技术领域Technical Field
本发明属于机电设备运行监测技术领域，具体涉及一种水厂机电设备运行工况监测方法、系统和设备。The present invention belongs to the technical field of electromechanical equipment operation monitoring, and in particular relates to a method, system and equipment for monitoring the operating conditions of electromechanical equipment in a water plant.
背景技术Background technique
在工业环境下，例如，对于水厂中的机电设备（主要是水泵和曝气用风机）的运行情况监测对于设备安全运行具有重要作用。目前对机电设备的运行情况进行监测的方式主要包括：（1）利用传感器采集相关运行数据，然后基于采集的运行数据进行阈值判定，从而判定设备是否存在异常；（2）基于大量历史数据，并结合人工智能算法对设备异常进行识别，例如使用神经网络或其他机器学习算法，以及机电设备的工况参数包括电流、电压、速度、温度等历史数据来学习判断识别是否处于正常运行状态。In an industrial environment, for example, monitoring the operation of electromechanical equipment (mainly water pumps and aeration fans) in a water plant plays an important role in the safe operation of the equipment. The current methods for monitoring the operation of electromechanical equipment mainly include: (1) using sensors to collect relevant operation data, and then making threshold judgments based on the collected operation data to determine whether the equipment is abnormal; (2) based on a large amount of historical data, combined with artificial intelligence algorithms to identify equipment anomalies, such as using neural networks or other machine learning algorithms, and the operating parameters of electromechanical equipment including current, voltage, speed, temperature and other historical data to learn and judge whether it is in normal operation.
然而，上述现有技术存在如下缺陷：单纯利用数值型参数进行异常识别，即传感器传输的数据来进行异常情况的监测和判断，这种方式容易存在误判，因为对于水厂水泵或风机这类机电设备，电流、电压等并不是判断异常的决定性因素，以及因传感器导致的数据异常等均会造成误判导致异常识别准确率不高。However, the above-mentioned existing technology has the following defects: simply using numerical parameters for abnormality identification, that is, using the data transmitted by the sensor to monitor and judge abnormal conditions, this method is prone to misjudgment, because for electromechanical equipment such as water plant pumps or fans, current, voltage, etc. are not the decisive factors for judging abnormalities, and data anomalies caused by sensors will cause misjudgment, resulting in low accuracy of abnormality identification.
发明内容Summary of the invention
为了解决现有监测技术存在的异常识别准确率不高的问题，本发明提出了一种水厂机电设备运行工况监测方法、系统和设备，本发明提出的方法针对不同类型传感器数据分别建立单独识别模型，然后根据单独识别体系的识别结果构建综合识别模型，同时对综合识别模型中各项参数权重进行动态更新，从而提高识别精度和异常预警精准度。In order to solve the problem of low accuracy of abnormality recognition in existing monitoring technologies, the present invention proposes a method, system and equipment for monitoring the operating conditions of electromechanical equipment in a water plant. The method proposed in the present invention establishes separate recognition models for different types of sensor data, and then constructs a comprehensive recognition model based on the recognition results of the separate recognition systems. At the same time, the weights of various parameters in the comprehensive recognition model are dynamically updated, thereby improving the recognition accuracy and the accuracy of abnormality warning.
本发明通过下述技术方案实现：The present invention is achieved through the following technical solutions:
一种水厂机电设备运行工况监测方法，所述方法包括：A method for monitoring the operating conditions of electromechanical equipment in a water plant, the method comprising:
针对机电设备各项监测参数分别建立各自的单源异常识别模型；其中，监测参数包括数值类监测参数和非数值类监测参数；Establish a single-source anomaly recognition model for each monitoring parameter of electromechanical equipment; the monitoring parameters include numerical monitoring parameters and non-numerical monitoring parameters;
基于各单源异常识别模型识别结果及相应的权重系数，构建综合异常识别模型并进行初始化，由此得到由单源异常识别模型和综合异常识别模型构成的设备异常识别模型；Based on the recognition results of each single-source anomaly recognition model and the corresponding weight coefficients, a comprehensive anomaly recognition model is constructed and initialized, thereby obtaining an equipment anomaly recognition model composed of a single-source anomaly recognition model and a comprehensive anomaly recognition model;
利用训练集对所述设备异常识别模型进行训练，根据机电设备实际异常情况对所述设备异常识别模型中的各项权重系数进行迭代调整，直到满足识别精度要求；其中，所述训练集通过机电设备各项监测参数的历史测量数据及其标记构建而成，所述标记通过对机电设备实际异常情况进行核查确定，所述标记包括异常和正常；The equipment abnormality recognition model is trained using a training set, and each weight coefficient in the equipment abnormality recognition model is iteratively adjusted according to the actual abnormal situation of the electromechanical equipment until the recognition accuracy requirement is met; wherein the training set is constructed by historical measurement data of various monitoring parameters of the electromechanical equipment and their labels, and the labels are determined by checking the actual abnormal situation of the electromechanical equipment, and the labels include abnormal and normal;
利用训练好的设备异常识别模型进行机电设备异常识别。Use the trained equipment anomaly recognition model to identify anomalies in electromechanical equipment.
在一些实施方式中，权重系数的迭代调整过程具体包括：In some implementations, the iterative adjustment process of the weight coefficient specifically includes:
当所述综合异常识别模型识别结果为异常，此时如果标记为异常，则增加所述综合异常识别模型中异常项权重，此时如果标记为正常，则降低所述综合异常识别模型中异常项权重；When the comprehensive anomaly recognition model identifies an abnormality, if it is marked as abnormal, the weight of the abnormal item in the comprehensive anomaly recognition model is increased; if it is marked as normal, the weight of the abnormal item in the comprehensive anomaly recognition model is reduced;
当所述综合异常识别模型识别结果为正常，此时如果标记为异常，则增加所述综合异常识别模型中异常识别正确的项的权重，同时降低所述综合异常识别模型中异常识别错误的项的权重，此时如果标记为正常，则保持当前权重不变。When the recognition result of the comprehensive anomaly recognition model is normal, if it is marked as abnormal, the weight of the item with correct abnormal recognition in the comprehensive anomaly recognition model is increased, and the weight of the item with incorrect abnormal recognition in the comprehensive anomaly recognition model is reduced. If it is marked as normal, the current weight is kept unchanged.
