WO2018176213A1 - 高温预警方法与装置 - Google Patents

高温预警方法与装置 Download PDF

Info

Publication number
WO2018176213A1
WO2018176213A1 PCT/CN2017/078396 CN2017078396W WO2018176213A1 WO 2018176213 A1 WO2018176213 A1 WO 2018176213A1 CN 2017078396 W CN2017078396 W CN 2017078396W WO 2018176213 A1 WO2018176213 A1 WO 2018176213A1
Authority
WO
WIPO (PCT)
Prior art keywords
value
temperature
equipment room
layer
intermediate layer
Prior art date
Application number
PCT/CN2017/078396
Other languages
English (en)
French (fr)
Inventor
向洁
杨金龙
唐璐
马会明
Original Assignee
深圳中兴力维技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳中兴力维技术有限公司 filed Critical 深圳中兴力维技术有限公司
Priority to PCT/CN2017/078396 priority Critical patent/WO2018176213A1/zh
Publication of WO2018176213A1 publication Critical patent/WO2018176213A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor

Definitions

  • the invention relates to the field of high temperature early warning technology, in particular to a high temperature early warning method and device.
  • the fault handling method for the core equipment of the equipment room mainly stays at the stage of fault generation and alarm to processing.
  • the core equipment fails, it will bring huge losses to the production, and high temperature is one of the core elements that directly cause equipment failure.
  • the object of the present invention is to provide a high temperature early warning method and device, aiming at realizing high temperature warning of equipment.
  • the present invention provides a high temperature warning method, comprising the following steps:
  • the obtained value is converted by a back propagation neural network model to calculate a predicted temperature value of the equipment room;
  • the step of obtaining the value of the parameter that affects the temperature of the equipment room equipment includes:
  • the resistance value of the device is obtained by a resistance sensor.
  • it also includes:
  • the back propagation neural network model including an input layer, an intermediate layer, and an output layer;
  • the back propagation neural network model is trained.
  • the step of training the back propagation neural network model includes:
  • the weights and thresholds of the input layer and the intermediate layer are updated by a gradient descent method.
  • the step of calculating, according to the value of the parameter that affects the temperature of the equipment room, and the weight and threshold of the input layer and the intermediate layer, the predicted temperature value of the equipment room includes:
  • the output value of the output layer is calculated by using an output value of the intermediate layer as an input value of the output layer, and the predicted temperature value is obtained by a normalization algorithm.
  • the present invention also provides a high temperature early warning device, comprising: an acquisition module, a calculation module, a determination module, and an alarm module;
  • the acquiring module is configured to obtain a value of a parameter that affects temperature of the equipment room;
  • the calculating module is configured to convert the obtained value by a back propagation neural network model to calculate a predicted temperature value of the equipment room device;
  • the determining module is configured to determine whether the predicted temperature value is greater than a preset temperature value
  • the alarm module is configured to issue a high temperature alarm when the predicted temperature value is greater than a preset temperature value.
  • the acquiring module includes a temperature sensor, a humidity sensor, a current sensor, and a resistance sensor;
  • the temperature sensor is configured to acquire a temperature in the equipment room and a temperature value of the device
  • the humidity sensor is configured to acquire humidity in the equipment room and a humidity value of the device
  • the current sensor is configured to acquire a current value of the device
  • the resistance sensor is configured to acquire a resistance value of the device.
  • the back propagation neural network model includes an input layer, an intermediate layer, and an output layer.
  • the high temperature early warning device further includes an adjustment module
  • the obtaining module is further configured to extract, from the equipment room history table, a record with a temperature value of the equipment of the organic room; the record includes a value of a parameter that affects the temperature of the equipment room, and an actual temperature of the corresponding equipment of the equipment room. value;
  • the calculation module is further configured to initialize a weight and a threshold of the input layer and the intermediate layer, and a value of a parameter that affects a temperature of the equipment room, and the input layer and the intermediate layer Weight and threshold calculate the predicted temperature value of the equipment room;
  • the determining module is further configured to determine whether an error of the predicted temperature value and an actual temperature value in the corresponding record is within a preset threshold range;
  • the adjusting module is configured to update a weight of the input layer and the intermediate layer by a gradient descent method when an error of the predicted temperature value and an actual temperature value in the corresponding record is not within a preset threshold range Threshold.
  • the computing module is specifically configured to:
  • the output value of the output layer is calculated by using an output value of the intermediate layer as an input value of the output layer, and the predicted temperature value is obtained by a normalization algorithm.
  • the data of the parameters affecting the temperature of the equipment room are monitored in real time, and the predicted temperature value of the equipment room in the current environment is calculated by the back propagation neural network model to determine whether the equipment room equipment will be interrupted by high temperature.
  • the high temperature interruption warning of the equipment room equipment effectively avoids the interruption caused by the high temperature of the equipment room, thereby effectively reducing the failure rate of the equipment room and improving the operational efficiency of the Internet of Things.
  • FIG. 1 is a flow chart of an embodiment of a high temperature early warning method according to the present invention.
  • FIG. 2 is a flow chart of steps of training a back propagation neural network model in the high temperature early warning method of the present invention
  • FIG. 3 is a flow chart of the steps of FIG. 2 according to the values of the parameters affecting the temperature of the equipment room, and the weights and thresholds of the input layer and the intermediate layer to calculate the predicted temperature values of the equipment room Figure
  • FIG. 4 is a schematic block diagram of an embodiment of a high temperature early warning device according to the present invention.
  • FIG. 5 is a schematic block diagram of another embodiment of the high temperature early warning device of the present invention.
  • a high temperature early warning method includes the following steps:
  • Step S110 Obtain a value of a parameter that affects the temperature of the equipment room.
