The Research Of Street Light's Time Prediction

Street lights play an important role in public travel, social order and traffic safety. In recent years, the city street lights grow at a rate of 10% to 20% per year. In 2014 the city street lights consume about 248.5 billion kWh of electricity. According to lighting in an average of 12 hours a day, if reduce lighting 1 minute every day, the annual saving about 345 million kWh, equivalent to about 344 thousand tons of carbon dioxide. Most of existing city street lights trigger on time according to the template table. In normal weather often turn off late in the morning and turn on early in the night, resulting in a waste of energy. In bad weather can't automatically control the lights and need staff manual issued light switch command. However, from sending the command to the street lights completely lighting takes about 10~15 minutes. Meanwhile, the sky has become very dark, seriously affecting people's normal life and may even cause traffic accidents.Combined with the existing street light switch control problems, this paper proposes put prediction into street light switch time, gets the light switch time in advance. Not only provides decision- making support for light control, but also can save energy. The paper studies street light switch time characteristics and its influences, and focuses on the analysis of the characteristics of the illumination, proposes weather numerical mapping. Street light system's environment is complex, the sample data influenced by various factors. Compared with the existing common prediction algorithm, because of the small sample properties of the data, using least squares support vector machine modeling. According to the characteristics of the training data, gives preprocessing methods, the model structure, kernel selection, parameter optimization, etc.Through the simulation analysis can be found that the least squares support vector machine's forecasting performance is better than BP in small sample. When the trend of the illumination change is more specific, the forecast will be better. Only according to the sunset time, lowest temperature, highest temperature and weather conditions, the mean absolute error of prediction is within 2 minutes and 15 seconds. Compared to turn on the light at sunset, according to the paper's prediction time, each light can save about 7 minutes and 33 seconds of lighting time every day.