UPS equipment performance degradation prediction method and system based on deep learning

By constructing a UPS device performance decay prediction model based on deep learning, the problem of insufficient data processing and feature extraction in the existing technology is solved, more accurate performance decay prediction and intelligent maintenance suggestions are achieved, and the stability of equipment operation is improved and maintenance efficiency.

CN118965212APending Publication Date: 2024-11-15STATE GRID JIBEI ELECTRIC POWER CO LTD TANGSHAN POWER SUPPLY CO +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410908407.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-08
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing technology lacks advanced algorithms and models in the performance decay prediction of UPS equipment, and cannot effectively process and analyze large amounts of data, resulting in insufficient accuracy and reliability of the prediction model, unable to provide intelligent decision-making support, affecting the normal operation and maintenance efficiency of the equipment. .

Method used

Using a deep learning-based method, the equipment information and historical operation data of UPS devices are collected, and the convolutional neural network-long and short-term memory model (CNN-LSTM), recurrent neural network model (RNN) and long-term short-term The performance decay prediction model combined with memory network model (LSTM) predicts the degree of performance decay of the device in real time and provides maintenance measures recommendations based on the prediction results.

Benefits of technology

It improves the accuracy and reliability of UPS equipment performance decay prediction, provides users with maintenance reminders in advance, avoids equipment failures, and optimizes maintenance work to the greatest extent.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118965212A_ABST
    Figure CN118965212A_ABST
Patent Text Reader

Abstract

According to the UPS equipment performance degradation prediction method and system based on deep learning provided by the invention, equipment information and historical operation data of UPS equipment are acquired through a sensor, the acquired historical operation data are cleaned, and high-quality data input is provided for feature extraction and model training. Features related to UPS equipment performance degradation are extracted from the preprocessed data, a UPS equipment performance degradation prediction model is constructed, and the model comprises a convolutional neural network-long and short-term memory model CNN-LSTM, a recurrent neural network model RNN and a long and short-term memory network model LSTM. The constructed model is trained; and inputting the real-time operation data of the UPS equipment into the trained UPS equipment performance degradation prediction model, and predicting the performance degradation degree of the UPS equipment in real time. Finally, whether the final prediction result is normal or not is judged, if not, a user is reminded to maintain the UPS equipment, equipment faults are avoided, and therefore the reliability and stability of the UPS equipment are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of UPS equipment performance degradation prediction, and in particular to a UPS equipment performance degradation prediction method based on deep learning. Background Art

[0002] UPS, which stands for Uninterrupted Power Supply, is a system device that connects lead-acid maintenance-free batteries and other batteries to the host, and converts the DC power in the battery into AC power through module circuits such as the host inverter. UPS equipment is mainly used to provide a stable and uninterrupted power supply to a single computer, a computer network system or other power electronic equipment such as solenoid valves, pressure transmitters, etc.

[0003] A UPS power shell heat resistance performance testing device proposed by Chinese invention patent application number: CN202210165124.X is used to test the thermal performance of the power shell, and the thermal performance testing device includes a sealing member for sealing the power shell, a heating component installed on the sealing member, and a detection mechanism for feeding back the test results. The present invention uses the sealing member and the power shell to form a closed shell, simulating the actual use environment of the power shell, so that the detected heat resistance results are more realistic, and secondly, by setting a fan to circulate the gas in the closed shell, the heat exchange between high temperature and low temperature is accelerated, so that the temperature in the power shell is more uniform; the detection unit is pushed back and forth by the No. 2 cylinder, so that the ball squeezes the softened inner wall of the power shell, thereby causing it to deform and dent, and multi-point detection is performed on the two side walls of the power shell at the same time in the same detection environment.

[0004] The existing technology only collects part of the operating data, such as focusing only on conventional parameters such as voltage, input current and output current, while ignoring features closely related to performance degradation such as temperature and load change rate. This may result in the prediction model being unable to fully understand the performance status of the UPS device. Due to the lack of advanced algorithms and models, the existing technology cannot effectively process and analyze a large amount of data, and it is difficult to extract key features related to performance degradation. This limits the accuracy and reliability of the prediction model. Due to the limitations of data processing and feature extraction, the prediction model of the existing technology cannot accurately predict the performance degradation of the UPS device. This leads to untimely or excessive maintenance measures, affecting the normal operation of the equipment. The existing technology cannot provide comprehensive intelligent decision support, such as early warning, fault diagnosis and countermeasures. This increases the burden on operation and maintenance personnel and may lead to untimely and inefficient maintenance work. The existing technology does not fully consider the diversity and variability of the operating environment of the UPS device, resulting in unstable prediction performance of the prediction model under different environmental conditions. Therefore, we make improvements to this and propose a UPS device performance degradation prediction method based on deep learning.