  • temperature sensors, humidity sensors, current sensors, and resistance sensors are installed in the equipment room and on the equipment to obtain temperature and humidity in the equipment room, and parameters such as temperature, humidity, current, and resistance of the equipment that affect the temperature of the equipment. The value.
  • Step S120 The obtained value is converted by a backpropagation (BP) neural network model to calculate a predicted temperature value of the equipment room.
  • BP backpropagation
  • the back propagation neural network model includes an input layer, an intermediate layer, and an output layer; wherein the input layer is configured to receive a temperature sensor, a humidity sensor, a current sensor, And the temperature, humidity, current, and resistance values detected by the resistance sensor.
  • the intermediate layer is used for calculations.
  • the output layer is used to output a predicted value.
  • the back propagation neural network model needs to be trained to obtain the required weights and thresholds of the input layer and the intermediate layer, thereby ensuring the The accuracy of the predicted value of the output layer output.
  • Step S130 Determine whether the predicted temperature value is greater than a preset temperature value; if yes, execute step S140, if otherwise, perform step S110.
  • Step S140 issuing a high temperature warning.
  • comparing the predicted temperature with a preset temperature value for example, a temperature at which the high temperature of the equipment room is interrupted
  • a preset temperature value for example, a temperature at which the high temperature of the equipment room is interrupted
  • the data of the parameter affecting the temperature of the equipment room is monitored in real time, and the predicted temperature value of the equipment room in the current environment is calculated by using the back propagation neural network model to determine whether the equipment room equipment is interrupted by high temperature.
  • the high-temperature interruption warning of the equipment room is realized, which effectively avoids the interruption caused by the high temperature of the equipment room, thereby effectively reducing the failure rate of the equipment room and improving the operational efficiency of the Internet of Things.
  • FIG. 2 is a flow chart of the steps of the back propagation neural network model training in the high temperature early warning method of the present invention.
  • Step S210 Initialize weights and thresholds of the input layer and the intermediate layer.
  • the weight v i and the threshold ⁇ i of the input layer, and the weight w j and the threshold b of the intermediate layer are initialized by a random number.
  • the weights v i and w j range from -1 to 1, and the thresholds ⁇ i and b range from 0 to 1.
  • Step S220 Extracting, from the computer room history record table, a record with a temperature value of the equipment of the organic room; the record includes a value of a parameter that affects the temperature of the equipment room, and an actual temperature value of the corresponding equipment of the equipment room.
  • the data is classified and filtered, and all the records with the temperature values of the organic room equipment are cyclically extracted.
  • the record includes the temperature value of the equipment room and the corresponding value of a group of other monitoring quantities that affect the temperature (for example, the temperature value and humidity value in the equipment room, as well as the temperature value, humidity value, current value and resistance value of the equipment).
  • high-temperature prediction is required for multiple computer rooms, it is necessary to extract corresponding data from multiple history tables as input for automatic cycle training, and the type and number of monitoring types of the input model need to be the same.
  • the trained model is applicable to the equipment room; if the historical data of multiple equipment rooms is used as the training sample, the trained model is common to multiple computer rooms.
  • Step S230 Calculate a predicted temperature value of the equipment room according to the value of the parameter that affects the temperature of the equipment room, and the weight and threshold of the input layer and the intermediate layer.
  • the value of the parameter that affects the temperature of the equipment room, and the weight and threshold of the input layer and the intermediate layer are used to calculate the equipment room.
  • the steps of predicting the temperature value include:
  • Step S231 using the value of the parameter that affects the temperature of the equipment room as an input value of the input layer, and calculating an output value of the input layer according to the weight of the input layer and a threshold.
  • Step S232 using the output value of the input layer as an input value of the intermediate layer, and calculating an output value of the intermediate layer according to the weight of the intermediate layer and a threshold.
  • Step S233 calculating an output value of the output layer by using an output value of the intermediate layer as an input value of the output layer, and obtaining the predicted temperature value by a normalization algorithm.
  • the value of H j is used as an input value of the output layer, according to a formula
  • the output value O of the output layer is calculated.
  • the predicted temperature value is then derived by a normalization algorithm.
  • Step S240 Determine whether an error of the predicted temperature value and an actual temperature value in the corresponding record is within a preset threshold range. If yes, step S250 is performed; if not, step S260 is performed.
  • Step S250 saving the weights and thresholds of the intermediate layer and the output layer, and then ending the training.
  • Step S260 updating the weights and thresholds of the input layer and the intermediate layer by a gradient descent method. It is then retrained based on the updated weights and thresholds.
  • the main principle of the gradient descent method is to deviate the weight and threshold of the intermediate layer and the output layer, so that the error function is minimized.
  • the derivation process is omitted, and the formula for adjusting the weight and the threshold is obtained as follows:
  • the weight of the input layer is:
  • the weight of the middle layer is:
  • the threshold of the middle layer is the threshold of the middle layer
  • the technical solution of the embodiment can effectively ensure the accuracy of the prediction by training the back propagation neural network model to determine the weight and threshold of the input layer and the intermediate layer.
  • a high temperature early warning device includes an acquisition module 310, a calculation module 320, a determination module 330, and an alarm module 340. among them,
  • the obtaining module 310 is configured to obtain a value of a parameter that affects the temperature of the equipment room.
  • temperature sensors, humidity sensors, current sensors, and resistance sensors are installed in the equipment room and on the equipment to obtain temperature and humidity in the equipment room, and parameters such as temperature, humidity, current, and resistance of the equipment that affect the temperature of the equipment. The value.
  • the calculating module 320 is configured to convert the obtained value by a back propagation neural network model to calculate a predicted temperature value of the equipment room.
