A current linkage early warning method for coal mine underground multi-motor coupling effect

Abstract

The present application belongs to the technical field of current safety production, and provides a current linkage early warning method for the coupling effect of multiple motors in a coal mine, comprising: setting a sensor network to collect historical current sequences and historical working condition sequences of each motor, marking abnormal working conditions and corresponding precursor working conditions in the historical working condition sequences of each motor, setting a monitoring window, judging whether there is an abnormal working condition or a precursor working condition in the monitoring window, if not, triggering current prediction, constructing a fusion prediction model, the fusion prediction model comprising a time series current prediction model, a time series working condition prediction model, and a motor group coupling model established by obtaining each motor basic information, predicting each motor through the fusion prediction model to obtain a predicted current sequence of each motor, analyzing the historical current sequence in the monitoring window and the predicted current sequence obtained by prediction, analyzing the current abnormality of the motor and giving an early warning.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of current safety production, and in particular relates to a current linkage early warning method for a multi-motor coupling effect in a coal mine. BACKGROUND

[0002] In coal mining operations, the coordinated operation of multiple motors in the mine is a key link to ensure the smooth progress of the production process. Numerous motors drive the belt transport, ventilation, drainage and other systems. However, with the increasing depth and scale of coal mining, the operating environment of the motors in the mine is becoming increasingly complex. Adverse conditions such as high temperature, humidity and dust not only affect the performance of the motors, but also significantly increase the risk of motor failure. At the same time, the electromagnetic coupling and mechanical coupling between multiple motors interact, making it easy to trigger a chain reaction when an abnormality occurs in one motor, affecting the stable operation of the entire system, and even causing serious safety accidents.

[0003] Traditional motor monitoring systems often rely on a single type of sensor to collect data, such as monitoring only current or temperature, which cannot fully reflect the operating state of the motor. Due to the complex causes of motor failure in coal mines, single parameter monitoring cannot capture potential fault signals in the early stage, resulting in many faults being discovered when they have developed to a serious stage, delaying maintenance opportunities and causing production interruptions and economic losses.

[0004] On the other hand, existing early warning methods mostly use fixed thresholds to judge abnormalities, lacking adaptability to dynamic changes in motor operating conditions. The operating scene in a coal mine is variable, and the motor load and environmental conditions can change at any time. Fixed thresholds cannot accurately distinguish between normal fluctuations and true abnormal states, and may easily cause false positives or false negatives. For example, when underground equipment starts or the load suddenly changes, the current and other parameters will fluctuate temporarily, and traditional threshold judgments may misjudge them as abnormal. When a motor has a slight fault and the parameter change does not reach the fixed threshold, it may not be able to issue an early warning in time.

[0005] In addition, for the coupling effect between multiple motors, existing technologies do not fully consider its impact on the operating state of the motor. In a multi-motor system, the interaction between motors can cause the changes in current, speed and other parameters to exhibit complex nonlinear relationships. Traditional methods cannot accurately predict the operating trend of the motor under such coupling effects, and cannot take effective measures to prevent the expansion and spread of faults.

[0006] To solve the above problems, the present application proposes a current linkage early warning method for a multi-motor coupling effect in a coal mine. SUMMARY

[0007] To compensate for the shortcomings of the prior art and solve at least one of the technical problems raised in the background art.

[0008] The technical scheme adopted by the present application to solve the technical problem is: a current linkage early warning method for a coal mine underground multi-motor coupling effect, comprising:

[0009] A sensor network is set for the coal mine underground multi-motor linkage system, and historical current sequences and historical working condition sequences of each motor are collected;

[0010] Abnormal working conditions and corresponding precursor working conditions are marked in the historical working condition sequences of each motor, a monitoring window is set, and it is determined whether there is an abnormal working condition or a precursor working condition in the monitoring window, if not, current prediction is triggered;

[0011] If current prediction is triggered, a fusion prediction model is constructed, the fusion prediction model includes a time series current prediction model and a time series working condition prediction model trained based on the historical current sequences and the historical working condition sequences, and a motor group coupling model established by obtaining basic information of each motor, each motor is predicted by the fusion prediction model, and a predicted current sequence of each motor is obtained;