[0005] Chinese invention patent application number: CN202310579934.4 discloses a method, device and storage medium for evaluating the health status of an uninterruptible power supply (UPS), including obtaining multiple sets of capacity decay cycle data of the UPS, the capacity decay cycle data including floating charge condition data and working condition data; obtaining the full life cycle data of the UPS based on multiple sets of the capacity decay cycle data; training a model based on the full life cycle data to obtain a UPS capacity decay evaluation model; and obtaining the health status evaluation value of the UPS through the evaluation model according to the current monitoring data of the UPS. This patent uses less prediction data and a single prediction model, so the prediction effect is poor. Summary of the invention

[0006] In order to address the deficiencies in the prior art, the present invention provides a UPS equipment performance degradation prediction method and system based on deep learning.

[0007] The present invention adopts the following technical solution.

[0008] Step 1: Collect the equipment information and historical operation data of the uninterruptible power supply UPS device to be predicted;

[0009] Step 2: Preprocess the collected equipment information and historical operation data;

[0010] Step 3: Extract UPS equipment performance degradation related features as a data set;

[0011] Step 4: Build a UPS equipment performance degradation prediction model based on deep learning;

[0012] Step 5: Use the data set to train the UPS equipment performance degradation prediction model;

[0013] Step 6: Collect the equipment information and real-time operation data of the UPS device to be predicted, pre-process the data according to steps 2-3 and extract the relevant features of UPS device performance degradation as the prediction data set, input it into the UPS device performance degradation prediction model trained in step 5, and predict the performance degradation degree of the UPS device in real time.

[0014] The present invention further includes the following preferred embodiments:

[0015] In step 1, the device information of the UPS device includes a UPS rated input voltage, a UPS rated output voltage, a UPS rated input current, and a UPS rated output current;

[0016] The historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of equipment connected to the UPS, and number of equipment connected to the UPS.

[0017] In step 2, One-Hot is used to encode the device type connected to the UPS device; then invalid values ​​in all data are removed, and missing values ​​and abnormal values ​​in all data are cleaned.

[0018] In step 3, the extracted UPS device performance degradation related features include voltage and current deviation characteristic values, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS.

[0019] The calculation formula of the voltage and current deviation characteristic value is as follows:

[0020] P IV =a p V IP +b p V OP +c p I IP +d p I OP

[0021] Among them, P IV is the voltage and current deviation characteristic value, V IP is the difference between the input voltage during operation and the rated input voltage, a p is the input voltage deviation weight, V OP is the difference between the output voltage during operation and the rated output voltage, b p is the output voltage deviation weight, I IP is the difference between the input current during operation and the rated input current, c p is the input current deviation weight, I OP is the difference between the output current during operation and the rated output current, d p is the output current deviation weight; where a p 、b p 、c p ,d p Satisfies the following relationship:

[0022] a P +b P +c P +d P =1.

[0023] Input voltage deviation weight a p Calculated using the following formula:

[0024]

[0025] in,

[0026] max() is the maximum value function, Vmax is the maximum input voltage in the historical operation data, V min is the minimum input voltage in the historical operation data, V IR is the rated input voltage value;

[0027] Output voltage deviation weight b p =0.5-a p .

[0028] Input current deviation weight c p Calculated using the following formula:

[0029]

[0030] in,

[0031] max() is the maximum value function, I max is the maximum input current in the historical operation data, I min is the minimum input current value in the historical operation data, I IR is the rated input current value;

[0032] Output current deviation weight d p =0.5-c p .

[0033] The calculation formula of the load change rate is as follows:

[0034]

[0035] Among them, L R is the load change rate, L V is the load difference within a set time interval, L A is the average load during the time interval;

[0036] The calculation formula of the temperature change rate is as follows:

[0037]

[0038] Among them, T R is the rate of temperature change, T V is the temperature difference of the equipment within the set time interval, and S is the number of sampling points within the set time interval.

[0039] In step 3, the data set also includes labels corresponding to performance degradation related features, and the labels include 0, 1, 2 and 3; among which 0 represents normal, 1 represents the early stage of performance degradation, 2 represents the middle stage of performance degradation, and 3 represents the late stage of performance degradation.

[0040] In step 4, the convolutional neural network-long short-term memory model CNN-LSTM, the recurrent neural network model RNN and the long short-term memory network model LSTM are combined into a UPS equipment performance degradation prediction model;

[0041] The input of the CNN-LSTM model is the load change rate and the type of equipment connected to the UPS.

[0042] Among them, the input of the recurrent neural network is the voltage and current deviation characteristic value and the number of devices connected to the UPS;

[0043] Among them, the input of the long short-term memory network is the temperature change rate.