  • the back propagation neural network model includes an input layer, an intermediate layer, and an output layer; wherein the input layer is configured to receive a temperature sensor, a humidity sensor, a current sensor, And the temperature, humidity, current, and resistance values detected by the resistance sensor.
  • the intermediate layer is used for calculations.
  • the output layer is used to output a predicted value.
  • the back propagation neural network model needs to be trained to obtain the required weights and thresholds of the input layer and the intermediate layer, thereby ensuring the The accuracy of the predicted value of the output layer output.
  • the determining module 330 is configured to determine whether the predicted temperature value is greater than a preset temperature value.
  • the alarm module 340 is configured to issue a high temperature alarm when the predicted temperature value is greater than a preset temperature value.
  • comparing the predicted temperature with a preset temperature value for example, a temperature at which the high temperature of the equipment room is interrupted
  • a preset temperature value for example, a temperature at which the high temperature of the equipment room is interrupted
  • the data of the parameter affecting the temperature of the equipment room is monitored in real time, and the predicted temperature value of the equipment room in the current environment is calculated by using the back propagation neural network model to determine whether the equipment room equipment is interrupted by high temperature. Realize the high temperature interruption warning of the equipment room, effectively avoiding The situation in which the temperature in the equipment room is too high causes an interruption, which effectively reduces the failure rate of the equipment room and improves the operational efficiency of the Internet of Things.
  • FIG. 5 is a schematic block diagram of another embodiment of the high temperature early warning device of the present invention.
  • the high temperature early warning device further includes an adjustment module 350;
  • the back propagation neural network model includes an input layer, an intermediate layer, and an output layer. among them,
  • the obtaining module 310 is further configured to extract, from the equipment room history table, a record with a temperature value of the equipment of the organic room; the record includes a value of a parameter that affects the temperature of the equipment room, and the actual value of the corresponding equipment of the equipment room. Temperature value.
  • the data is classified and filtered, and all the records with the temperature values of the organic room equipment are cyclically extracted.
  • the record includes the temperature value of the equipment room and the corresponding value of a group of other monitoring quantities that affect the temperature (for example, the temperature value and humidity value in the equipment room, as well as the temperature value, humidity value, current value and resistance value of the equipment).
  • high-temperature prediction is required for multiple computer rooms, it is necessary to extract corresponding data from multiple history tables as input for automatic cycle training, and the type and number of monitoring types of the input model need to be the same.
  • the trained model is applicable to the equipment room; if the historical data of multiple equipment rooms is used as the training sample, the trained model is common to multiple computer rooms.
  • the calculation module 320 is further configured to initialize a weight and a threshold of the input layer and the intermediate layer, and a value of a parameter that affects a temperature of the equipment room, and the input layer and the middle
  • the weight and threshold of the layer calculate the predicted temperature value of the equipment room.
  • the weight v i and the threshold ⁇ i of the input layer, and the weight w j and the threshold b of the intermediate layer are initialized by a random number.
  • the weights v i and w j range from -1 to 1, and the thresholds ⁇ i and b range from 0 to 1.
  • the calculating module 320 is specifically configured to: use the value of the parameter that affects the temperature of the equipment room as an input value of the input layer, and calculate the input layer according to the weight and the threshold of the input layer.
  • Output value of the input layer, the output value of the intermediate layer is used as an input value of the intermediate layer, and an output value of the intermediate layer is calculated according to a weight value and a threshold value of the intermediate layer;
  • the output value of the output layer is calculated, and the predicted temperature value is obtained by a normalization algorithm.
  • the determining module 330 is further configured to determine whether an error of the predicted temperature value and an actual temperature value in the corresponding record is within a preset threshold range.
  • the adjusting module 350 is configured to update the weights of the input layer and the intermediate layer by a gradient descent method when an error of the predicted temperature value and an actual temperature value in the corresponding record is not within a preset threshold range And threshold.
  • the main principle of the gradient descent method is to deviate the weight and threshold of the intermediate layer and the output layer, so that the error function is minimized.
  • the derivation process is omitted, and the formula for adjusting the weight and the threshold is obtained as follows:
  • the weight of the input layer is:
  • the weight of the middle layer is:
  • the threshold of the middle layer is the threshold of the middle layer
  • the technical solution of the embodiment can effectively ensure the accuracy of the prediction by training the back propagation neural network model to determine the weight and threshold of the input layer and the intermediate layer.
  • the foregoing embodiment method can be implemented by means of software plus a necessary general hardware platform, and can also be implemented by hardware, but in many cases, the former is A better implementation.
  • the technical solution of the present invention which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium (such as ROM/RAM, disk,
  • the optical disc includes a number of instructions for causing a terminal device (which may be a cell phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the methods described in various embodiments of the present invention.
  • the data of the parameters affecting the temperature of the equipment room are monitored in real time, and the predicted temperature value of the equipment room in the current environment is calculated by the back propagation neural network model to determine whether the equipment room equipment will be interrupted by high temperature.
  • the high temperature interruption warning of the equipment room equipment effectively avoids the interruption caused by the high temperature of the equipment room, thereby effectively reducing the failure rate of the equipment room and improving the operational efficiency of the Internet of Things.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Air Conditioning Control Device (AREA)
  • Testing And Monitoring For Control Systems (AREA)