[0012] The historical current sequences in the monitoring window and the predicted current sequences obtained by prediction are analyzed, current anomalies of the motor are analyzed and early warning is performed;

[0013] The acquisition method of the historical current sequences and the historical working condition sequences is:

[0014] A sensor network is set for the coal mine underground multi-motor linkage system, and motor parameters of each motor are collected in real time according to the set data collection frequency, the sensor network is marked in real time as the collection time according to the data collection frequency, and at each collection time, a three-phase stator current vector and a working condition data vector of each motor are constructed;

[0015] The time period between the time when the sensor network starts collecting and the current time is marked as a historical period, for each motor, the three-phase stator current vector and the working condition data vector collected in the historical period are integrated according to the time sequence, and the historical current sequence and the historical working condition sequence of each motor are obtained;

[0016] The determination method of whether there is an abnormal working condition or a precursor working condition is:

[0017] A monitoring window with the current time as the end point and the same time length is set, it is determined whether there is an overlap between the monitoring window and the time period corresponding to any abnormal working condition, if there is, it is determined that there is an abnormal working condition in the monitoring window;

[0018] If not, a plurality of precursor analysis time periods are set in the monitoring window, the precursor analysis time periods all have the current time as the end point and different time lengths, the historical working condition sequences in all the precursor analysis time periods of the motor are intercepted, the precursor analysis sequences of the motor are obtained and normalized respectively, and the normalized precursor analysis sequences are obtained.

[0019] obtaining all normalized precursor sequences of the precursor working conditions, and determining that the precursor working condition exists in the monitoring window if the DTW distance between any normalized precursor analysis sequence and any normalized precursor sequence in the monitoring window is less than the distance threshold value;

[0020] The abnormal working condition is determined in the following manner:

[0021] The mean and standard deviation of each parameter of the working condition data vector in the historical working condition sequence of each motor under the normal working condition data range are obtained and calculated, and the dynamic abnormal threshold value is obtained through data processing. If any parameter of the working condition data vector at any collection time exceeds the set dynamic abnormal threshold value, the collection time is marked with an abnormal time marker including the motor number;

[0022] For any motor number, if the abnormal time marker exists at not less than KL consecutive collection times, it is determined that the corresponding motor is in an abnormal working condition, and the collection times with the abnormal time marker are marked as an abnormal working condition of the corresponding motor. The adjacent abnormal working conditions of the same motor with a time distance less than a preset time threshold value are merged into one abnormal working condition;

[0023] The precursor working condition is determined in the following manner:

[0024] All precursor sequence groups and normalized precursor sequences contained therein are obtained. The number of normalized precursor sequences in the same precursor sequence group and the total number of abnormal sequences in the corresponding same type abnormal working condition are processed to obtain a stability ratio.

[0025] If the DTW distance between any two normalized precursor sequences not belonging to the same precursor sequence group meets the similarity standard, the two precursor sequence groups are classified into the same similarity set. The number of precursor sequence groups in the similarity set where the precursor sequence groups are located is obtained, and data processing is performed with the total number of all precursor sequence groups to obtain a similarity ratio.

[0026] If the stability ratio is greater than a preset stability standard, and the similarity ratio is less than a preset similarity standard, it is determined that the precursor sequence corresponding to the normalized precursor sequence in the precursor sequence group is a precursor working condition.

[0027] The precursor sequence group is obtained in the following manner:

[0028] In the historical working condition sequence of the motor, the abnormal sequence of the motor is obtained and normalized respectively to obtain the normalized abnormal sequence and the normalized working condition data vector contained therein. The DTW distance between the normalized abnormal sequences corresponding to any two abnormal sequences of the same motor is calculated. If it is less than the distance threshold value, the two abnormal sequences are the same type of abnormal working condition.