[0044] In step 6, the prediction result of the convolutional neural network-long short-term memory model is recorded as the first prediction result z r1 , the prediction result of the recurrent neural network model is recorded as the second prediction result z r2 , the prediction result of the long short-term memory network model is recorded as the third prediction result z r3 , the prediction algorithm is used to obtain the final prediction result of the UPS equipment performance degradation prediction model according to the first prediction result, the second prediction result and the third prediction result. The formula of the prediction algorithm is as follows:

[0045]

[0046] Among them, P k is the probability of event k occurring, where the values ​​of k are 0, 1, 2, and 3, corresponding to the four labels in the data set; z k is the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k; e is the base of the natural logarithm;

[0047] The calculation formula of the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k is as follows:

[0048]

[0049] Among them, λ1 is the prediction relative accuracy of the first prediction result, λ2 is the prediction relative accuracy of the second prediction result; λ3 is the prediction relative accuracy of the third prediction result;

[0050] λ i The calculation formula is as follows, i is 1, 2, 3;

[0051]

[0052] Among them, z ri is the i-th prediction result, λ i is the relative accuracy of the prediction of the i-th prediction result, is the mean of the first prediction result, the second prediction result and the third prediction result, m z is the maximum value of the absolute value obtained by subtracting any two of the first prediction result, the second prediction result, and the third prediction result;

[0053] Calculate the probability P of event k occurring through the prediction algorithm k , select the one with the highest probability of occurrence k, and record it as the risk probability P lk , and its corresponding k value is recorded as the final prediction result k L ; In particular, when m z =0, the first prediction result is recorded as the final prediction result, and the risk probability is recorded as 1.

[0054] Determine whether the final prediction result is 0. If so, the UPS is normal; if not, remind the user to adopt the selected maintenance measures for maintenance;

[0055] The selected maintenance measures include conventional maintenance measures and additional maintenance measures; wherein the additional maintenance measures are selected based on the risk probability and the final prediction result.

[0056] When the final prediction result is 1 or 2, the conventional maintenance measures are selected as: enhancing heat dissipation, reducing UPS load, and ensuring that the remaining power percentage of the UPS is higher than 20%;

[0057] When the final prediction result is 2, routine maintenance measures also include dust removal, electrostatic protection, dehumidification, and controlling the internal temperature of the UPS to no more than 45 degrees;

[0058] When the final prediction result is 3, the routine maintenance measure is selected as replacing the UPS device;

[0059] The UPS remaining power percentage calculation formula is divided into charging calculation formula and power consumption calculation formula;

[0060] The calculation formula for charging is as follows:

[0061]

[0062] Among them, s c Indicates the remaining power percentage of the UPS; c0 Indicates the initial remaining power percentage of the UPS; Q i Indicates the amount of electricity obtained by UPS charging; Q n Indicates the maximum amount of electricity that the UPS can store;

[0063] The calculation formula for electricity consumption is as follows:

[0064]

[0065] Among them, Q lIndicates the amount of power lost by the UPS.

[0066] Determine whether the hazard prediction value exceeds the set warning threshold. If so, set additional maintenance measures as follows:

[0067] When k L When it is 1: the internal temperature of UPS is controlled not to exceed 45 degrees;

[0068] When k L When the value is 2: Please ask professionals to maintain the UPS regularly;

[0069] When k L When it is 3: UPS will stop running immediately;

[0070] The warning threshold is set to 0.6;

[0071] The calculation formula of the risk prediction value is:

[0072]

[0073] Among them, F r The risk prediction value.

[0074] The present application also claims protection for a UPS device performance degradation prediction system based on deep learning based on the aforementioned UPS device performance degradation prediction method, characterized in that it includes a data acquisition module, a data preprocessing module, a feature extraction module, a degradation performance prediction module, a maintenance measure selection module and a communication module;

[0075] The data acquisition module is used to collect device information and real-time operation data of the UPS device to be predicted; wherein the device information of the UPS device includes UPS rated input voltage, UPS rated output voltage, UPS rated input current, and UPS rated output current; the historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of device connected to the UPS, and number of devices connected to the UPS;

[0076] The data preprocessing module uses One-Hot to encode the device type connected to the UPS device, and then cleans the abnormal values ​​in all the data and fills the missing values;

[0077] The feature extraction module extracts UPS device performance degradation related features, including voltage and current deviation characteristic values, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS; and uses the extracted UPS device performance degradation related features as a prediction data set;

[0078] The degradation performance prediction module combines the trained UPS equipment performance degradation prediction model and the prediction data set obtained by the feature extraction module to perform degradation performance prediction on the UPS equipment and obtain the final prediction result and risk probability;

[0079] The maintenance measure selection module determines whether the final prediction result obtained by the degradation performance prediction module is normal. If not, it selects maintenance measures according to the risk probability and the final prediction result, and sends information through the communication module to remind the user to adopt the selected maintenance measures for maintenance.

[0080] On the other hand, the present application discloses an electronic device, including a processor and a storage medium; the characteristics are:

[0081] The storage medium is used to store instructions;

[0082] The processor is used to operate according to the instructions to execute the UPS device performance degradation prediction method based on deep learning.

[0083] The present application also discloses a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the UPS device performance degradation prediction method based on deep learning.