在一些实施方式中，对于数值类监测参数，单源异常识别模型构建过程具体包括：In some implementations, for numerical monitoring parameters, the single-source anomaly recognition model construction process specifically includes:
将该数值类监测参数的测量值与相应的阈值范围进行比较，根据比较结果输出识别结果；Compare the measured value of the numerical monitoring parameter with the corresponding threshold range, and output the recognition result according to the comparison result;
其中，阈值范围根据所述机电设备的实际运行要求和工作参数进行确定；Wherein, the threshold range is determined according to the actual operation requirements and working parameters of the electromechanical equipment;
对于非数值类监测参数，单源异常识别模型构建过程具体包括：For non-numerical monitoring parameters, the single-source anomaly recognition model construction process specifically includes:
按照预设周期获取若干个机电设备正常运行时的波形，将若干个波形分别转换为数据矩阵并标记为正常模型构建初始的模型库；Acquire waveforms of several electromechanical devices during normal operation according to a preset cycle, convert the waveforms into data matrices respectively and mark them as normal models to build an initial model library;
按照预设周期获取机电设备运行时的波形作为待识别波形；Acquire the waveform of the electromechanical equipment during operation as the waveform to be identified according to a preset period;
将待识别波形转换为待识别数据矩阵并将其与模型库中的正常模型一一进行相似度计算，根据相似度计算结果输出识别结果。The waveform to be identified is converted into a data matrix to be identified and its similarity is calculated one by one with the normal models in the model library, and the identification result is output according to the similarity calculation result.
在一些实施方式中，对于数值类参数，模型训练过程还包括：In some implementations, for numerical parameters, the model training process further includes:
根据机电设备的实际异常情况，对单源异常识别模型的识别结果进行核查，如果识别结果与机电设备的实际异常情况不相符，则对单源异常识别模型的阈值范围进行更新；According to the actual abnormal situation of the electromechanical equipment, the recognition result of the single-source abnormality recognition model is verified, and if the recognition result does not match the actual abnormal situation of the electromechanical equipment, the threshold range of the single-source abnormality recognition model is updated;
对于非数值类参数，模型训练过程还包括：For non-numeric parameters, the model training process also includes:
根据机电设备的实际异常情况，对单源异常识别模型的识别结果进行核查，如果识别结果与机电设备的实际异常情况不相符，则对单源异常识别模型的模型库进行更新。According to the actual abnormal situation of the electromechanical equipment, the recognition result of the single-source abnormality recognition model is checked. If the recognition result does not match the actual abnormal situation of the electromechanical equipment, the model library of the single-source abnormality recognition model is updated.
在一些实施方式中，所述综合异常识别模型构建过程具体包括：In some embodiments, the comprehensive abnormality recognition model construction process specifically includes:
建立异常函数，异常函数表示为：异常值等于各单源异常识别结果与其对应的权重系数的乘积之和除以各单源异常识别结果对应的权重系数的和；An abnormal function is established, and the abnormal function is expressed as follows: the abnormal value is equal to the sum of the products of each single-source abnormal recognition result and its corresponding weight coefficient divided by the sum of the weight coefficients corresponding to each single-source abnormal recognition result;
根据异常值与阈值的比较结果输出最终的识别结果；Output the final recognition result according to the comparison result between the outlier value and the threshold value;
对所述综合异常识别模型进行初始化的过程具体包括：The process of initializing the comprehensive anomaly recognition model specifically includes:
将所有权重系数初始化为相同。Initialize all weight coefficients to be the same.
在一些实施方式中，所述方法还包括：In some embodiments, the method further comprises:
利用设备异常识别模型进行机电设备异常识别过程中，动态调整所述设备异常识别模型中各项权重系数。In the process of using the equipment abnormality identification model to identify the abnormality of electromechanical equipment, the weight coefficients of each item in the equipment abnormality identification model are dynamically adjusted.
在一些实施方式中，所述方法还包括：In some embodiments, the method further comprises:
采用训练好的设备异常预警模型进行机电设备异常预警。Use the trained equipment abnormality warning model to provide abnormal warning for electromechanical equipment.
在一些实施方式中，所述设备异常预警模型训练过程为：In some implementations, the equipment abnormality warning model training process is:
构建时间序列模型和训练样本，所述训练样本通过各单源异常识别模型和综合异常识别输出的历史识别结果构成的时间序列及其机电设备实际异常标记构建而成；Constructing a time series model and training samples, wherein the training samples are constructed by the time series consisting of the historical recognition results output by each single-source anomaly recognition model and the comprehensive anomaly recognition output and the actual anomaly marks of the electromechanical equipment;
利用所述训练样本对时间序列模型进行训练，得到设备异常预警模型。The training samples are used to train the time series model to obtain an equipment abnormality warning model.
第二方面，本发明提出了一种水厂机电设备运行工况监测系统，所述系统包括：In a second aspect, the present invention provides a water plant electromechanical equipment operating condition monitoring system, the system comprising:
单源异常模型构建模块，所述单源异常模型构建模块针对机电设备各项监测参数分别建立各自的单源异常识别模型；其中，监测参数包括数值类监测参数和非数值类监测参数；A single-source anomaly model building module, wherein the single-source anomaly model building module establishes respective single-source anomaly recognition models for various monitoring parameters of electromechanical equipment; wherein the monitoring parameters include numerical monitoring parameters and non-numerical monitoring parameters;
综合异常模型构建模块，所述综合异常模型构建模块基于各单源异常识别模型识别结果及相应的权重系数，构建综合异常识别模型并进行初始化，由此得到由单元异常识别模型和综合异常识别模型构成的设备异常识别模型；A comprehensive abnormality model construction module, which constructs and initializes a comprehensive abnormality recognition model based on the recognition results of each single-source abnormality recognition model and the corresponding weight coefficient, thereby obtaining an equipment abnormality recognition model composed of a unit abnormality recognition model and a comprehensive abnormality recognition model;
模型训练模块，所述模型训练模块利用训练集对所述设备异常识别模型进行训练，根据机电设备实际异常情况对所述设备异常识别模型中的各项权重系数进行迭代调整，直到满足识别精度要求；其中，所述训练集通过机电设备各项监测参数的历史测量数据及其标记构建而成，所述标记通过对机电设备实际异常情况进行核查确定，所述标记包括异常和正常；A model training module, wherein the model training module trains the equipment abnormality recognition model using a training set, and iteratively adjusts various weight coefficients in the equipment abnormality recognition model according to the actual abnormal conditions of the electromechanical equipment until the recognition accuracy requirement is met; wherein the training set is constructed by historical measurement data of various monitoring parameters of the electromechanical equipment and their labels, and the labels are determined by checking the actual abnormal conditions of the electromechanical equipment, and the labels include abnormal and normal;
以及设备异常识别模块，所述设备异常识别模块利用训练好的设备异常识别模型进行机电设备异常识别。And an equipment anomaly recognition module, wherein the equipment anomaly recognition module uses a trained equipment anomaly recognition model to perform anomaly recognition of electromechanical equipment.
第三方面，本发明提出了一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现本发明所述方法的步骤。In a third aspect, the present invention proposes a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method of the present invention when executing the computer program.
