The invention relates to a feature cross elevator trapping time sequence prediction model construction method based on deep learning, an obtained model and a prediction method. The construction method comprises the steps of forming a time sequence by historical information data, transmitting input information into a single-layer full-connection layer and converting the input information into primary abstract features, performing feature interaction, transmitting the input information into an LSTM neural network and obtaining comprehensive time sequence abstract features, transmitting the input information into the full-connection layer and converting the comprehensive time sequence abstract features into predicted trapped person probability, and calculating predicted loss by a cross entropy loss function. And LSTM neural network parameters are updated through back propagation to train the model. The prediction method comprises the steps of obtaining a plurality of prediction models by adopting a construction method, performing comprehensive processing by utilizing the plurality of prediction models with slightly different emphasis, and obtaining a comprehensive trapped person probability as a prediction result. According to the method, on the premise that real-time operation data of elevator assemblies are lacked, only historical record data of the elevator are used, the time sequence characteristics of the elevator are extracted, and the probability that people are trapped in the elevator at the next time point is predicted.

    本发明涉及基于深度学习的特征交叉电梯困人时间序列预测模型构建方法、所得模型及预测方法。该构建方法包括将历史信息数据组成时间序列,将输入信息传入单层全连接层转换为初级抽象特征,特征交互,传入LSTM神经网络并获得综合时序抽象特征,传入全连接层并转换为预测困人概率,以交叉熵损失函数计算预测损失,并通过反向传播更新LSTM神经网络参数以训练模型。该预测方法包括采用构建方法获得多个预测模型,利用侧重点略有差别的多个预测模型进行综合处理并得到综合困人概率作为预测结果。本发明能在缺少电梯组件实时运行数据的前提下,仅使用电梯的历史记录数据,抽取电梯的时序特征,预测电梯在下一个时间点的困人概率。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Deep learning-based feature cross elevator people trapping time sequence prediction model construction method, obtained model and prediction method


    Weitere Titelangaben:

    基于深度学习的特征交叉电梯困人时间序列预测模型构建方法、所得模型及预测方法


    Beteiligte:
    CHEN WU (Autor:in) / XU WEIQUAN (Autor:in)

    Erscheinungsdatum :

    2022-06-14


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06K Erkennen von Daten , RECOGNITION OF DATA / B66B Aufzüge , ELEVATORS / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Elevator people flow prediction and elevator dispatching system based on deep learning

    WEN WU / TANG NI / PU FEIYU et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Elevator emergency stop people trapping fault detection method based on multi-model fusion

    ZHONG CHAOWEN / WAN MIN / CAI WEIWEI et al. | Europäisches Patentamt | 2021

    Freier Zugriff

    Bus arrival time prediction method based on spatial-temporal feature deep learning

    LI YI / ZHANG MINGZE / YANG WENQIANG et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Elevator fault early warning method based on prediction model

    CHEN SHUFANG / LIU LIMEI / LI JUAN et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Traffic flow prediction method and device based on space-time sequence deep learning

    DU SHENGDONG / WU QIUCHI / HU JIE et al. | Europäisches Patentamt | 2023

    Freier Zugriff