历史数据下优化GRU模型的测控设备相位预测

Improved GRU Model for Phase Prediction of Measurement and Control Equipment Based on Historical Data

中文关键词:  环境受限  历史记录  相位预测  鲸鱼算法  GRU

英文关键词:environmental constraints  historical records  phase prediction  whale algorithm  GRU

中文摘要:

      航天测控设备目标跟踪测量对相位具有依赖性，在无近远场标校环境或场坪周边环境受限等情况下，人工经验装填相位难以满足参数准确性和设备稳定跟踪需求。探索依赖记录的历史数据进行相位预测方法，通过数据筛选、清洗和标准化等预处理后，利用群智能鲸鱼算法，以预测值和真实值的均方根误差作为适应度函数，自适应寻优长短时序列模型神经元个数和步长两个超参数，得到超参数优化值，对历史数据随机划分进行训练和验证，实现最终相位预测模型，对某型号设备历史数据和当前校相条件开展多次验证，同时与循环神经网络（RNN）和门控循环单元（GRU）对比，模型预测值和真实值相差小，算法准确性优于传统统计计算，满足测控设备跟踪、日常比对以及应急辅助需求。

英文摘要:

      Target tracking measurement of aerospace measurement and control equipment is dependent on phase,and it is difficult to meet the requirements of parameter accuracy and equipment stability tracking by relying on manual experience when there is no near and far field calibration environment or the surrounding environment of equipment field is limited.The phase prediction method relying on the recorded historical data is explored.After preprocessing such as data screening,cleaning and standardization,the swarm intelligence whale algorithm is used to adaptively optimize the number of neural elements and the step size of the long and short time series model by taking the root mean square error of the predicted value and the real value as the fitness function to obtain the optimized value of the super parameter.The random division of the historical data is trained and verified to achieve the final phase prediction model.The historical data and current phase calibration conditions of a certain type of equipment are verified for many times.At the same time,the recurrent neural network (RNN) and gated recurrent unit (GRU) are compared,the difference between the predicted value of the model and the true value is small,and the accuracy of the algorithm is better than the traditional statistical calculation,meeting the tracking,daily comparison and emergency auxiliary needs of the measurement and control equipment.