Abstract

本发明公开了一种高温预警方法与装置，属于高温预警技术领域。所述方法包括：获取对机房设备的温度造成影响的参数的数值；将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值；判断所述预测温度值是否大于预设温度值；如果是，则发出高温预警。本发明的技术方案，通过实时监测影响机房设备的温度的参数的数据，并通过反向传播神经网络模型计算出在当前环境下机房设备的预测温度值，以判断机房设备是否会高温中断，实现了机房设备的高温中断预警，有效避免由于机房温度过高而导致中断的状况发生，进而有效降低了机房设备的故障率、提高了物联网的运营效率。

Description

高温预警方法与装置 技术领域
本发明涉及高温预警技术领域，尤其涉及一种高温预警方法与装置。
背景技术
随着物联物技术的快速发展，互联网数据中心应运而生。在线上线下业务的快速推动下，机房建设规模也越来越大，负责后台计算的核心设备任务越来越重，如何保障核心设备的安全运行、降低设备故障率将变得重中之重。
目前针对机房核心设备故障处理方式主要还是停留在故障产生、告警到处理的阶段。在目前的大数据时代，核心设备若发生故障，将给生产带来巨大的损失，而高温是直接导致设备发生故障的核心元素之一。
综上所述，如何及时对设备发生高温进行预警，降低设备的故障率、提高运营效率成为一个亟待解决的技术问题。
发明内容
有鉴于此，本发明的目的在于提供一种高温预警方法与装置，旨在实现设备高温预警。
为实现上述目的，本发明提供一种高温预警方法，包括如下步骤：
获取对机房设备的温度造成影响的参数的数值；
将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值；
判断所述预测温度值是否大于预设温度值；
如果是，则发出高温预警。
可选的，所述获取会对机房设备的温度造成影响的参数的数值的步骤包括：
通过温度传感器获取机房内的温度与设备的温度值；
通过湿度传感器获取机房内的湿度与设备的湿度值；
通过电流传感器获取设备的电流值；
通过电阻传感器获取设备的电阻值。
可选的，还包括：
创建所述反向传播神经网络模型，所述反向传播神经网络模型包括输入层、中间层与输出层；
对所述反向传播神经网络模型进行训练。
可选的，所述对所述反向传播神经网络模型进行训练的步骤包括：
初始化所述输入层与所述中间层的权值和阈值；
从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值；
根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值；
判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内；
如果否，则通过梯度下降法更新所述输入层与所述中间层的权值和阈值。
可选的，所述根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值的步骤包括：
以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值，根据所述输入层的权值和阈值计算所述输入层的输出值；
以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值；
以所述中间层的输出值作为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
为了实现上述目的，本发明还提出一种高温预警装置，包括：获取模块、计算模块、判断模块，以及告警模块；
所述获取模块，用于获取对机房设备的温度造成影响的参数的数值；
所述计算模块，用于将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值；
所述判断模块，用于判断所述预测温度值是否大于预设温度值；
所述告警模块，用于在所述预测温度值大于预设温度值时，发出高温告警。
可选的，所述获取模块包括温度传感器、湿度传感器、电流传感器，以及电阻传感器；
所述温度传感器，用于获取机房内的温度与设备的温度值；
所述湿度传感器，用于获取机房内的湿度与设备的湿度值；
所述电流传感器，用于获取设备的电流值；
所述电阻传感器，用于获取设备的电阻值。
可选的，所述反向传播神经网络模型包括输入层、中间层与输出层。
可选的，所述高温预警装置还包括调整模块；
所述获取模块，还用于从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值；
所述计算模块，还用于初始化所述输入层与所述中间层的权值和阈值，并根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值；
所述判断模块，还用于判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内；
所述调整模块，用于在所述预测温度值与对应记录中的实际温度值的误差不在预设的阈值范围内时，通过梯度下降法更新所述输入层与所述中间层的权值和阈值。
可选的，所述计算模块具体用于，
以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值， 根据所述输入层的权值和阈值计算所述输入层的输出值；
以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值；
以所述中间层的输出值作为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
本发明的技术方案，通过实时监测影响机房设备的温度的参数的数据，并通过反向传播神经网络模型计算出在当前环境下机房设备的预测温度值，以判断机房设备是否会高温中断，实现了机房设备的高温中断预警，有效避免由于机房温度过高而导致中断的状况发生，进而有效降低了机房设备的故障率、提高了物联网的运营效率。
附图说明
图1为本发明高温预警方法一实施例的流程图；
图2为本发明高温预警方法中反向传播神经网络模型训练的步骤流程图；
图3为图2中根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值的步骤的流程图；
图4为本发明高温预警装置一实施例的模块示意图；
图5为本发明高温预警装置另一实施例的模块示意图。
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
具体实施方式
为了使本发明所要解决的技术问题、技术方案及工业实用性更加清楚、明白，以下结合附图和实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。
如图1所示，本发明第一实施例提出的一种高温预警方法，包括以下步骤：
步骤S110、获取对机房设备的温度造成影响的参数的数值。
具体的，在机房内和设备上安装温度传感器、湿度传感器、电流传感器，以及电阻传感器，以获取机房内的温度和湿度，以及设备的温度、湿度、电流与电阻等会影响设备的温度的参数的数值。
步骤S120、将获取到的数值通过反向传播(backpropagation，BP)神经网络模型进行换算，以计算出所述机房设备的预测温度值。