本发明提出的一种水厂机电设备运行工况监测方法、系统和设备，首先分别针对不同监测参数建立各自的单源异常识别模型，然后基于各单源异常识别模型识别结果及其对应的权重系数构建综合异常识别模型并进行初始化，初始化时各项权重系数相同，利用历史测量数据对模型进行训练，此过程中，迭代调整模型中各项权重系数，提高异常识别精度和可靠性；此外，在模型训练过程中，不仅仅利用数据集训练模型，还结合异常核查，对单源异常识别模型进行迭代修正，进一步提高了异常识别精度。The present invention proposes a method, system and equipment for monitoring the operating conditions of electromechanical equipment in a water plant. First, respective single-source anomaly recognition models are established for different monitoring parameters. Then, based on the recognition results of each single-source anomaly recognition model and its corresponding weight coefficient, a comprehensive anomaly recognition model is constructed and initialized. During initialization, each weight coefficient is the same. The model is trained using historical measurement data. During this process, each weight coefficient in the model is iteratively adjusted to improve the accuracy and reliability of anomaly recognition. In addition, during the model training process, not only the data set is used to train the model, but also the single-source anomaly recognition model is iteratively corrected in combination with anomaly verification, thereby further improving the accuracy of anomaly recognition.
本发明提出的一种水厂机电设备运行工况监测方法、系统和设备，利用设备异常识别模型进行实时监测过程中，还可进一步动态调整模型中各项权重系数，进一步提高识别精度；The present invention proposes a method, system and device for monitoring the operating conditions of electromechanical equipment in a water plant. During the real-time monitoring process using the equipment abnormality recognition model, the weight coefficients of each item in the model can be further dynamically adjusted to further improve the recognition accuracy.
本发明提出的一种水厂机电设备运行工况监测方法、系统和设备，还基于历史的异常识别结果，结合时间序列模型对机电设备是否会发生异常进行较为精准地预测，从而实现提前预警，为机电设备监测和运维决策提供依据。The present invention proposes a method, system and equipment for monitoring the operating conditions of electromechanical equipment in a water plant. Based on historical abnormality identification results and combined with a time series model, it can make a relatively accurate prediction of whether an abnormality will occur in the electromechanical equipment, thereby achieving early warning and providing a basis for electromechanical equipment monitoring and operation and maintenance decision-making.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
此处所说明的附图用来提供对本发明实施例的进一步理解，构成本申请的一部分，并不构成对本发明实施例的限定。在附图中：The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
图1为本发明实施例的方法流程图；FIG1 is a flow chart of a method according to an embodiment of the present invention;
图2为本发明实施例构建的设备异常识别模型架构示意图；FIG2 is a schematic diagram of the architecture of a device anomaly recognition model constructed in an embodiment of the present invention;
图3为本发明实施例的权重系数迭代调整流程图；FIG3 is a flowchart of iterative adjustment of weight coefficients according to an embodiment of the present invention;
图4为本发明实施例的系统原理框图。FIG. 4 is a system principle block diagram of an embodiment of the present invention.
具体实施方式Detailed ways
为使本发明的目的、技术方案和优点更加清楚明白，下面结合实施例和附图，对本发明作进一步的详细说明，本发明的示意性实施方式及其说明仅用于解释本发明，并不作为对本发明的限定。In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary implementation modes of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.
实施例：现有技术主要是基于传感器传输的数值类数据进行异常情况的监测和判断，容易产生误识别，因为很多时候，例如电流、电压这些数值类参数并不是判断异常的决定性因素，同时由于传感器自身故障导致电流、电压这些数值类数据出现异常，从而造成识别精度较低的问题。针对此，本实施例提出了一种水厂机电设备运行工况监测方法，本实施例提出的方法首先将针对不同参数构建各自的单源异常识别模型，然后再基于若干单源异常识别模型的异常识别结果，构建综合预测模型，且该综合预测模型中异常因素的权重迭代更新，提高了识别精度。Embodiment: The existing technology mainly monitors and judges abnormal conditions based on numerical data transmitted by sensors, which is prone to misidentification, because in many cases, numerical parameters such as current and voltage are not the decisive factors for judging abnormalities. At the same time, due to the failure of the sensor itself, numerical data such as current and voltage are abnormal, resulting in low recognition accuracy. In view of this, this embodiment proposes a method for monitoring the operating conditions of electromechanical equipment in a water plant. The method proposed in this embodiment first constructs respective single-source abnormality recognition models for different parameters, and then constructs a comprehensive prediction model based on the abnormality recognition results of several single-source abnormality recognition models. The weights of abnormal factors in the comprehensive prediction model are iteratively updated to improve the recognition accuracy.
如图1所示，本实施例提出的方法具体包括如下步骤：As shown in FIG1 , the method proposed in this embodiment specifically includes the following steps:
步骤100，针对机电设备各项监测参数分别建立各自的单源异常识别模型。其中，对于数值类参数，如温度、湿度、电流、电压等，利用阈值判定逻辑构建相应的单源异常识别模型；对于非数值类参数，如振动或噪声波形，利用相似度比较方式构建相应的单源异常识别模型。Step 100: Establish a single-source anomaly recognition model for each monitoring parameter of the electromechanical equipment. For numerical parameters, such as temperature, humidity, current, voltage, etc., a corresponding single-source anomaly recognition model is constructed using threshold judgment logic; for non-numerical parameters, such as vibration or noise waveforms, a corresponding single-source anomaly recognition model is constructed using a similarity comparison method.
可选的，对于数值类参数，单源异常识别模型构建过程具体包括：将传感器采集到的数值类数据与相应的阈值范围进行比较，根据比较结果输出识别结果。具体的，如果超出阈值范围则识别为异常，否则识别为正常。阈值范围可根据机电设备的实际运行要求和工作参数等进行选择。Optionally, for numerical parameters, the single-source anomaly recognition model construction process specifically includes: comparing the numerical data collected by the sensor with the corresponding threshold range, and outputting the recognition result according to the comparison result. Specifically, if it exceeds the threshold range, it is recognized as abnormal, otherwise it is recognized as normal. The threshold range can be selected according to the actual operation requirements and working parameters of the electromechanical equipment.
对于非数值类参数，如设备振动、噪声等波形，单源异常识别模型构建过程具体包括：For non-numerical parameters, such as equipment vibration, noise and other waveforms, the single-source anomaly recognition model construction process specifically includes:
步骤101，按照预设周期获取若干个机电设备正常运行时的波形，将若干个波形分别转换为数据矩阵并标记为正常模型构建初始模型库；具体的，预设周期可以设置为30秒，即可以每30秒记录一段波形，预先收集一段时间的数据，例如15天内每30秒记录一次波形，就可以获得若干个正常模型；Step 101, according to a preset period, the waveforms of several electromechanical devices during normal operation are obtained, and the several waveforms are respectively converted into data matrices and marked as normal models to build an initial model library; specifically, the preset period can be set to 30 seconds, that is, a waveform can be recorded every 30 seconds, and data for a period of time is collected in advance, for example, a waveform is recorded every 30 seconds within 15 days, and several normal models can be obtained;
步骤102，按照预设周期获取机电设备运行时的波形作为待识别波形；Step 102, obtaining the waveform of the electromechanical device during operation as the waveform to be identified according to a preset period;
步骤103，将该待识别波形转换为待识别数据矩阵并将其与模型库中的正常模型一一进行相似度计算，根据相似度计算结果输出识别结果。具体的，如果匹配度大于阈值上限，则识别为正常，如果匹配度小于阈值下限，则识别为异常，否则识别为疑似异常；具体的，本实施例可以采用相似度计算法进行匹配，即分别计算待识别波形的数据矩阵与模型库中的正常模型进行相似度计算，如果相似度大于95%（可根据实际情况设置），则识别为正常，如果相似度小于80%（可根据实际情况设置），则识别为异常，处于90%和80%之间则识别为疑似异常。Step 103, convert the waveform to be identified into a data matrix to be identified and perform similarity calculations on it with the normal models in the model library one by one, and output the identification result according to the similarity calculation result. Specifically, if the matching degree is greater than the upper threshold, it is identified as normal, if the matching degree is less than the lower threshold, it is identified as abnormal, otherwise it is identified as suspected abnormal; specifically, this embodiment can use a similarity calculation method for matching, that is, respectively calculate the data matrix of the waveform to be identified and the normal model in the model library for similarity calculation, if the similarity is greater than 95% (can be set according to actual conditions), it is identified as normal, if the similarity is less than 80% (can be set according to actual conditions), it is identified as abnormal, and if it is between 90% and 80%, it is identified as suspected abnormal.
