DOI: 10.21276/sjet

Pages: 407-412

Saving energy is becoming a trend in today's society, office as our research object, the power consumption is very huge. We can effectively predict and track the office electricity situation is more and more important. In this paper, we play the data advantage, use the neural network to do training, get the corresponding forecast value. In order to overcome the shortcomings of the slow convergence of the neural network and easy to fall into the local error, we combine the genetic algorithm to filter the weights and thresholds of the input layer and the hidden layer to speed up the training accuracy and efficiency. Through the actual data validation, we get the training error within 5.5%, the robustness is strong, and the genetic algorithm after the combination of time increased by 31%. Effectively proving the reliability of our model.