具体的，首先需要创建一个反向传播神经网络模型，所述反向传播神经网络模型包括输入层、中间层与输出层；其中，所述输入层用于接受温度传感器、湿度传感器、电流传感器，以及电阻传感器检测到的温度、湿度、电流以及电阻值。所述中间层用于进行计算。所述输出层用于输出预测值。当所述反向传播神经网络模型创建完成后，需要对所述反向传播神经网络模型进行训练，以得到所述输入层与所述中间层的符合要求的权值与阈值，进而保证所述输出层输出的预测值的准确性。
步骤S130、判断所述预测温度值是否大于预设温度值；如果是，则执行步骤S140，如果否则执行步骤S110。
步骤S140、发出高温预警。
具体的，将所述预测温度与预设温度值(比如，机房设备的高温中断的温度)做比较，如果所述预测温度大于所述预设温度值，则判定机房设备可能会高温中断，这时，则发出告警，告警的方式具体可以为声光报警等。而所述预测温度小于或等于所述预设温度值，则判定机房设备不会高温中断，则继续获取对机房设备的温度造成影响的参数的数值。
本实施例的技术方案，通过实时监测影响机房设备的温度的参数的数据，并通过反向传播神经网络模型计算出在当前环境下机房设备的预测温度值，以判断机房设备是否会高温中断，实现了机房设备的高温中断预警，有效避免由于机房温度过高而导致中断的状况发生，进而有效降低了机房设备的故障率、提高了物联网的运营效率。
进一步的，如图2所示，图2为本发明高温预警方法中反向传播神经网络模型训练的步骤流程图。
在本实施例中，所述对所述反向传播神经网络模型进行训练的步骤包括：
步骤S210、初始化所述输入层与所述中间层的权值和阈值。
具体的，通过随机数初始化所述输入层的权值vi和阈值θi，以及所述中间层的权值wj和阈值b。权值vi与wj的范围在-1～1之间，阈值θi与b的范围在0～1之间。并初始化学习率η；学习率的范围在0～1之间。其中，i＝0、1、2......n-1；j＝0、1、2......m-1；n为所述输入层节点个数，m为所述中间层节点个数，而
Figure PCTCN2017078396-appb-000001
且1≤a≤10。
步骤S220、从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值。
具体的，从机房历史记录表中，对数据进行分类筛选，循环提取所有带有机房设备温度值的记录。该记录内包括机房设备温度值以及相应的影响该温度的一组其他监控量的值(比如，机房内的温度值和湿度值，以及设备的温度值、湿度值、电流值与电阻值)。若需要对多个机房进行高温预测，则需要从多个历史记录表中提取相应的数据作为输入进行自动循环训练，且输入模型的监控量类型与数目需要相同。若采用某一机房内的历史数据作为训练样本，则所训练出的模型适用于该机房；若采用多个机房的历史数据作为训练样本，则训练的模型通用于多个机房。
值得一提的是，上述步骤S210与步骤S220之间的顺序不分先后。
步骤S230、根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值。
如图3所示，在本实施例中，所述根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值的步骤包括：
步骤S231、以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值，根据所述输入层的权值和阈值计算所述输入层的输出值。
具体的，选取样本里的一条数据作为所述输入层的输入值，比如，第i个监 控量的值xi(也即，温度传感器、湿度传感器、电流传感器，以及电阻传感器监测到的值)；然后根据公式
Figure PCTCN2017078396-appb-000002
计算所述输入层的输出值，其中，i＝0、1、2......n-1；j＝0、1、2......m-1；vji为所述输入层的第i个节点到所述中间层的第j个节点的权值，θj为所述输入层的第j个节点的阈值。
步骤S232、以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值。
具体的，以yj的值作为所述中间层的输入值，根据公式
Figure PCTCN2017078396-appb-000003
计算所述中间层的输出值Hj＝f(yj)；其中，e为输出结果与真实值之间的差，计算公式为：e＝d-O；j＝0、1、2......m-1。
步骤S233、以所述中间层的输出值作为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
具体的，以Hj的值作为所述输出层的输入值，根据公式
Figure PCTCN2017078396-appb-000004
计算所述输出层的输出值O。然后通过归一算法得出所述预测温度值。
步骤S240、判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内。如果是，则执行步骤S250；如果否，则执行步骤S260。
具体的，根据设定的误差公式
Figure PCTCN2017078396-appb-000005
计算预测温度值与对应记录中的实际温度值的误差；其中，m＝1；d是历史高温告警下的温度通过归一算法处理后的值，O为预测值。
步骤S250、保存所述中间层和所述输出层的权值和阈值，然后结束训练。
步骤S260、通过梯度下降法更新所述输入层与所述中间层的权值和阈值。然后根据更新后的权值与阈值重新训练。
具体的，梯度下降法的主要原理是对中间层和输出层的权值和阈值求偏导,使得误差函数最小，这里略去求导过程，得到调整权值和阈值的公式如下：
输入层的权值为：
Figure PCTCN2017078396-appb-000006
输入层的阈值：
Figure PCTCN2017078396-appb-000007
中间层的权值为：
Figure PCTCN2017078396-appb-000008
中间层的阈值：
Figure PCTCN2017078396-appb-000009
本实施例的技术方案，通过对反向传播神经网络模型进行训练，以确定输入层与中间层的权值与阈值，可以有效保证预测的精确度。
如图4所示，本发明第一实施例提出的一种高温预警装置，包括获取模块310、计算模块320、判断模块330，以及告警模块340。其中，
所述获取模块310，用于获取对机房设备的温度造成影响的参数的数值。
具体的，在机房内和设备上安装温度传感器、湿度传感器、电流传感器，以及电阻传感器，以获取机房内的温度和湿度，以及设备的温度、湿度、电流与电阻等会影响设备的温度的参数的数值。
所述计算模块320，用于将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值。
具体的，首先需要创建一个反向传播神经网络模型，所述反向传播神经网络模型包括输入层、中间层与输出层；其中，所述输入层用于接受温度传感器、湿度传感器、电流传感器，以及电阻传感器检测到的温度、湿度、电流以及电阻值。所述中间层用于进行计算。所述输出层用于输出预测值。当所述反向传播神经网络模型创建完成后，需要对所述反向传播神经网络模型进行训练，以得到所述输入层与所述中间层的符合要求的权值与阈值，进而保证所述输出层输出的预测值的准确性。