步骤200，基于各单源异常识别模型识别结果及相应的权重系数，构建综合异常识别模型并进行初始化，由此得到由单源异常识别模型和综合异常识别模型构成的设备异常识别模型。Step 200, based on the recognition results of each single-source abnormality recognition model and the corresponding weight coefficient, a comprehensive abnormality recognition model is constructed and initialized, thereby obtaining a device abnormality recognition model composed of the single-source abnormality recognition model and the comprehensive abnormality recognition model.
可选的，本实施例的综合异常识别模型构建及初始化过程具体包括：Optionally, the comprehensive anomaly recognition model construction and initialization process of this embodiment specifically includes:
步骤201，基于各单源异常识别模型的识别结果及其相应的权重系数构建综合异常识别模型；具体的，各项监测参数对应的单源异常识别结果分别为P1、P2、…、Pn，对应各项监测参数对应设置权重为w1、w2、…、wn，则构建的综合异常识别模型为：Step 201, construct a comprehensive anomaly recognition model based on the recognition results of each single-source anomaly recognition model and its corresponding weight coefficient; specifically, the single-source anomaly recognition results corresponding to each monitoring parameter areP1 ,P2 , …,Pn , and the corresponding weights of each monitoring parameter arew1 ,w2 , …,wn , then the constructed comprehensive anomaly recognition model is:
其中，P为异常值，如果P大于阈值，则综合异常识别模型识别结果为异常（记为1），P小于等于阈值，则综合异常识别模型识别结果为正常（记为0）；若根据第i项监测参数预测为异常，则其对应的单源异常预测结果Pi（i=1,2,…,n）为1，否则为0；n为机电设备的监测参数项数。Among them, P is the abnormal value. IfP is greater than the threshold, the identification result of the comprehensive abnormality identification model is abnormal (denoted as 1); ifP is less than or equal to the threshold, the identification result of the comprehensive abnormality identification model is normal (denoted as 0); if it is predicted to be abnormal according to thei- th monitoring parameter, the corresponding single-source abnormality prediction resultPi (i=1,2,…,n ) is 1, otherwise it is 0;n is the number of monitoring parameters of electromechanical equipment.
步骤202，综合异常识别模型中的所有权重初始为相同，得到初始化的综合异常识别模型。即初始化时，各识别结果的权重（即各监测参数）对于最终的异常识别影响程度相同。基于此，得到的设备异常识别模型如图2所示。In step 202, all weights in the comprehensive anomaly recognition model are initially the same, and an initialized comprehensive anomaly recognition model is obtained. That is, during initialization, the weights of each recognition result (i.e., each monitoring parameter) have the same influence on the final anomaly recognition. Based on this, the obtained device anomaly recognition model is shown in FIG2.
步骤300，利用训练集对设备异常识别模型进行训练，根据机电设备实际异常情况对异常识别模型中的各项权重系数进行迭代调整，直到满足识别精度要求。训练集通过传感器采集的历史数据及其标记构建而成，该标记可通过对机电设备实际异常情况进行核查实现，分为异常（记为1）和正常（记为0）。Step 300, train the equipment abnormality recognition model using the training set, and iteratively adjust the weight coefficients in the abnormality recognition model according to the actual abnormality of the electromechanical equipment until the recognition accuracy requirement is met. The training set is constructed by the historical data collected by the sensor and its labeling, which can be realized by checking the actual abnormality of the electromechanical equipment and is divided into abnormal (marked as 1) and normal (marked as 0).
具体如图3所示，权重系数迭代调整过程具体可包括：As shown in FIG3 , the weight coefficient iterative adjustment process may specifically include:
当综合异常识别模型识别结果为异常，此时如果标记为异常，则增加综合异常识别模型中异常项权重；如果标记为正常，则降低综合异常识别模型中异常项权重；通过上述动态调整权重的方式，可以在下一次发生异常时，正确地综合识别异常的概率增加，错误地综合识别异常的概率降低，从而降低了误判率，提高了识别精度；When the comprehensive anomaly recognition model identifies an anomaly, if it is marked as an anomaly, the weight of the anomaly item in the comprehensive anomaly recognition model is increased; if it is marked as normal, the weight of the anomaly item in the comprehensive anomaly recognition model is reduced; by dynamically adjusting the weights, the next time an anomaly occurs, the probability of correctly identifying an anomaly is increased, and the probability of incorrectly identifying an anomaly is reduced, thereby reducing the misjudgment rate and improving the recognition accuracy;
当综合异常识别模型识别结果为正常，此时如果标记为异常，则增加综合异常识别模型中异常识别正确的项的权重，同时降低综合异常识别模型中异常识别错误的项的权重；如果标记为正常，则保持当前权重不变；综合识别为正常，但是机电设备实际异常，这种情况即异常识别完全没有做出响应，说明综合异常识别模型中各项权重设置不合理，则需要对权重的调整强度更大，因此，此时可将发生异常时异常判断正确的项的权重提高，同时也降低异常判断错误的项的权重，从而快速调整各项对异常预测的影响程度，实现及时的异常响应，提高异常识别精度；而在综合识别为正常，且机电设备实际也正常的情况下，说明综合异常识别模型正确运行，此时可以不进行权重的调整。采用上述权重动态实时调整技术，经过一段时间的迭代，综合异常识别模型中各项权重会发生变化，针对机电设备，异常发生时，对异常影响程度更高的项的权重增加，对异常影响程度更小的项的权重减小，从而得到异常识别精准度更高的模型，提高了机电设备异常识别精度准和可靠性，保障了机电设备运行的安全性和可靠性。When the recognition result of the comprehensive anomaly recognition model is normal, if it is marked as abnormal, the weight of the item with correct anomaly recognition in the comprehensive anomaly recognition model is increased, and the weight of the item with incorrect anomaly recognition in the comprehensive anomaly recognition model is reduced; if it is marked as normal, the current weight remains unchanged; the comprehensive recognition is normal, but the electromechanical equipment is actually abnormal. In this case, the abnormal recognition has no response at all, indicating that the weight settings of each item in the comprehensive anomaly recognition model are unreasonable, and the weight adjustment intensity needs to be stronger. Therefore, at this time, the weight of the item with correct abnormal judgment when an abnormality occurs can be increased, and the weight of the item with incorrect abnormal judgment can be reduced, so as to quickly adjust the influence of each item on abnormal prediction, realize timely abnormal response, and improve the accuracy of abnormal recognition; when the comprehensive recognition is normal and the electromechanical equipment is actually normal, it means that the comprehensive anomaly recognition model is running correctly, and the weight adjustment does not need to be performed at this time. By adopting the above-mentioned dynamic real-time weight adjustment technology, after a period of iteration, the weights of each item in the comprehensive abnormality recognition model will change. For electromechanical equipment, when an abnormality occurs, the weights of items with a greater degree of impact on the abnormality will increase, and the weights of items with a smaller degree of impact on the abnormality will decrease, thereby obtaining a model with higher abnormality recognition accuracy, improving the accuracy and reliability of abnormality recognition of electromechanical equipment, and ensuring the safety and reliability of the operation of electromechanical equipment.
