Study On Cigarette Factory Compressed Air Production Forecast Based On Artificial Neural Network

The energy system of a firm is complicated and it includes energy of wide variety. To manage the entire system effectively, many firms have built the energy center which base on computer system, data acquisition and control instruments. Most energy systems can only deal the data with most basic processing, but can not have a comprehensive analysis, optimal scheduling, balance and predict. Scientific prediction is the basis and guarantee of right decisions, and accurate predictions have great significance on the safe and reliable operation of energy systems. Scientific prediction is also the source of data that EMS needed. So seeking of effective energy forecasting methods to improve the prediction accuracy is of great significance.In the paper, the daily volume of compressed air is short-term forecasted. The paper first analyses the factors that affect the yield and analyses the cyclical characteristics of production series, and production series is preprocessed. Then four prediction models have been built to predict the yield. At last, compare the difference between forecast results and built the model with the highest prediction accuracy.The paper uses the BP Artificial Neural Network prediction model, according to24hours,24original data series are built, and uses the neural network to train and predict the24series, and then it shows that the result is better-off. But the BP Artificial Neural Network had some shortcomings, so it should be optimized with Genetic Algorithm and Particle Swarm Optimization, thus build the BP Artificial Neural Network prediction model based on Genetic Algorithm and Particle Swarm Optimization, experiments show that prediction is better than BP neural network. In order to seek higher prediction accuracy, a combination prediction model based on wavelet is built. Use the wavelet analysis to decompose the date series, then establish the combination prediction model according to subsequence, the final prediction sequence is obtained after wavelet reconstruction, this combination model makes the prediction more accuracy, the feasibility and effectiveness of the model is verified, so it can provide a reliable data support for the energy management.