[0029] In the historical working condition sequence of the motor, the precursor sequence of the motor is intercepted and normalized to obtain a normalized precursor sequence, and in the same type of abnormal working condition, the DTW distance between any two abnormal sequence corresponding normalized precursor sequences is calculated, and if it is less than the distance threshold, it is classified into the same precursor sequence group;

[0030] The acquisition method of the predicted current sequence is:

[0031] The root mean square error of the time series current prediction model and the motor group coupling model in the time series current training set is calculated, and the weight of the time series current prediction model and the motor group coupling model is set according to the root mean square error,

[0032] The fusion prediction model weights the prediction results of the time series current prediction model and the motor group coupling model based on the set weight, that is, the time series predicted current sequence and the physical predicted current sequence are weighted and fused by the calculated weight, to obtain the predicted current sequence of each motor;

[0033] The acquisition method of the time series predicted current sequence is:

[0034] All working condition data vectors and three-phase stator current vectors in the abnormal working condition and the precursor working condition in the historical working condition sequence and the historical current sequence are respectively removed, the historical working condition sequence and the historical current sequence are supplemented and normalized by interpolation method, to obtain the time series working condition training set and the time series current training set, and the monitoring window is obtained and the window working condition sequence and the window current sequence are obtained by intercepting;

[0035] The long short-term memory network is used, the time series working condition training set and the time series current training set are used, the time series working condition prediction model and the time series current prediction model are respectively constructed and trained, and the window working condition sequence and the window current sequence are combined to obtain the time series predicted working condition sequence and the time series predicted current sequence of the motor;

[0036] The acquisition method of the physical predicted current sequence is:

[0037] The basic information of each motor is obtained, and the loss parameters of each motor are obtained by statistical analysis of the historical current sequence and the historical working condition sequence, and based on the motor dynamics equation and the Kirchhoff law, the multi-motor coupling effect is combined to establish a motor group coupling model with the input of each motor working condition data vector and the output of each motor three-phase stator current vector, and the time series predicted working condition sequence of each motor is input into the motor group coupling model to obtain the physical predicted current sequence of each motor;

[0038] The analysis of the current anomaly of the motor includes:

[0039] The window current sequence and the predicted current sequence are acquired, the window current sequence includes a three-phase stator current vector, and the predicted current sequence includes a predicted three-phase stator current vector; the current effective value and the predicted current effective value are calculated through data processing; if the current effective value or the predicted current effective value meets the current overload abnormal condition at continuous not less than KL collection time points, it is determined that the corresponding motor exists a current overload abnormality.

[0040] On the contrary, the three-phase unbalance degree of the motor and the predicted three-phase unbalance degree are calculated; if the three-phase unbalance degree or the predicted three-phase unbalance degree is greater than the national standard critical value at continuous not less than KL collection time points, it is determined that the corresponding motor exists a three-phase unbalance abnormality.

[0041] The beneficial effects of the present application are as follows:

[0042] 1. The present application comprehensively collects the historical current and working condition sequence of each motor through the setting of the sensor network, and marks the abnormal working condition and the precursor working condition, can accurately capture the key information in the motor operation, based on these data, a fusion prediction model is constructed, the current, working condition and motor group coupling effect are comprehensively considered, the prediction result is more close to the actual situation, compared with the traditional single monitoring mode, the potential current abnormality can be found in advance, the production interruption caused by misjudgment is effectively avoided, the continuity and stability of coal mine production are significantly improved, and the economic loss caused by motor failure is reduced.

[0043] 2. The present application can monitor the motor running state in real time and dynamically through the setting of the monitoring window and analysis and early warning, when there is no abnormal or precursor working condition in the monitoring window, the current prediction is triggered, the future current change trend of the motor is grasped in time, once the current abnormality is found, the early warning can be quickly sent, the valuable processing time of the staff is saved, which helps to take maintenance measures in advance, prevents the fault from expanding, ensures the safety of the coal mine underground operating personnel and equipment, improves the safety and reliability of coal mine production, and reduces the probability of safety accidents. BRIEF DESCRIPTION OF DRAWINGS

[0044] The present application will be further described below in conjunction with the drawings.

