Load consumption forecasting plays a crucial role for power systems operators to ensure reliable and efficient operation. Traditional load forecasting methods, such as time series analysis and regression, have been shown to be inadequate for capturing the complex dynamics of load consumption. In this paper, we propose a novel approach for short-|term load forecasting by hybridizying a state-ANFIS (SANFIS) model with particle swarm optimization algorithm (PSO). The hybrid model set up contains 5 layers: fuzzyfication layer, rule layer, normalization layer, consequence layer and output layer for a total of 120 parameters to be determined during training using the PSO algortihm instead of gradient descent. Multivariate time series data from the city of Johor in Malaysia generated in 2009 and 2010, which include hourly load data and hourly temperature time series were used to feed in the SANFIS model in order to determine using the PSO algorithm good values of parameters that minimize forecasting errors made by the models. We evaluated the proposed approach on an unsed part of load consumption of the city of Johor in Malaysia dataset. The results show that the proposed approach can significantly improve the accuracy of load forecasting by obtaining a root mean squared error of 0.000648 and a mean absolute percentage error of 0.043 while using the whole dataset.