所述判断模块330，用于判断所述预测温度值是否大于预设温度值。
所述告警模块340，用于在所述预测温度值大于预设温度值时，发出高温告警。
具体的，将所述预测温度与预设温度值(比如，机房设备的高温中断的温度)做比较，如果所述预测温度大于所述预设温度值，则判定机房设备可能会高温中断，这时，则发出告警，告警的方式具体可以为声光报警等。而所述预测温度小于或等于所述预设温度值，则判定机房设备不会高温中断，则继续获取对机房设备的温度造成影响的参数的数值。
本实施例的技术方案，通过实时监测影响机房设备的温度的参数的数据，并通过反向传播神经网络模型计算出在当前环境下机房设备的预测温度值，以判断机房设备是否会高温中断，实现了机房设备的高温中断预警，有效避免由 于机房温度过高而导致中断的状况发生，进而有效降低了机房设备的故障率、提高了物联网的运营效率。
进一步的，如图5所示，图5为本发明高温预警装置另一实施例的模块示意图。
在本实施例中，所述高温预警装置还包括调整模块350；所述反向传播神经网络模型包括输入层、中间层与输出层。其中，
所述获取模块310，还用于从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值。
具体的，从机房历史记录表中，对数据进行分类筛选，循环提取所有带有机房设备温度值的记录。该记录内包括机房设备温度值以及相应的影响该温度的一组其他监控量的值(比如，机房内的温度值和湿度值，以及设备的温度值、湿度值、电流值与电阻值)。若需要对多个机房进行高温预测，则需要从多个历史记录表中提取相应的数据作为输入进行自动循环训练，且输入模型的监控量类型与数目需要相同。若采用某一机房内的历史数据作为训练样本，则所训练出的模型适用于该机房；若采用多个机房的历史数据作为训练样本，则训练的模型通用于多个机房。
所述计算模块320，还用于初始化所述输入层与所述中间层的权值和阈值，并根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值。
具体的，通过随机数初始化所述输入层的权值vi和阈值θi，以及所述中间层的权值wj和阈值b。权值vi与wj的范围在-1～1之间，阈值θi与b的范围在0～1之间。并初始化学习率η；学习率的范围在0～1之间。其中，i＝0、1、2......n-1；j＝0、1、2......m-1；n为所述输入层节点个数，m为所述中间层节点个数，而
Figure PCTCN2017078396-appb-000010
且1≤a≤10。
进一步的，所述计算模块320具体用于，以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值，根据所述输入层的权值和阈值计算所述输入层的输出值；以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值；以所述中间层的输出值作 为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
具体的，选取样本里的一条数据作为所述输入层的输入值，比如，第i个监控量的值xi(也即，温度传感器、湿度传感器、电流传感器，以及电阻传感器监测到的值)；然后根据公式
Figure PCTCN2017078396-appb-000011
计算所述输入层的输出值，其中，i＝0、1、2......n-1；j＝0、1、2......m-1；vji为所述输入层的第i个节点到所述中间层的第j个节点的权值，θj为所述输入层的第j个节点的阈值。以yj的值作为所述中间层的输入值，根据公式
Figure PCTCN2017078396-appb-000012
计算所述中间层的输出值Hj＝f(yj)；其中，e为输出结果与真实值之间的差，计算公式为：e＝d-O；j＝0、1、2......m-1。以Hj的值作为所述输出层的输入值，根据公式
Figure PCTCN2017078396-appb-000013
计算所述输出层的输出值O。然后通过归一算法得出所述预测温度值。
所述判断模块330，还用于判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内。
具体的，根据设定的误差公式
Figure PCTCN2017078396-appb-000014
计算预测温度值与对应记录中的实际温度值的误差；其中，m＝1；d是历史高温告警下的温度通过归一算法处理后的值，O为预测值。
所述调整模块350，用于在所述预测温度值与对应记录中的实际温度值的误差不在预设的阈值范围内时，通过梯度下降法更新所述输入层与所述中间层的权值和阈值。
具体的，梯度下降法的主要原理是对中间层和输出层的权值和阈值求偏导,使得误差函数最小，这里略去求导过程，得到调整权值和阈值的公式如下：
输入层的权值为：
Figure PCTCN2017078396-appb-000015
输入层的阈值：
Figure PCTCN2017078396-appb-000016
中间层的权值为：
Figure PCTCN2017078396-appb-000017
中间层的阈值：
Figure PCTCN2017078396-appb-000018
本实施例的技术方案，通过对反向传播神经网络模型进行训练，以确定输入层与中间层的权值与阈值，可以有效保证预测的精确度。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者装置不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者装置所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者装置中还存在另外的相同要素。
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件来实现，但很多情况下前者是更佳的实施方式。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本发明各个实施例所述的方法。
以上参照附图说明了本发明的优选实施例，并非因此局限本发明的权利范围。上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。另外，虽然在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于此处的顺序执行所示出或描述的步骤。
本领域技术人员不脱离本发明的范围和实质，可以有多种变型方案实现本发明，比如作为一个实施例的特征可用于另一实施例而得到又一实施例。凡在运用本发明的技术构思之内所作的任何修改、等同替换和改进，均应在本发明的权利范围之内。
工业实用性说明
本发明的技术方案，通过实时监测影响机房设备的温度的参数的数据，并通过反向传播神经网络模型计算出在当前环境下机房设备的预测温度值，以判断机房设备是否会高温中断，实现了机房设备的高温中断预警，有效避免由于机房温度过高而导致中断的状况发生，进而有效降低了机房设备的故障率、提高了物联网的运营效率。