[0045] Figure 1 It is a step flow chart of a current linkage early warning method for coal mine underground multi-motor coupling effect according to an embodiment of the present application;

[0046] Figure 2 It is a step flow chart of the acquisition of the predicted current sequence in the current linkage early warning method for coal mine underground multi-motor coupling effect according to an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application is further described below in conjunction with specific embodiments.

[0048] Embodiment 1

[0049] Please refer to Figure 1 The current linkage early warning method for the coupling effect of multiple motors in a coal mine underground, as shown in the embodiment of the present application, comprises the following steps:

[0050] S1: Set up a sensor network to collect the historical current sequence and historical working condition sequence of each motor;

[0051] A sensor network is set up for the multiple motor linkage system in a coal mine underground, which includes current sensors, vibration sensors, temperature sensors and speed sensors. According to the working characteristics and monitoring requirements of the multiple motor linkage system in a coal mine underground, the data acquisition frequency is determined. According to the data acquisition frequency, the motor parameters of each motor of the multiple motor linkage system in a coal mine underground are collected in real time through the sensor network. The motor parameters include three-phase stator current, bearing seat vibration acceleration, winding temperature, bearing temperature, rotor speed and input voltage. The time when the sensor network collects the motor parameters of each motor in real time is marked as the collection time;

[0052] Among them, the current sensor is connected in series to the three-phase circuit of the power distribution cabinet of each motor in the multiple motor linkage system in a coal mine underground, ensuring the consistency of the current direction. The current sensor uses a closed-loop high-precision explosion-proof Hall current sensor to collect the three-phase stator current of each motor in real time 、 and , where m represents the number of the motor in the multiple motor linkage system in a coal mine underground, and t represents the collection time of the sensor network for data collection;

[0053] Among them, the vibration sensor is rigidly fixed to the radial bearing area of each motor bearing seat in the multiple motor linkage system in a coal mine underground through a special bolt. The vibration sensor uses a wideband accelerometer to capture the vibration acceleration of each motor bearing seat .

[0054] Among them, the temperature sensor is embedded in the stator winding of each motor in the multiple motor linkage system in a coal mine underground, and a surface-mounted temperature sensor is installed at each motor bearing to synchronously collect the winding temperature and bearing temperature .

[0055] The rotation speed sensor adopts a photoelectric / magnetoelectric dual-mode probe, the distance between the photoelectric probe and the reflective code disc is (1.0±0.2) mm, the air gap of the magnetoelectric type is (1.5±0.5) mm, and the rotation speed sensor is used to monitor the rotation speed of each motor rotor in the multi-motor linkage system in the coal mine in real time ;

[0056] The voltage sensor adopts a high-precision explosion-proof voltage transformer and is connected in parallel to the power supply circuit of each motor in the multi-motor linkage system in the coal mine, the variable ratio is adapted to the rated voltage, and the input voltage of each motor is collected ;

[0057] For the collected motor parameters of each motor in the multi-motor linkage system in the coal mine, at each collection time, the three-phase stator current vector of each motor is constructed and the working condition data vector ;

[0058] ;

[0059] ;

[0060] The edge preprocessing and hybrid networking architecture are used to realize efficient data transmission, eliminate random noise and power grid interference, and the GPS / IRIG-B dual-mode time module is used to synchronize the time scale of the motor parameters of each motor

[0061] The time period between the time when the sensor network starts collecting and the current time is marked as a historical period, and for each motor, the three-phase stator current vector and the working condition data vector collected in the historical period are integrated according to the time sequence to obtain the historical current sequence and the historical working condition sequence of each motor

[0062] It should be noted that the purpose of this step is to construct a sensor network containing multiple sensors, determine the collection frequency according to the working characteristics of the motor in the coal mine, real-time acquire the parameters such as three-phase stator current, vibration acceleration, temperature, rotation speed and voltage of the motor, construct the three-phase stator current vector and the working condition data vector of each motor, and perform data transmission optimization and time scale synchronization, finally integrate the historical current sequence and the historical working condition sequence to provide basic data for subsequent analysis. The edge preprocessing and hybrid networking architecture are used to eliminate noise and interference and improve data quality; time scale synchronization ensures data time consistency, which is convenient for subsequent time sequence analysis

