基于粗糙径向基神经网络的刮板输送机负载预测方法研究

Load prediction method of scraper conveyor based on rough RBF neural network

刮板输送机负载的准确预测对实现采煤机和刮板输送机的协同控制至关重要。刮板输送机短期负载受工作面环境、冲击载荷等不确定性因素的影响,具有很强的非线性和非平稳性,难以准确预测。针对此问题,本研究提出一种基于粗糙径向基神经网络的刮板输送机负载预测方法。该方法首先建立刮板输送机电流去噪模型,得到反映综采工作面刮板输送机真实负载的电流分量;然后针对刮板输送机负载电流波动大导致的神经网络预测模型训练误差增大、预测精度低的问题,引入表征负载变化波动的上下输入粗糙神经元,提出一种粗糙径向基神经网络(RRBFNN)模型;最后基于粗糙径向基神经网络建立刮板输送机短期负载预测模型,并进行仿真实验验证。结果表明:本研究提出的RRBFNN刮板输送机短期负载预测模型,比传统RBF模型的平均绝对误差(MAE)、平均绝对百分误差(MAPE)和均方根误差(RMSE)分别降低26.22%,25.39%和14.72%,该方法能有效提高刮板输送机负载的预测精度。

Accurate prediction of scraper conveyor load is crucial to the cooperative control of shearer and scraper conveyor.The short-term load of scraper conveyor is influenced by uncertain factors such as working face environment and impact load,which is difficult to predict accurately,with the strong nonlinear and non-stationary properties.To solve this problem,a load forecasting method of scraper conveyor based on rough radial basis function neural network was proposed.Firstly,the current denoising model of scraper conveyor was established,and the current component reflecting the real load of scraper conveyor in fully mechanized mining face was obtained.Then,in view of the the increasing training error of neural network prediction model and the low prediction accuracy caused by the large fluctuation of load current of scraper conveyor,a rough radial basis function neural network(RRBFNN)model was proposed by introducing rough neurons representing the fluctuation of load change.Finally,based on RRBFNN,the short-term load forecasting model of scraper conveyor was established and verified by simulation experiments.The results show that the RRBFNN forecasting model of scraper conveyor short-term load is better compared with the conventional RBF model,as the average absolute error(MAE),average absolute percentage error(MAPE)and root mean square error(RMSE)was 26.22%,25.39%and 14.72%lower,which indicates the proposed method can effectively improve the forecasting accuracy of scraper conveyor load.