本实施例以水泵为例对上述综合异常监测过程进行详细说明，参与异常识别的各项监测参数包括：温度（P1）、湿度（P2）、烟雾（P3）、电流（P4）、电压（P5）、水压（P6）、振动（P7）和噪声强度（P8）共八项，这八项监测参数如果预测为异常，则返回1，否则返回0。初始权重均设置为1，则综合异常识别模型为：This embodiment takes a water pump as an example to explainthe above-mentioned comprehensive abnormality monitoring process in detail.The monitoring parameters involvedin abnormality identification include: temperature (P1) , humidity (P2 ), smoke (P3 ), current (P4 ), voltage (P5 ), waterpressure (P6 ), vibration (P7 ) and noise intensity (P8 ). If these eight monitoring parameters are predicted to be abnormal, 1 is returned, otherwise0 is returned. The initial weights are all set to 1, and the comprehensive abnormality identification model is:
利用上述综合异常识别模型进行水泵运行监测，当监测到大于0.5时，则认为发生异常，小于等于0.5时，则认为未发生异常（即正常）。The above-mentioned comprehensive abnormality recognition model is used to monitor the operation of the pump. When it is greater than 0.5, it is considered abnormal; when it is less than or equal to 0.5, it is considered normal.
（1）当综合识别出现异常，且核查水泵实际异常，则表明识别正确，此次预警过程中，温度、电流、电压、振动的异常预测为1，即异常，其余项异常判断为0，即不异常，则将此次预警过程中异常项的权重提升3%，即调整后的权重为温度1.03，电流1.03，电压1.03，振动1.03，调整后的综合异常识别模型为：(1) When an abnormality occurs in the comprehensive identification and the actual abnormality of the water pump is verified, it indicates that the identification is correct. During this early warning process, the abnormal predictions of temperature, current, voltage, and vibration are 1, that is, abnormal, and the abnormal judgments of the other items are 0, that is, not abnormal. The weight of the abnormal items in this early warning process is increased by 3%, that is, the adjusted weights are temperature 1.03, current 1.03, voltage 1.03, and vibration 1.03. The adjusted comprehensive abnormality identification model is:
此种情况下，异常识别值不会出现负值，且偶尔成功参与预警的项权重变动较小，连续多次成功参与的项权重会快速上升。In this case, the anomaly identification value will not be negative, and the weight of the items that occasionally successfully participate in the early warning will change little, while the weight of the items that successfully participate for many consecutive times will increase rapidly.
（2）当综合识别出现异常，但是核查水泵实际未发生异常，则表明识别错误，此次预警过程中，温度、电流、电压、振动的异常预测为1，即异常，其余项异常判断为0，即不异常，则将此次预警过程中判断为异常的项的权重下调3%，即调整后的权重为温度0.97，电流0.97，电压0.97，振动0.97，调整后的综合异常识别模型为：(2) When an abnormality occurs in the comprehensive identification, but the water pump is not actually abnormal after verification, it indicates that the identification is wrong. During this early warning process, the abnormal predictions of temperature, current, voltage, and vibration are 1, that is, abnormal, and the abnormal judgments of other items are 0, that is, not abnormal. The weights of the items judged as abnormal during this early warning process are reduced by 3%, that is, the adjusted weights are 0.97 for temperature, 0.97 for current, 0.97 for voltage, and 0.97 for vibration. The adjusted comprehensive abnormality identification model is:
（3）当综合识别为正常，但是核查水泵实际异常，则表明预警未响应，则权重设置不合理，因此对于权重的更改强度会加大，如上所述，此次预警过程中，温度、电流、电压、振动的异常预测为1，即异常，其余项异常判断为0，即不异常，则将此次预警过程中异常判断正确的项的权重提高10%，同时也将异常判断错误的项的权重降低5%，即调整后的权重为温度1.1，电流1.1，电压1.1，振动1.1，湿度0.95，水压0.95，噪声强度0.95，烟雾0.95，从而对于异常发生不响应的项的权重会发生快速的变化，调整后的综合异常识别模型为：(3) When the comprehensive identification is normal, but the actual abnormality of the water pump is verified, it means that the warning has not responded, and the weight setting is unreasonable. Therefore, the intensity of the weight change will be increased. As mentioned above, during this warning process, the abnormal prediction of temperature, current, voltage, and vibration is 1, that is, abnormal, and the abnormal judgment of other items is 0, that is, not abnormal. In this case, the weight of the items with correct abnormal judgment in this warning process will be increased by 10%, and the weight of the items with incorrect abnormal judgment will be reduced by 5%. That is, the adjusted weights are temperature 1.1, current 1.1, voltage 1.1, vibration 1.1, humidity 0.95, water pressure 0.95, noise intensity 0.95, and smoke 0.95. As a result, the weights of the items that do not respond to abnormalities will change rapidly. The adjusted comprehensive abnormality recognition model is:
（4）当综合识别为正常，且核查水泵实际正常，则表明正确预警，此时可不进行权重的调整。(4) When the comprehensive identification is normal and the water pump is actually normal, it indicates a correct warning and no weight adjustment is required.
可选的，对于数值类参数，模型训练过程具体还可包括：根据机电设备的实际异常反馈情况，对单源异常识别模型的识别结果进行核查，如果识别结果与机电设备实际异常情况不相符，则对阈值范围进行更新，进一步提高识别精准度。Optionally, for numerical parameters, the model training process may also specifically include: verifying the recognition results of the single-source abnormality recognition model based on the actual abnormal feedback of the electromechanical equipment; if the recognition results do not match the actual abnormal situation of the electromechanical equipment, updating the threshold range to further improve the recognition accuracy.