[0063] S2: Label abnormal working conditions and corresponding precursor working conditions in the historical working condition sequence of each motor, set a monitoring window, and determine whether there is an abnormal working condition or a precursor working condition in the monitoring window, if not, trigger current prediction

[0064] In the multi-motor linkage system of a coal mine at the current moment, the historical operating condition sequence of each motor is obtained. Based on the normal operating condition data range screened by preliminary rules and manually marked, the mean value of the nth parameter of the operating condition data vector in the historical operating condition sequence of motor number m is calculated under the normal operating condition data range. and standard deviation , set the dynamic anomaly threshold to:

[0065] ;

[0066] Where k is a manually preset empirical coefficient. For each motor, in the historical operating condition sequence, if any parameter of the operating condition data vector at any acquisition moment exceeds the preset dynamic abnormality threshold, the motor is determined to have a parameter abnormality at that acquisition moment, and an abnormal moment mark including the motor number is assigned to the acquisition moment.

[0067] To avoid misjudgment due to individual noise data, a continuous judgment mechanism is adopted. For any motor number, if there are abnormal time mark at no less than KL consecutive collection moments, the corresponding motor is determined to be in an abnormal working condition. Otherwise, the abnormal time mark of the corresponding motor number at the collection moment is cleared, and the collection moments with continuous abnormal time mark of the same motor number are demarcated as an abnormal working condition of the corresponding motor. For any motor, adjacent abnormal working conditions with a time distance less than a preset time threshold are merged into one abnormal working condition.

[0068] Based on any motor, all abnormal operating conditions are extracted from the motor's historical operating condition sequence, and the motor's abnormal sequence is intercepted. Within each abnormal sequence, each parameter of the operating condition data vector is normalized using the minimum-maximum normalization method to obtain a normalized abnormal sequence and a normalized operating condition data vector within the normalized abnormal sequence;

[0069] For any two abnormal sequences of the same motor, the DTW distance between the two corresponding normalized abnormal sequences is calculated using the dynamic time warping (DTW) algorithm to analyze the similarity of the change trends between the two abnormal sequences.

[0070] Specifically, an M×N distance matrix D is constructed, where M and N are the lengths of the two normalized anomaly sequences, respectively. The matrix element D(i,j) represents the Euclidean distance between the i-th and j-th normalized operating condition data vectors in the two normalized anomaly sequences. The shortest path from the upper left corner to the lower right corner of the distance matrix is ​​calculated using a dynamic programming method. The cumulative distance of the shortest path is the DTW distance of the two normalized anomaly sequences. The smaller the DTW distance, the more similar the change trends of the two normalized anomaly sequences are.

[0071] The abnormal sequences corresponding to the normalized abnormal sequences of the same motor whose DTW distance is less than the distance threshold are divided into the same type of abnormal working conditions;

[0072] a prognosis collection period is set for each abnormal sequence, the end of the prognosis collection period is the start of the corresponding abnormal sequence, the length of the prognosis collection period is determined by the length of the abnormal sequence, for the same motor, a reference length is set for the prognosis collection period of the first abnormal sequence in the historical period, the length ratio of all abnormal sequences of the same motor to the corresponding prognosis collection period is the same;

[0073] In the historical working condition sequence of the motor, the historical working condition sequence in all prognosis collection periods of the motor is intercepted to obtain the prognosis sequence of the motor, and in each prognosis sequence, each parameter of the working condition data vector is normalized by using the minimum-maximum normalization method to obtain a normalized prognosis sequence;

[0074] Similarly, in the same type of abnormal working condition, the normalized prognosis sequences corresponding to any two abnormal sequences are obtained, the DTW distance between the two normalized prognosis sequences is calculated by using the dynamic time warping algorithm, and the two normalized prognosis sequences with a DTW distance less than a distance threshold are classified into the same prognosis sequence group, and if there is a prognosis sequence that is not classified into any prognosis sequence group, the prognosis sequence itself is a prognosis sequence group.

