Abstract: To address the modeling problem of high-dimensional functional time series objects with observation errors and exogenous variables, this paper proposes an analytical framework for a highdimensional functional time series autoregressive model (VFARX) with exogenous variables. Considering potential observation errors, a positive definite operator based on a non-zero lag order autocovariance function is used to reduce the dimensionality of the function. The parameter estimation is transformed into a set of conditional optimization problems through the block regularized minimum distance (block RMD) method, and the theoretical properties of the estimator are given. In high-frequency stock trading scenarios, four VFARX (1, 1) models with different exogenous variables exhibit significant advantages.