[0084] The beneficial effect of the present invention is that, compared with the prior art, the present invention collects equipment information and historical operation data of UPS equipment, the equipment information of UPS equipment includes UPS rated input voltage, UPS rated output voltage, UPS rated input current and UPS rated output current; the historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of equipment connected to UPS, etc., which provides a basic data source for subsequent data processing and analysis. The collected data is preprocessed to ensure the accuracy and reliability of the data. Features related to the performance degradation of UPS equipment are extracted from the preprocessed data, such as voltage and current deviation characteristic values, load change rate, temperature change rate, etc. These features can reflect the performance status and change trend of UPS equipment. Labels are set for the normalized UPS equipment performance degradation related features; and a data set is constructed using the normalized UPS equipment performance degradation related features and corresponding labels.

[0085] The present invention constructs a UPS device performance degradation prediction model, which includes a convolutional neural network-long short-term memory model CNN-LSTM, a recurrent neural network model RNN and a long short-term memory network model LSTM, etc., which can effectively learn and predict the performance degradation of UPS devices. The UPS device performance degradation prediction model is trained using a data set. The trained model is applied to real-time data, and the performance degradation of the UPS device is predicted according to the prediction results of the three deep learning models in the UPS device performance degradation prediction model. Then, it is determined whether the final prediction result is normal. If not, the user is reminded to adopt the selected maintenance measures for maintenance to avoid equipment failure. The prediction effect of the present invention is better than other prediction models, and maintenance reminders can be given to users in advance, which avoids the occurrence of UPS device failures to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 This is a flow chart of the UPS equipment performance degradation prediction method based on deep learning. DETAILED DESCRIPTION

[0087] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0088] Embodiment 1:

[0089] See attached Figure 1 , an embodiment of the present invention provides a UPS device performance degradation prediction method based on deep learning, including data collection, data preprocessing, feature extraction, building a UPS device performance degradation prediction model, training the UPS device performance degradation prediction model, and using the trained UPS device performance degradation prediction model to predict the degree of UPS performance degradation:

[0090] Step 1: Collect the equipment information and historical operation data of the uninterruptible power supply UPS device to be predicted;

[0091] UPS equipment information, including UPS rated input voltage, UPS rated output voltage, UPS rated input current and UPS rated output current, a total of four items of data;

[0092] Historical operation data, including eight items of data: input voltage, output voltage, input current, output current, temperature and load, type of equipment connected to the UPS, and number of equipment connected to the UPS;

[0093] The types of equipment connected to the UPS in historical operation data include computers, servers, network equipment, medical equipment and other critical business equipment;

[0094] Historical operation data is collected and acquired through sensors, with a sampling interval of 1 second;

[0095] Step 2: Perform data preprocessing on the equipment information and historical operation data collected in step 1.

[0096] One-Hot is used to encode the device type connected to the UPS device; then the outliers in all the data are cleaned and missing values ​​are filled, thereby reducing the impact of missing values ​​and outliers on the data, making the data accurate and reliable;

[0097] 2.1 Use One-Hot to encode the device type connected to the UPS device collected in step 1;

[0098] 2.2 After encoding the device type connected to the UPS device in step 2.1, clean the outliers in all data;

[0099] The method for cleaning outliers in any data is as follows:

[0100] when

[0101]

[0102] When

[0103] Among them, Z (h,t) is the data of the hth day at time t; n is the number of data sample points collected in one day, i is a natural number between 1 and n, and Z (h,i) is the data of the hth day and the i-th sampling time;

[0104] when

[0105]

[0106] When

[0107] 2.3 After cleaning all the outliers in the data in step 2.2, fill in the missing values.

[0108] The missing value filling is to fill the missing value with the average value of the corresponding data of the previous day.

[0109] Step 3: Extract the UPS device performance degradation related features, then normalize the extracted UPS device performance degradation related features respectively, and then set labels for the normalized UPS device performance degradation related features; use the normalized UPS device performance degradation related features and corresponding labels to construct a data set.

[0110] 3.1 Extracting features related to UPS equipment performance degradation;

[0111] Feature extraction is an important part of data analysis, especially when monitoring and maintaining critical equipment such as UPS. The performance degradation of UPS equipment may lead to unstable equipment output, or even fail to provide necessary backup power in the event of a power failure, thus affecting the operation of key businesses. Therefore, extracting features related to UPS equipment performance degradation from preprocessed data is crucial for predicting and preventing equipment failures.

[0112] Extract features related to UPS device performance degradation from the preprocessed data. The features include voltage and current deviation characteristic value, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS. There are five features in total.

[0113] The voltage and current deviation characteristic value is calculated in real time based on the importance of input voltage deviation, output voltage deviation, input current deviation and output current deviation. The smaller the characteristic value, the more stable the UPS equipment operation. The calculation formula of the voltage and current deviation characteristic value is as follows:

[0114] P IV =a p V IP +b p V OP +c p I IP +d p I OP

[0115] Among them, P IV is the voltage and current deviation characteristic value, V IP It is the difference between the input voltage during operation and the rated input voltage, that is, the input voltage deviation, a p is the input voltage deviation weight, V OP is the difference between the output voltage during operation and the rated output voltage, that is, the output voltage deviation, b p is the output voltage deviation weight, I IP It is the difference between the input current during operation and the rated input current, that is, the input current deviation, c p is the input current deviation weight, I OP It is the difference between the output current in operation and the rated output current, that is, the output current deviation, dp is the output current deviation weight; where a p 、b p 、c p ,d p Satisfies the following relationship:

[0116] a P +b P +c P +d P =1.