可选的，对于非数值类参数，模型训练过程具体还可包括：Optionally, for non-numerical parameters, the model training process may further include:
根据机电设备实际异常反馈情况，对识别结果进行进一步核查；该核查是为了避免异常误判，例如根据异常情况的反馈，设备并没有异常，但是识别为异常，则核查为正常；将核查后的待识别波形进行相应预测结果标记并更新模型库，即正常运行状态的标记为正常模型录入模型库，异常运行状态的标记为异常模型录入模型库。可选的，模型库中的数据也可以作为训练数据用于辅助异常预警模型的建立，以实现较为精准的异常识别预警。According to the actual abnormal feedback of the electromechanical equipment, the identification results are further verified; this verification is to avoid misjudgment of abnormalities. For example, according to the feedback of abnormal conditions, the equipment is not abnormal, but it is identified as abnormal, so it is verified as normal; the waveform to be identified after verification is marked with the corresponding prediction result and the model library is updated, that is, the normal operating state is marked as a normal model and entered into the model library, and the abnormal operating state is marked as an abnormal model and entered into the model library. Optionally, the data in the model library can also be used as training data to assist in the establishment of an abnormal warning model to achieve more accurate abnormal identification and warning.
步骤400，利用训练好的设备异常识别模型进行机电设备异常识别。可选地，该步骤400还包括该设备异常识别模型在实际监测过程中，可动态实时调整模型中的各项权重系数，从而进一步提高识别准确性和可靠性。需要说明的是，步骤400中权重的具体调整方式如步骤300中所述，此处不再过多赘述。Step 400, using the trained equipment anomaly recognition model to identify mechanical and electrical equipment anomalies. Optionally, step 400 also includes the equipment anomaly recognition model being able to dynamically and real-time adjust the weight coefficients of each item in the model during the actual monitoring process, thereby further improving the recognition accuracy and reliability. It should be noted that the specific adjustment method of the weight in step 400 is as described in step 300, and will not be repeated here.
相较于现有的仅以传统数据集作为训练集训练模型进行异常识别的技术，本实施例提出的方法首先分别针对不同监测参数建立各自的单源异常识别模型；然后基于各单源异常识别模型识别结果及其对应的权重系数构建综合异常识别模型并进行初始化，初始化时各项权重系数相同，即各项监测参数对异常识别结果影响程度相同，利用训练集对模型进行训练，此过程中，迭代调整各项权重系数，提高异常识别精度和可靠性。此外，本实施例提出的方法同时在模型训练过程中，不仅仅利用数据集训练模型，还结合异常核查，对单源异常识别模型进行迭代修正，进一步提高了异常识别精度。本实施例提出的方法在利用设备异常识别模型进行实时监测过程中，也可进一步动态调整模型中各项权重系数，进一步提高识别精度。Compared with the existing technology of using only traditional data sets as training sets to train models for anomaly identification, the method proposed in this embodiment first establishes separate single-source anomaly identification models for different monitoring parameters; then, based on the identification results of each single-source anomaly identification model and its corresponding weight coefficient, a comprehensive anomaly identification model is constructed and initialized. During initialization, each weight coefficient is the same, that is, each monitoring parameter has the same degree of influence on the anomaly identification result. The model is trained using the training set. During this process, each weight coefficient is iteratively adjusted to improve the accuracy and reliability of anomaly identification. In addition, during the model training process, the method proposed in this embodiment not only uses the data set to train the model, but also combines anomaly verification to iteratively correct the single-source anomaly identification model, thereby further improving the accuracy of anomaly identification. During the real-time monitoring process using the equipment anomaly identification model, the method proposed in this embodiment can also further dynamically adjust the weight coefficients of each item in the model to further improve the identification accuracy.
可选的，本实施例还基于历史的异常识别结果，结合时间序列模型对于水泵和风机等水务场景下会的机电设备是否发生异常进行预测，从而实现提前预警。基于此，本实施例提出的方法还包括：Optionally, this embodiment also predicts whether abnormalities will occur in electromechanical equipment such as water pumps and fans in water service scenarios based on historical abnormality recognition results and in combination with a time series model, thereby achieving early warning. Based on this, the method proposed in this embodiment also includes:
步骤500，采用训练好的设备异常预警模型进行机电设备异常预警。Step 500: Use the trained equipment abnormality warning model to perform abnormal warning of electromechanical equipment.
具体的，设备异常预警模型训练过程具体包括：Specifically, the equipment abnormality warning model training process includes:
步骤501，构建时间序列模型和训练样本。其中，该时间序列模型可采用ARIMA、LSTM等。训练样本通过历史异常识别结果（包括单源异常识别结果和综合异常识别结果）的时间序列和设备实际异常标记构建而成。Step 501, construct a time series model and training samples. The time series model may adopt ARIMA, LSTM, etc. The training samples are constructed by the time series of historical anomaly recognition results (including single-source anomaly recognition results and comprehensive anomaly recognition results) and actual anomaly marks of the equipment.
步骤502，利用训练样本对时间序列模型进行训练，从而得到设备异常预警模型。Step 502: Use training samples to train the time series model, thereby obtaining an equipment abnormality warning model.
将实时获取的异常识别结果时间序列输入到设备异常预警模型中，即可得到预测结果（异常或正常），从而为机电设备监测以及运维策略制定提供依据。By inputting the real-time abnormal identification result time series into the equipment abnormality warning model, the prediction result (abnormal or normal) can be obtained, thus providing a basis for electromechanical equipment monitoring and operation and maintenance strategy formulation.
本实施例以某一水厂中的水泵相关数据对上述设备异常预警模型的训练和应用过程进行进一步说明。This embodiment further illustrates the training and application process of the above-mentioned equipment abnormality warning model using water pump related data in a water plant.
（1）在异常发生时，即通过设备异常识别模型监测完成后，记录异常发生前24h的出水水压情况，流量情况以及管网水力分布中各压力监测点的压力数据。形成异常数据组，这些数据是标签为异常的数据的特征值。例如，收集频率为一小时一次，则形成一次异常的数据。假设异常次数为20次，则有20组 24×X的异常数据组，如下表1所示。(1) When an abnormality occurs, that is, after the monitoring is completed through the equipment abnormality identification model, the outlet water pressure, flow rate and pressure data of each pressure monitoring point in the hydraulic distribution of the pipe network 24 hours before the abnormality occurs are recorded. An abnormal data group is formed, which is the characteristic value of the data labeled as abnormal. For example, if the collection frequency is once an hour, an abnormal data is formed once. Assuming that the number of abnormalities is 20 times, there are 20 groups of 24×X abnormal data groups, as shown in Table 1 below.
表1Table 1
（2）水泵连续工作两周以上未发生异常的情况，视为正常情况，抽取与异常发生数量相等的连续数据量形成正常数据组。例如历史共发生20次异常，总计异常发生前的20天异常数据，则对应选取正常日中20天的正常数据形成相应数据库，根据上述异常表格形成对应的正常数据的特征值（即在正常一段时间内不发生异常的正常日的数据，一般情况下会有部分报告异常，但整体逐渐自动回归为0的情况，体现的是对部分点位异常值可忽略的情况），则有20组24×X的正常数据组，如下表2所示。(2) If the pump works continuously for more than two weeks without any abnormality, it is considered normal. The number of continuous data equal to the number of abnormalities is extracted to form a normal data group. For example, if there are 20 abnormalities in history, and the total number of abnormal data is 20 days before the abnormality occurs, then the normal data of 20 days in normal days are selected to form a corresponding database. The characteristic values of the corresponding normal data are formed according to the above abnormal table (that is, the data of normal days without abnormalities within a normal period of time, generally there will be some abnormal reports, but the overall data will gradually return to 0 automatically, reflecting the situation that the abnormal values of some points can be ignored). There are 20 groups of 24×X normal data groups, as shown in Table 2 below.
