Metals are transported within the deep Earth at high pressure and temperature by fluids possessing a complex chemical makeup. Reliable knowledge of metal behaviour in ore-forming fluids is essential for understanding the enrichment, transport and precipitation of metals and the formation of ore deposits: we can only predict how fluids of different compositions, pressures, and temperatures transport metals if reliable thermodynamic properties are available. The development of quantum chemistry approach especially the pseudopotentials within Density Functional Theory (DFT) together with the development of new generation force field have made Molecular Dynamics (MD) simulation a solid tool in studying the metal speciation and thermodynamic properties, and the results are in good agreement with the experiments. In geochemistry, MD can also help with interpreting the experiments in the aspect of providing the knowledge of the basic features of metal complexation, i.e. the bond distances and coordination numbers. In the recent 10 years, the combination of in-situ X-ray Absorption Spectroscopy (XAS) and MD has been proved a solid method to study the metal speciation in geochemical system. Recently, we employed a new automatic workflow that accelerated by machine learning potentials to calculate thermodynamic properties. Within this workflow, the calculation was accelerated by machine learning (ML) potentials from a concurrent learning scheme, within the accuracy of AIMD level, which enables a much longer time scale at lower costs. This will enable the geochemical simulations to a much larger size and longer time scales.