[0117] Input voltage deviation weight a p Calculated using the following formula:

[0118]

[0119] in,

[0120] max() is the maximum value function, V max is the maximum input voltage in the historical operation data, V min is the minimum input voltage in the historical operation data, V IR is the rated input voltage value;

[0121] Output voltage deviation weight b p =0.5-a p .

[0122] Input current deviation weight c p Calculated using the following formula:

[0123]

[0124] in,

[0125] max() is the maximum value function, I max is the maximum input current in the historical operation data, I min is the minimum input current value in the historical operation data, I IR is the rated input current value;

[0126] Output current deviation weight d p =0.5-c p .

[0127] The load change rate is the degree of change of the UPS device load within a set time interval. It reflects the UPS device's response capability and stability to load fluctuations; the lower the load change rate, the better the UPS device's adaptability to load changes; the set time interval is ten minutes.

[0128] The calculation formula for load change rate is:

[0129]

[0130] Among them, L R is the load change rate, L V is the load variation, i.e. the load difference within a set time interval, L A is the average load during the time interval.

[0131] The temperature change rate is the degree of change of the internal temperature of the UPS device within a set time interval; it reflects the heat dissipation performance and temperature stability of the UPS device; the lower the temperature change rate, the better the heat dissipation performance of the UPS device and the more stable the temperature control; the set time interval is one hour.

[0132] The temperature change rate is calculated as follows:

[0133]

[0134] Among them, T R is the rate of temperature change, T V is the temperature change, that is, the difference in the internal temperature of the device within the set time interval, and S is the number of sampling points within the set time interval.

[0135] 3.2 Normalize the performance degradation related features of each UPS device extracted in step 3.1 respectively.

[0136] The normalization formula is as follows:

[0137]

[0138] Among them, x′ is the normalized result, x is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0139] 3.3 Setting labels for the UPS device performance degradation related features normalized in step 3.2;

[0140] The label is the target value or result corresponding to the feature. There are four labels: 0, 1, 2, and 3. 0 means normal, 1 means the early stage of performance degradation, 2 means the middle stage of performance degradation, and 3 means the late stage of performance degradation.

[0141] 3.4 A data set is constructed using the normalized equipment performance degradation related features in step 3.2 and the labels of the corresponding normalized UPS equipment performance degradation related features set in step 3.3.

[0142] Step 4: Build a UPS equipment performance degradation prediction model based on deep learning;

[0143] The convolutional neural network-long short-term memory model CNN-LSTM, the recurrent neural network model RNN and the long short-term memory network model LSTM are combined into a UPS equipment performance degradation prediction model;

[0144] Build a CNN-LSTM model, including the input layer, Conv1D layer, first Pooling layer, Dropout layer, LSTM layer, second Pooling layer, Dense layer, Softmax layer and output layer;

[0145] The input data first passes through the Conv1D layer to extract features from the time series data; the Conv1D layer is connected to the first Pooling layer to compress the features extracted by the Conv1D layer; the first Pooling layer is connected to the Dropout layer to reduce overfitting; the Dropout layer sends the extracted features to the LSTM layer, which effectively captures the temporal dependencies of these features, and is connected to the second Pooling layer to compress the data again; the second Pooling layer is connected to the Dense layer to integrate all feature representations into one value, and finally the Dense layer is connected to the Softmax layer to output the classification results;

[0146] The input of the CNN-LSTM model is the load change rate in the dataset and the type of equipment connected to the UPS;

[0147] Build an RNN model whose input is the voltage and current deviation feature value in the dataset and the number of devices connected to the UPS;

[0148] Build an LSTM model whose input is the temperature change rate in the data set;

[0149] Step 5: Use the data set constructed in step 3 to train the UPS equipment performance degradation prediction model constructed in step 4;

[0150] Use the data set obtained in step 3.4 to train the CNN-LSTM, RNN, and LSTM in the UPS equipment performance degradation prediction model together, and adjust the parameters of the three models separately through the cross entropy loss function to minimize the prediction error.

[0151] Step 6: Collect the equipment information and real-time operation data of the UPS device to be predicted, pre-process the data according to steps 2-3 and extract the relevant features of UPS device performance degradation as the prediction data set, input it into the UPS device performance degradation prediction model trained in step 5, and predict the performance degradation degree of the UPS device in real time; finally, determine whether the final prediction result is normal. If not, remind the user to adopt the selected maintenance measures to avoid equipment failure.