Claims (10)

  1. 一种高温预警方法，包括如下步骤：
    获取对机房设备的温度造成影响的参数的数值；
    将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值；
    判断所述预测温度值是否大于预设温度值；
    如果是，则发出高温预警。
  2. 如权利要求1所述的高温预警方法，其中，所述获取会对机房设备的温度造成影响的参数的数值的步骤包括：
    通过温度传感器获取机房内的温度与设备的温度值；
    通过湿度传感器获取机房内的湿度与设备的湿度值；
    通过电流传感器获取设备的电流值；
    通过电阻传感器获取设备的电阻值。
  3. 如权利要求1所述的高温预警方法，其中，还包括：
    创建所述反向传播神经网络模型，所述反向传播神经网络模型包括输入层、中间层与输出层；
    对所述反向传播神经网络模型进行训练。
  4. 如权利要求3所述的高温预警方法，其中，所述对所述反向传播神经网络模型进行训练的步骤包括：
    初始化所述输入层与所述中间层的权值和阈值；
    从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值；
    根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所 述中间层的权值和阈值计算出所述机房设备的预测温度值；
    判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内；
    如果否，则通过梯度下降法更新所述输入层与所述中间层的权值和阈值。
  5. 如权利要求4所述的高温预警方法，其中，所述根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值的步骤包括：
    以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值，根据所述输入层的权值和阈值计算所述输入层的输出值；
    以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值；
    以所述中间层的输出值作为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
  6. 一种高温预警装置，包括：获取模块、计算模块、判断模块，以及告警模块；
    所述获取模块，用于获取对机房设备的温度造成影响的参数的数值；
    所述计算模块，用于将获取到的数值通过反向传播神经网络模型进行换算，以计算出所述机房设备的预测温度值；
    所述判断模块，用于判断所述预测温度值是否大于预设温度值；
    所述告警模块，用于在所述预测温度值大于预设温度值时，发出高温告警。
  7. 如权利要求6所述的高温预警装置，其中，所述获取模块包括温度传感器、湿度传感器、电流传感器，以及电阻传感器；
    所述温度传感器，用于获取机房内的温度与设备的温度值；
    所述湿度传感器，用于获取机房内的湿度与设备的湿度值；
    所述电流传感器，用于获取设备的电流值；
    所述电阻传感器，用于获取设备的电阻值。
  8. 如权利要求6所述的高温预警装置，其中，所述反向传播神经网络模型包括输入层、中间层与输出层。
  9. 如权利要求8所述的高温预警装置，其中，还包括调整模块；
    所述获取模块，还用于从机房历史记录表中提取带有机房设备温度值的记录；所述记录中包含有对机房设备的温度造成影响的参数的数值，以及对应的机房设备的实际温度值；
    所述计算模块，还用于初始化所述输入层与所述中间层的权值和阈值，并根据所述对机房设备的温度造成影响的参数的数值，以及所述输入层与所述中间层的权值和阈值计算出所述机房设备的预测温度值；
    所述判断模块，还用于判断所述预测温度值与对应记录中的实际温度值的误差是否在预设的阈值范围内；
    所述调整模块，用于在所述预测温度值与对应记录中的实际温度值的误差不在预设的阈值范围内时，通过梯度下降法更新所述输入层与所述中间层的权值和阈值。
  10. 如权利要求8所述的高温预警装置，其中，所述计算模块具体用于，
    以所述对机房设备的温度造成影响的参数的数值作为所述输入层的输入值，根据所述输入层的权值和阈值计算所述输入层的输出值；
    以所述输入层的输出值作为所述中间层的输入值，根据所述中间层的权值和阈值计算所述中间层的输出值；
    以所述中间层的输出值作为所述输出层的输入值，计算所述输出层的输出值，并通过归一算法得出所述预测温度值。
PCT/CN2017/078396 2017-03-28 2017-03-28 高温预警方法与装置 WO2018176213A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/078396 WO2018176213A1 (zh) 2017-03-28 2017-03-28 高温预警方法与装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/078396 WO2018176213A1 (zh) 2017-03-28 2017-03-28 高温预警方法与装置

