The invention relates to a building instantaneous cold load prediction method based on a Monte Carlo method using Latin hypercube sampling. The method comprises: using a Monte Carlo simulation methodto calculate a building cold load, and using the Latin hypercube sampling method to perform random sampling of the Monte Carlo simulation method; the Monte Carlo simulation method being a numerical calculation method guided by a probability statistics theory, performing random sampling on all uncertain parameters forming the building cooling load, recombining samples of all the uncertain parameters, and obtaining probability distribution of the building instantaneous cooling load after multiple times of Monte Carlo simulation. In the statistical sampling process, the Latin hypercube is the popularization of a Latin square matrix in multiple dimensions, each hyperplane perpendicular to the axis contains at most one sample, and therefore the Monte Carlo simulation frequency is reduced whilethe randomness of the samples is guaranteed. According to the method, the building instantaneous cooling load can be effectively calculated and predicted, and a reference basis is provided for installation and model selection of the building air conditioning system and operation optimization of the air conditioning system.