表2Table 2
（3）分析正常数据组和异常数据组中的特征值，即单源异常识别模型和综合异常识别模型识别结果，采用时间序列模型训练形成设备异常预警模型，即水泵异常前24h内的数据模型。(3) Analyze the characteristic values in the normal data group and the abnormal data group, that is, the recognition results of the single-source anomaly recognition model and the comprehensive anomaly recognition model, and use the time series model training to form an equipment anomaly warning model, that is, the data model within 24 hours before the water pump anomaly occurs.
（4）将实时数据输入到设备异常预警模型中进行预测，即输入数据为当前时刻前24h的对应数据，得出当前时刻的预测结果为正常或异常。(4) Input the real-time data into the equipment abnormality warning model for prediction, that is, the input data is the corresponding data 24 hours before the current moment, and the prediction result at the current moment is normal or abnormal.
本实施例以某一水厂中的风机相关数据对上述设备异常预警模型的训练和应用过程进行进一步说明。This embodiment further illustrates the training and application process of the above-mentioned equipment abnormality warning model by using the relevant data of the fans in a water plant.
首先，结合治水过程中的一些相关水质参数，如cod，水流量等，如表3所示。First, some relevant water quality parameters in the water treatment process, such as COD, water flow, etc., are combined, as shown in Table 3.
表3table 3
同样，按照上述水泵的模型建立和预测流程进行后续的流程，即可得到污水厂环境下曝气风机的异常预测结果。Similarly, by following the model building and prediction process of the water pump mentioned above, the abnormal prediction results of the aeration fan in the sewage treatment plant environment can be obtained.
基于上述相同技术构思，本实施例还提出了一种水厂机电设备运行工况监测系统，如图4所示，本实施例提出的系统具体包括：Based on the same technical concept as above, this embodiment also proposes a water plant electromechanical equipment operating condition monitoring system, as shown in FIG4 , the system proposed in this embodiment specifically includes:
单源异常模型构建模块，该模块针对机电设备各项监测参数分别建立各自的单源异常识别模型，其中，对于数值类参数，如温度、湿度、电流、电压等，利用阈值判定逻辑构建相应的单源异常识别模型；对于非数值类参数，如振动或噪声波形，利用相似度比较方式构建相应的单源异常识别模型；Single-source anomaly model building module, which builds single-source anomaly recognition models for each monitoring parameter of electromechanical equipment. For numerical parameters such as temperature, humidity, current, voltage, etc., the corresponding single-source anomaly recognition model is built using threshold judgment logic; for non-numerical parameters such as vibration or noise waveform, the corresponding single-source anomaly recognition model is built using similarity comparison method;
综合异常模型构建模块，该模块基于各单源异常识别模型识别结果与其相应的权重系数，构建综合异常识别模型并进行初始化，由此得到由单源异常识别模型和综合异常识别模型构成的设备异常识别模型；A comprehensive anomaly model construction module, which constructs and initializes a comprehensive anomaly recognition model based on the recognition results of each single-source anomaly recognition model and its corresponding weight coefficient, thereby obtaining an equipment anomaly recognition model composed of a single-source anomaly recognition model and a comprehensive anomaly recognition model;
模型训练模块，该模块利用训练集对设备异常识别模型进行训练，根据机电设备异常反馈情况对综合异常识别模型中的各项权重进行迭代调整，直到满足识别精度。The model training module uses the training set to train the equipment anomaly recognition model, and iteratively adjusts the weights in the comprehensive anomaly recognition model according to the feedback of electromechanical equipment anomalies until the recognition accuracy is met.
以及，设备异常识别模块，该模块利用训练好的设备异常识别模型进行机电设备异常识别。And, an equipment anomaly recognition module, which uses a trained equipment anomaly recognition model to perform anomaly recognition of electromechanical equipment.
可选的，本实施例提出的系统还包括：Optionally, the system provided in this embodiment further includes:
设备异常预警模块，该模块采用训练好的设备异常预警模型进行机电设备异常预警。The equipment abnormality warning module uses the trained equipment abnormality warning model to provide abnormal warning for electromechanical equipment.
本领域内的技术人员应明白，本申请的实施例可提供为方法、系统、或计算机程序产品。因此，本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质（包括但不限于磁盘存储器、CD-ROM、光学存储器等）上实施的计算机程序产品的形式。Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
本申请是参照根据本申请实施例的方法、设备（系统）、和计算机程序产品的流程图和／或方框图来描述的。应理解可由计算机程序指令实现流程图和／或方框图中的每一流程和／或方框、以及流程图和／或方框图中的流程和／或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能的装置。The present application is described with reference to the flowcharts and/or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and/or box in the flowchart and/or block diagram, as well as the combination of the processes and/or boxes in the flowchart and/or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能。These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和／或方框图一个方框或多个方框中指定的功能的步骤。These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.