[0152] 6.1 Apply the trained UPS equipment performance degradation prediction model to real-time operation data and record the predicted value;

[0153] The real-time operation data includes input voltage, output voltage, input current, output current, temperature, load, type of equipment connected to the UPS and number of equipment connected to the UPS, a total of 8 items of data.

[0154] The prediction result of the convolutional neural network-long short-term memory model is recorded as the first prediction result z r1 , the prediction result of the recurrent neural network model is recorded as the second prediction result z r2 , the prediction result of the long short-term memory network model is recorded as the third prediction result z r3 .

[0155] 6.2 Build a prediction algorithm and obtain the final prediction result through the prediction algorithm:

[0156] The formula of the prediction algorithm is as follows:

[0157]

[0158] Among them, P k is the probability of event k occurring, where k takes values ​​of 0, 1, 2, and 3, corresponding to the four labels in the data set; zk is the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k; e is the base of the natural logarithm.

[0159] The calculation formula of the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k is as follows:

[0160]

[0161] Among them, λ1 is the first prediction result z r1 The relative accuracy of the prediction, λ2 is the second prediction result z r2 The relative accuracy of the prediction; λ3 is the third prediction result z r3 The relative accuracy of the predictions;

[0162] λ i The calculation formula is as follows, i is 1, 2, 3;

[0163]

[0164] Among them, z ri is the i-th prediction result, λ i is the i-th prediction result z ri The relative accuracy of the predictions, is the first prediction result z r1 , the second prediction result z r2 and the third prediction result z r3The mean value, m z is the first prediction result z r1 , the second prediction result z r2 and the third prediction result z r3 The maximum absolute value of any two subtracted from each other.

[0165] Calculate the probability P of event k occurring through the prediction algorithm k , select the one with the highest probability of occurrence k, and record it as the risk probability P lk , and its corresponding k value is recorded as the final prediction result k L ; In particular, when m z =0, the first prediction result is recorded as the final prediction result, and the risk probability is recorded as 1;

[0166] 6.3 Determine whether the final prediction result obtained in step 6.2 is 0. If so, the UPS is normal; if not, remind the user to adopt the selected maintenance measures for maintenance.

[0167] The selected maintenance measures include conventional maintenance measures and additional maintenance measures.

[0168] When the final prediction result is 1 or 2, the conventional maintenance measures are selected as follows:

[0169] Improve heat dissipation, reduce UPS load, and ensure that the remaining power percentage of the UPS is higher than 20%.

[0170] When the final prediction result is 2, routine maintenance measures also include dust removal, static electricity protection, dehumidification, and controlling the internal temperature of the UPS to no more than 45 degrees.

[0171] When the final prediction result is 3, the routine maintenance measure is selected as replacing the UPS device.

[0172] The calculation formula for the remaining power percentage of the UPS is divided into the calculation formula when charging and the calculation formula when power is used.

[0173] The calculation formula for charging is as follows:

[0174]

[0175] Among them, s c Indicates the remaining power percentage of the UPS; c0 Indicates the initial remaining power percentage of the UPS; Q i Indicates the amount of electricity obtained by UPS charging; Q n Indicates the maximum amount of electricity that the UPS can store;

[0176] The calculation formula for electricity consumption is as follows:

[0177]

[0178] Among them, Q l Indicates the amount of power lost by the UPS.

[0179] Determine whether the hazard prediction value exceeds the set warning threshold. If so, set additional maintenance measures as follows:

[0180] When k L When it is 1: the internal temperature of UPS is controlled not to exceed 45 degrees;

[0181] When k L When the value is 2: Please ask professionals to maintain the UPS regularly;

[0182] When k L When it is 3: UPS will stop running immediately;

[0183] The warning threshold is set to 0.6;

[0184] The calculation formula of the risk prediction value is:

[0185]

[0186] Among them, F r The risk prediction value.

[0187] The present application also discloses a system based on the aforementioned deep learning-based UPS equipment performance degradation prediction method, including a data acquisition module, a data preprocessing module, a feature extraction module, a degradation performance prediction module, a maintenance measure selection module, and a communication module;

[0188] The data acquisition module is used to collect device information and real-time operation data of the UPS device to be predicted; wherein the device information of the UPS device includes UPS rated input voltage, UPS rated output voltage, UPS rated input current, and UPS rated output current; the historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of device connected to the UPS, and number of devices connected to the UPS;

[0189] The data preprocessing module uses One-Hot to encode the device type connected to the UPS device, and then cleans the abnormal values ​​in all the data and fills the missing values;

[0190] The feature extraction module extracts UPS device performance degradation related features, including voltage and current deviation characteristic values, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS; the extracted UPS device performance degradation related features are normalized and then used as a prediction data set;

[0191] The degradation performance prediction module combines the trained UPS equipment performance degradation prediction model and the prediction data set obtained by the feature extraction module to perform degradation performance prediction on the UPS equipment and obtain the final prediction result and risk probability;

[0192] The maintenance measure selection module determines whether the final prediction result obtained by the decay performance prediction module is normal. If not, the maintenance measure is selected according to the risk probability and the final prediction result, and the communication module sends information to remind the user to adopt the selected maintenance measure for maintenance. The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the above specific implementation methods. Therefore, any modification or replacement of the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the invention are included in the scope of the claims of the present invention.

