An estimation method incorporating a normal-inverse Wishart conjugate prior distribution was proposed in conjunction with Bayesian theory to address the problem of high-dimensional estimation of traditional vector autoregressive models VARs . In this estimation method the proposed model introduced the Metropolis-Hastings MH algorithm to determine the hyperparameters of the prior distribution from the previous dataset and to estimate them by setting the shrinkage coefficients associated with the model dimensions. Compared with traditional VAR models the estimation method based on Bayesian theory can control overfitting while retaining relevant sample information and has better robustness and effectiveness. In addition based on the improved BVAR model the MS-BVECM model was proposed by combining the zone system transfer technique with the error correction model which can effectively analyze the long-run and short-run equilibrium state changes between the variables during the economic cycle. When the economic variables are subject to fluctuations in the short run and deviate from the long-run equilibrium the error correction model mechanism will gradually bring them back to the long-run equilibrium to ensure the robustness of the model. Finally taking Chongqing as an example the proposed model was used to analyze and predict the dynamic relationship between energy consumption industrial structure upgrading and economic growth in Chongqing and to provide feasible suggestions.