijing Zhongguancun Double Carbon Digital Energy Industrial Park in Jiaoxin Village, Shimen Street, Baiyun District, Guangzhou City, and establishes a foundation pit settlement prediction model based on particle swarm optimization BP neural network (PSOBP). And the accuracy was compared and tested with the foundation pit surface settlement prediction model established by using BP neural network and long short-term memory (LSTM) neural network alone. The results show that the bias value of the PSOBP neural network model is reduced by 84.6% and 87.3% respectively compared with the BP neural network model and the LSTM neural network model, and the RMSE value is reduced by 41.2% and 32.0% respectively, significantly improving the prediction accuracy. The PSOBP neural network algorithm demonstrates higher accuracy and effectiveness in the prediction of foundation pit surface settlement, and can provide reliable decision-making basis for the safe construction of foundation pit engineering.