[0193] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0194] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0195] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0196] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0197] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A UPS equipment performance degradation prediction method based on deep learning, characterized in that: The method comprises the following steps: Step 1: Collect the equipment information and historical operation data of the uninterruptible power supply UPS device to be predicted; Step 2: Preprocess the collected equipment information and historical operation data; Step 3: Extract UPS equipment performance degradation related features as a data set; Step 4: Build a UPS equipment performance degradation prediction model based on deep learning; Step 5: Use the data set to train the UPS equipment performance degradation prediction model; Step 6: Collect the equipment information and real-time operation data of the UPS device to be predicted, pre-process the data according to steps 2-3 and extract the relevant features of UPS device performance degradation as the prediction data set, input it into the UPS device performance degradation prediction model trained in step 5, and predict the performance degradation degree of the UPS device in real time.

2. The UPS device performance degradation prediction method based on deep learning according to claim 1 is characterized in that: In step 1, the device information of the UPS device includes a UPS rated input voltage, a UPS rated output voltage, a UPS rated input current, and a UPS rated output current; The historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of equipment connected to the UPS, and number of equipment connected to the UPS.

3. The UPS device performance degradation prediction method based on deep learning according to claim 1 is characterized in that: In step 2, One-Hot is used to encode the device type connected to the UPS device; then the outliers in all the data are cleaned and the missing values ​​are filled.

4. The UPS device performance degradation prediction method based on deep learning according to claim 1 is characterized in that: In step 3, the extracted UPS device performance degradation related features include voltage and current deviation characteristic values, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS.

5. The UPS device performance degradation prediction method based on deep learning according to claim 4 is characterized in that: The calculation formula of the voltage and current deviation characteristic value is as follows: P IV =a p V IP +b p V OP +c p I IP +d p I OP Among them, P IV is the voltage and current deviation characteristic value, V IP is the difference between the input voltage during operation and the rated input voltage, a p is the input voltage deviation weight, V OP is the difference between the output voltage during operation and the rated output voltage, b p is the output voltage deviation weight, I IP is the difference between the input current during operation and the rated input current, c p is the input current deviation weight, I OP is the difference between the output current during operation and the rated output current, d p is the output current deviation weight; where a p 、b p 、c p d p Satisfies the following relationship: a P +b P +c P +d P =1。 6. The UPS device performance degradation prediction method based on deep learning according to claim 5 is characterized in that: Input voltage deviation weight a p Calculated using the following formula: in, max() is the maximum value function, V max is the maximum input voltage in the historical operation data, V min is the minimum input voltage in the historical operation data, V IR is the rated input voltage value; Output voltage deviation weight b p =0.5-a p .

7. The UPS device performance degradation prediction method based on deep learning according to claim 6 is characterized in that: Input current deviation weight c p Calculated using the following formula: in, max() is the maximum value function, I max is the maximum input current in the historical operation data, I min is the minimum input current value in the historical operation data, I IR is the rated input current value; Output current deviation weight d p =0.5-c p .

8. The UPS device performance degradation prediction method based on deep learning according to claim 4 or 5, characterized in that: The calculation formula of the load change rate is as follows: Among them, L R is the load change rate, L V is the load difference within a set time interval, L A is the average load during the time interval; The calculation formula of the temperature change rate is as follows: Among them, T R is the rate of temperature change, T V is the temperature difference of the equipment within the set time interval, and S is the number of sampling points within the set time interval.

9. The UPS device performance degradation prediction method based on deep learning according to claim 1, characterized in that: In step 3, the data set also includes labels corresponding to performance degradation related features, and the labels include 0, 1, 2 and 3; among which 0 represents normal, 1 represents the early stage of performance degradation, 2 represents the middle stage of performance degradation, and 3 represents the late stage of performance degradation.

10. The UPS equipment performance degradation prediction method based on deep learning according to claim 1, characterized in that: In step 4, the convolutional neural network-long short-term memory model CNN-LSTM, the recurrent neural network model RNN and the long short-term memory network model LSTM are combined into a UPS equipment performance degradation prediction model; The input of the CNN-LSTM model is the load change rate and the type of equipment connected to the UPS. Among them, the input of the recurrent neural network is the voltage and current deviation characteristic value and the number of devices connected to the UPS; Among them, the input of the long short-term memory network is the temperature change rate.

