Abstract: Coal wall spalling is a major factor restricting the safe and efficient development of coal mining. Accurate monitoring of coal wall support conditions is essential to improving wall stability and ensuring safe mining. To address the issue of low precision, limited range, and poor real-time performance in traditional coal wall deformation monitoring technologies, an intelligent support device based on Fiber Bragg Grating (FBG) was developed. The FBG was innovatively embedded into the hydraulic support shield plate, and a concentrated load inversion model for the coal wall based on Particle Swarm Optimization (PSO)-Radial Basis Function (RBF) neural networks was established to monitor the support conditions of the coal wall. To overcome the limitations of theoretical calculations for stress conditions of structural components with varying cross-sections under complex working conditions, a corresponding relationship between micro-mechanical and macro-mechanical parameters was established using distributed grating strain data based on the PSO-RBF neural network model. This enabled high-precision inversion of the position and magnitude of concentrated load on the support shield plate of the coal wall. Experimental results showed that the model achieved low prediction errors and high goodness of fit in multi-point calibration samples. The average absolute errors of position coordinates and load magnitude in the training set were 0.460 6, 0.248 7, and 0.973 2, respectively. Compared to the load magnitude prediction, the model exhibited smaller errors in predicting the position of concentrated loads. The research findings provide an important theoretical and technical foundation for monitoring coal wall support conditions and ensuring safe mining, as well as a reliable method for accurate measurement of surrounding rock loads on other structural components of hydraulic supports.