以上所述的具体实施方式，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施方式而已，并不用于限定本发明的保护范围，凡在本发明的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410864625.6ACN118392255B (en) | 2024-07-01 | 2024-07-01 | Method, system and equipment for monitoring operation condition of electromechanical equipment of water plant |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410864625.6ACN118392255B (en) | 2024-07-01 | 2024-07-01 | Method, system and equipment for monitoring operation condition of electromechanical equipment of water plant |
| Publication Number | Publication Date |
|---|---|
| CN118392255Atrue CN118392255A (en) | 2024-07-26 |
| CN118392255B CN118392255B (en) | 2024-09-27 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410864625.6AActiveCN118392255B (en) | 2024-07-01 | 2024-07-01 | Method, system and equipment for monitoring operation condition of electromechanical equipment of water plant |
| Country | Link |
|---|---|
| CN (1) | CN118392255B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542262A (en)* | 2012-01-04 | 2012-07-04 | 东南大学 | Waveform identification method based on operating-characteristic working condition waveform library of high-speed rail |
| US8306931B1 (en)* | 2009-08-06 | 2012-11-06 | Data Fusion & Neural Networks, LLC | Detecting, classifying, and tracking abnormal data in a data stream |
| CN109524139A (en)* | 2018-10-23 | 2019-03-26 | 中核核电运行管理有限公司 | A kind of real-time device performance monitoring method based on equipment working condition variation |
| US20200097810A1 (en)* | 2018-09-25 | 2020-03-26 | Oracle International Corporation | Automated window based feature generation for time-series forecasting and anomaly detection |
| CN112258620A (en)* | 2018-12-28 | 2021-01-22 | 英特尔公司 | Apparatus and method for ray tracing instruction processing and execution |
| CN113673600A (en)* | 2021-08-23 | 2021-11-19 | 中海石油气电集团有限责任公司 | Industrial signal abnormity early warning method, system, storage medium and computing equipment |
| CN114548236A (en)* | 2022-01-28 | 2022-05-27 | 郑州信大捷安信息技术股份有限公司 | Production condition monitoring method and system |
| US20220187140A1 (en)* | 2020-12-14 | 2022-06-16 | Yokogawa Electric Corporation | Abnormal temperature detection device, abnormal temperature detection method, and non-transitory computer readable storage medium |
| CN115438747A (en)* | 2022-09-29 | 2022-12-06 | 中国农业银行股份有限公司 | Abnormal account recognition model training method, device, equipment and medium |
| JP2023018316A (en)* | 2021-07-27 | 2023-02-08 | 京セラドキュメントソリューションズ株式会社 | Image recognition method, image recognition device, and image recognition program |
| CN115902646A (en)* | 2023-01-06 | 2023-04-04 | 中国电力科学研究院有限公司 | Energy storage battery fault identification method and system |
| CN116434341A (en)* | 2023-04-23 | 2023-07-14 | 昆明理工大学 | Student classroom abnormal behavior identification method and system |
| CN117762913A (en)* | 2023-12-14 | 2024-03-26 | 广西电网有限责任公司 | An automatic identification method and system for offshore wind power abnormal power data |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8306931B1 (en)* | 2009-08-06 | 2012-11-06 | Data Fusion & Neural Networks, LLC | Detecting, classifying, and tracking abnormal data in a data stream |
| CN102542262A (en)* | 2012-01-04 | 2012-07-04 | 东南大学 | Waveform identification method based on operating-characteristic working condition waveform library of high-speed rail |
| US20200097810A1 (en)* | 2018-09-25 | 2020-03-26 | Oracle International Corporation | Automated window based feature generation for time-series forecasting and anomaly detection |
| CN109524139A (en)* | 2018-10-23 | 2019-03-26 | 中核核电运行管理有限公司 | A kind of real-time device performance monitoring method based on equipment working condition variation |
| CN112258620A (en)* | 2018-12-28 | 2021-01-22 | 英特尔公司 | Apparatus and method for ray tracing instruction processing and execution |
| US20220187140A1 (en)* | 2020-12-14 | 2022-06-16 | Yokogawa Electric Corporation | Abnormal temperature detection device, abnormal temperature detection method, and non-transitory computer readable storage medium |
| JP2023018316A (en)* | 2021-07-27 | 2023-02-08 | 京セラドキュメントソリューションズ株式会社 | Image recognition method, image recognition device, and image recognition program |
| CN113673600A (en)* | 2021-08-23 | 2021-11-19 | 中海石油气电集团有限责任公司 | Industrial signal abnormity early warning method, system, storage medium and computing equipment |
| CN114548236A (en)* | 2022-01-28 | 2022-05-27 | 郑州信大捷安信息技术股份有限公司 | Production condition monitoring method and system |
| CN115438747A (en)* | 2022-09-29 | 2022-12-06 | 中国农业银行股份有限公司 | Abnormal account recognition model training method, device, equipment and medium |
| CN115902646A (en)* | 2023-01-06 | 2023-04-04 | 中国电力科学研究院有限公司 | Energy storage battery fault identification method and system |
| CN116434341A (en)* | 2023-04-23 | 2023-07-14 | 昆明理工大学 | Student classroom abnormal behavior identification method and system |
| CN117762913A (en)* | 2023-12-14 | 2024-03-26 | 广西电网有限责任公司 | An automatic identification method and system for offshore wind power abnormal power data |
| Title |
|---|
| 柳青秀;马红占;褚学宁;马斌彬;王峥;: "基于长短时记忆―自编码神经网络的风电机组性能评估及异常检测", 计算机集成制造系统, no. 12, 15 December 2019 (2019-12-15), pages 233 - 243* |
| Publication number | Publication date |
|---|---|
| CN118392255B (en) | 2024-09-27 |
| Publication | Publication Date | Title |
|---|---|---|
| JP7069269B2 (en) | Semi-supervised methods and systems for deep anomaly detection for large industrial surveillance systems based on time series data using digital twin simulation data | |
| US11169514B2 (en) | Unsupervised anomaly detection, diagnosis, and correction in multivariate time series data | |
| CN111275288A (en) | XGboost-based multi-dimensional data anomaly detection method and device | |
| CN117010263A (en) | Residual life prediction method based on convolutional neural network and long-term and short-term memory network | |
| WO2020242553A1 (en) | Failure prediction using gradient-based sensor identification | |
| CN111275136B (en) | Fault prediction system based on small sample and early warning method thereof | |
| CN119398253A (en) | A smart management platform for power plants | |
| CN112611584B (en) | Fatigue failure detection method, device, equipment and medium for wind generating set | |
| CN116235121A (en) | Apparatus and method for identifying anomalies in an industrial facility for performing a production process | |
| CN119129446B (en) | A real-time fault diagnosis method and device for a turbine cooler | |
| CN114970709B (en) | Improved GA-based data-driven AHU multi-fault diagnosis feature selection method | |
| CN115361424B (en) | Medical equipment monitoring method and system based on data analysis of Internet of things | |
| CN117803580A (en) | Energy-saving control method and device for centrifugal pump based on machine learning | |
| CN119357917B (en) | Data-driven over-temperature monitoring method and device, storage medium, and terminal | |
| CN119004157B (en) | Wind turbine generator converter temperature abnormality early warning method and device and electronic equipment | |
| CN119727468A (en) | A control method and system for a brushless DC reduction motor | |
| CN118392255B (en) | Method, system and equipment for monitoring operation condition of electromechanical equipment of water plant | |
| CN117365923B (en) | Intelligent control method and system for air compressor | |
| CN118115081A (en) | Unattended cloud cabin health state assessment method | |
| US11144046B2 (en) | Fault signal recovery apparatus and method | |
| CN115840919A (en) | Fault prediction method and device for sensing control system | |
| CN118714168B (en) | Remote maintenance method based on Internet of things equipment | |
| CN120372461B (en) | Concept drift detection method and system based on self-adaptive data driving | |
| CN118625744B (en) | Automatic environment control method in chicken farm | |
| CN119046861B (en) | Method and system for predicting residual life of air compressor driven by real-time data |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | Address after:Room 201-208, 2nd Floor, Unit 3, Building 3, No. 68 Bailuwan 1st Street, Jinjiang District, Chengdu City, Sichuan Province, 610000 Patentee after:Aotu Technology Co.,Ltd. Country or region after:China Address before:Room 201-208, 2nd Floor, Unit 3, Building 3, No. 68 Bailuwan 1st Street, Jinjiang District, Chengdu City, Sichuan Province, 610000 Patentee before:Sichuan Aotu Technology Co.,Ltd. Country or region before:China | |
| CP03 | Change of name, title or address |