11. The UPS device performance degradation prediction method based on deep learning according to claim 10, characterized in that: In step 6, the prediction result of the convolutional neural network-long short-term memory model is recorded as the first prediction result z r1 , the prediction result of the recurrent neural network model is recorded as the second prediction result z r2 , the prediction result of the long short-term memory network model is recorded as the third prediction result z r3 , the prediction algorithm is used to obtain the final prediction result of the UPS equipment performance degradation prediction model according to the first prediction result, the second prediction result and the third prediction result. The formula of the prediction algorithm is as follows: Among them, P k is the probability of event k occurring, where the values ​​of k are 0, 1, 2, and 3, corresponding to the four labels in the data set; z k is the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k; e is the base of the natural logarithm; The calculation formula of the degree to which the predicted value of the UPS equipment performance degradation prediction model deviates from k is as follows: Among them, λ1 is the prediction relative accuracy of the first prediction result, λ2 is the prediction relative accuracy of the second prediction result; λ3 is the prediction relative accuracy of the third prediction result; λ i The calculation formula is as follows, i is 1, 2, 3; Among them, z ri is the i-th prediction result, λ i is the relative accuracy of the prediction of the i-th prediction result, is the mean of the first prediction result, the second prediction result and the third prediction result, m z is the maximum value of the absolute value obtained by subtracting any two of the first prediction result, the second prediction result, and the third prediction result; Calculate the probability P of event k occurring through the prediction algorithm k , select the one with the highest probability of occurrence k, and record it as the risk probability P lk , and its corresponding k value is recorded as the final prediction result k L ; In particular, when m z =0, the first prediction result is recorded as the final prediction result, and the risk probability is recorded as 1.

12. The UPS device performance degradation prediction method based on deep learning according to claim 11, characterized in that: Determine whether the final prediction result is 0. If so, the UPS is normal; if not, remind the user to adopt the selected maintenance measures for maintenance; The selected maintenance measures include conventional maintenance measures and additional maintenance measures; wherein the additional maintenance measures are selected based on the risk probability and the final prediction result.

13. The UPS device performance degradation prediction method based on deep learning according to claim 12, characterized in that: When the final prediction result is 1 or 2, the conventional maintenance measures are selected as: enhancing heat dissipation, reducing UPS load, and ensuring that the remaining power percentage of the UPS is higher than 20%; When the final prediction result is 2, routine maintenance measures also include dust removal, electrostatic protection, dehumidification, and controlling the internal temperature of the UPS to no more than 45 degrees; When the final prediction result is 3, the routine maintenance measure is selected as replacing the UPS device; The UPS remaining power percentage calculation formula is divided into charging calculation formula and power consumption calculation formula; The calculation formula for charging is as follows: Among them, s c Indicates the remaining power percentage of the UPS; c0 Indicates the initial remaining power percentage of the UPS; Q i Indicates the amount of electricity obtained by UPS charging; Q n Indicates the maximum amount of electricity that the UPS can store; The calculation formula for electricity consumption is as follows: Among them, Q l Indicates the power lost by the UPS.

14. The UPS device performance degradation prediction method based on deep learning according to claim 12, characterized in that: Determine whether the hazard prediction value exceeds the set warning threshold. If so, set additional maintenance measures as follows: When k L When it is 1: the internal temperature of UPS is controlled not to exceed 45 degrees; When k L When the value is 2: Please ask professionals to maintain the UPS regularly; When k L When it is 3: UPS will stop running immediately; The warning threshold is set to 0.6; The calculation formula of the risk prediction value is: Among them, F r The risk prediction value.

15. A UPS device performance degradation prediction system based on deep learning using the UPS device performance degradation prediction method based on deep learning according to any one of claims 1 to 14, characterized in that: It includes data acquisition module, data preprocessing module, feature extraction module, degradation performance prediction module, maintenance measure selection module and communication module; The data acquisition module is used to collect device information and real-time operation data of the UPS device to be predicted; wherein the device information of the UPS device includes UPS rated input voltage, UPS rated output voltage, UPS rated input current, and UPS rated output current; the historical operation data includes input voltage, output voltage, input current, output current, temperature, load, type of device connected to the UPS, and number of devices connected to the UPS; The data preprocessing module uses One-Hot to encode the device type connected to the UPS device, and then cleans the abnormal values ​​in all the data and fills the missing values; The feature extraction module extracts UPS device performance degradation related features, including voltage and current deviation characteristic values, load change rate, temperature change rate, device type connected to the UPS, and the number of devices connected to the UPS; and uses the extracted UPS device performance degradation related features as a prediction data set; The degradation performance prediction module combines the trained UPS equipment performance degradation prediction model and the prediction data set obtained by the feature extraction module to perform degradation performance prediction on the UPS equipment and obtain the final prediction result and risk probability; The maintenance measure selection module determines whether the final prediction result obtained by the degradation performance prediction module is normal. If not, it selects maintenance measures according to the risk probability and the final prediction result, and sends information through the communication module to remind the user to adopt the selected maintenance measures for maintenance.

16. An electronic device, comprising a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is used to operate according to the instructions to execute the UPS device performance degradation prediction method based on deep learning according to claims 1-14.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the UPS device performance degradation prediction method based on deep learning described in claims 1-14 is implemented.

Citation Information

Cited By

  • UPS equipment capacity prediction method and system

    CN121325000A