Publications (1)

Publication Number Publication Date
WO2018176213A1 true WO2018176213A1 (zh) 2018-10-04

Family

ID=63673969

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/078396 WO2018176213A1 (zh) 2017-03-28 2017-03-28 高温预警方法与装置

Country Status (1)

Country Link
WO (1) WO2018176213A1 (zh)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102842909A (zh) * 2012-09-12 2012-12-26 湖南大学 一种电力电子混合系统控制方法
CN104359565A (zh) * 2014-10-17 2015-02-18 中国农业大学 一种冷链运输温度监测与预警的方法及系统
CN104679195A (zh) * 2015-03-16 2015-06-03 芜湖凯博实业股份有限公司 一种电脑散热冷却系统及其控制方法

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102842909A (zh) * 2012-09-12 2012-12-26 湖南大学 一种电力电子混合系统控制方法
CN104359565A (zh) * 2014-10-17 2015-02-18 中国农业大学 一种冷链运输温度监测与预警的方法及系统
CN104679195A (zh) * 2015-03-16 2015-06-03 芜湖凯博实业股份有限公司 一种电脑散热冷却系统及其控制方法

Similar Documents

Publication Publication Date Title
CN108197845B (zh) 一种基于深度学习模型lstm的交易指标异常的监测方法
Meng et al. Rating the crisis of online public opinion using a multi-level index system
WO2022047658A1 (zh) 日志异常检测系统
CN109523021A (zh) 一种基于长短时记忆网络的动态网络结构预测方法
CN113572625B (zh) 故障预警方法、预警装置、设备及计算机介质
CN110909125B (zh) 推文级社会媒体谣言检测方法
CN110210660B (zh) 一种超短期风速预测方法
CN109002781B (zh) 一种储能变流器故障预测方法
CN113723010A (zh) 一种基于lstm温度-位移相关模型的桥梁损伤预警方法
CN103197983A (zh) 基于概率图模型的服务组件可靠性在线时间序列预测方法
CN108833376A (zh) 面向软件定义网络的DoS攻击检测方法
CN114723285A (zh) 一种电网设备安全性评估预测方法
CN112363896A (zh) 日志异常检测系统
CN109558298A (zh) 基于深度学习模型的告警执行频率优化方法及相关设备
CN115329799A (zh) 桥梁安全状态监测方法、装置、计算机设备和存储介质
CN114401145A (zh) 一种网络流量检测系统及方法
CN114266602A (zh) 电力物联网多源数据融合的深度学习电价预测方法和装置
WO2018176213A1 (zh) 高温预警方法与装置
CN109635008B (zh) 一种基于机器学习的设备故障检测方法
CN113151842B (zh) 风光互补电解水制氢的转化效率的确定方法和确定装置
Wang et al. Research of the early warning analysis of crop diseases and insect pests
CN112529637B (zh) 基于情景感知的服务需求动态预测方法及系统
CN110929800B (zh) 一种基于sax算法的商业体异常用电检测方法
Yang et al. Root Cause Location Based on Prophet and Kernel Density Estimation
CN112816211A (zh) 一种基于pso-bp算法的带式输送机故障诊断的方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17904030

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17904030

Country of ref document: EP

Kind code of ref document: A1