Short-term Heating Load Forecasting And Analysis Of Thermo-electrical Plant Based On Improved Artificial Neutral Network

Central heating heating load management system for prediction is the premise and the basis for economic operation and the rational design of heating systems, and provides an important basis for setting the heating operation parameters, and ensure the quality heating, energy conservation has important significance.In this paper, a heat transfer station in Xinjiang, for example, operating parameters for the regulation of heating in the past are by man manual adjustment of this phenomenon, with the use of BP networks prediction model heating, and on this basis to some historical operating data on the BP neural network learning and training. In this process, given historical performance data bytes are the lack of or deviation from the range of circumstances, using data fitting method the preliminary estimates of non-performing, then we can then adopt the method of Kalman filter on preliminary estimates of the value of bad data are filtered, and then get a better estimate of the data. In addition, the BP algorithm to improve the forecasting results easily fall into local minimum and forecast precision is difficult to raise the issue, the probability of adaptive change in the method of cross-basic genetic algorithm is improved by improving the genetic algorithm neural network BP Initial to optimize the weights, get a group of optimum weights, set the initial weights on the basis of this initial re-use neural network to predict.Verified that using this method to predict the thermal load, the prediction error within 4%, shows that this method can set the operating parameters for the heating to provide a scientific basis and have practical significance.