[0075] The number of normalized prognosis sequences in the same prognosis sequence group is obtained, and a ratio is processed with the total number of abnormal sequences in the corresponding same type of abnormal working condition to obtain a stability ratio.

[0076] For two different prognosis sequence groups, if there is any two normalized prognosis sequences not belonging to the same prognosis sequence group with a DTW distance less than a distance threshold, it is judged that the two prognosis sequence groups are similar, and the two prognosis sequence groups are classified into the same similar set, and if there is any prognosis sequence group that is not classified into any similar set, the prognosis sequence group itself is a similar set.

[0077] The number of prognosis sequence groups in the similar set where any prognosis sequence group is located is obtained, and a ratio is processed with the total number of all prognosis sequence groups to obtain a similarity ratio.

[0078] If the stability ratio is greater than a preset stability standard and the similarity ratio is less than a preset similarity standard, it is judged that the prognosis sequence corresponding to the normalized prognosis sequence in the prognosis sequence group is a prognosis working condition of the same type of abnormal working condition.

[0079] It should be noted that based on the preset proportion threshold, there can be multiple prognosis working conditions for the same type of abnormal working condition, or no prognosis working condition can be found.

[0080] A monitoring window is set, the monitoring window is a period with a fixed length ending at the current time, the monitoring window slides with the change of time, and it is judged whether there is an overlap between the monitoring window and the period corresponding to any abnormal sequence.

[0081] If the abnormal working condition exists, it is determined that the abnormal working condition exists in the monitoring window;

[0082] If the abnormal working condition does not exist, a plurality of precursor analysis periods are set in the monitoring window, the precursor analysis period ends at the current time, and the lengths of all precursor analysis periods in the monitoring window are different, wherein the longest precursor analysis period has the same length as the monitoring window;

[0083] In the historical working condition sequence of the motor, the historical working condition sequence in all precursor analysis periods of the motor is intercepted to obtain a precursor analysis sequence of the motor, and each parameter of the working condition data vector is normalized by using the minimum-maximum normalization method in each precursor analysis sequence to obtain a normalized precursor analysis sequence;

[0084] If the DTW distance between any normalized precursor analysis sequence in the monitoring window and any normalized precursor sequence of the precursor working condition is less than the distance threshold, it is determined that the precursor working condition occurs in the monitoring window;

[0085] For any motor, if the abnormal working condition exists or the precursor working condition occurs in the monitoring window, an alarm is generated and the motor number is sent to the administrator terminal;

[0086] If all motors in the coal mine multi-motor linkage system do not have abnormal working conditions and precursor working conditions in the monitoring window, it is determined that the coal mine multi-motor linkage system is running normally, and the current prediction is triggered;

[0087] It should be noted that the purpose of this step is to mark abnormal working conditions and their precursor working conditions based on historical working condition data through statistical analysis and clustering algorithm. The dynamic time warping (DTW) algorithm is introduced to cluster and analyze abnormal working conditions and precursor working conditions, breaking through the limitations of traditional single threshold judgment, capturing complex change trends, judging precursor working conditions based on stable proportion and similar proportion, establishing a more scientific abnormal precursor judgment standard, setting a monitoring window to determine whether there is an abnormal working condition or a precursor working condition in real time, and triggering the current prediction if there is no precursor working condition. The mining of precursor working conditions realizes early warning of failure, which saves time for preventive maintenance. The monitoring window mechanism ensures real-time monitoring, timely discovery of potential risks, early identification of abnormalities, and early warning triggering decision;

[0088] S3: If the current prediction is triggered, a fusion prediction model is constructed, the fusion prediction model includes a time series current prediction model and a time series working condition prediction model trained based on historical current sequences and historical working condition sequences respectively, and a motor group coupling model established based on basic information of each motor. The fusion prediction model is used to predict each motor to obtain a predicted current sequence of each motor;

[0089] As shown in Figure 2 , the specific acquisition steps of the predicted current sequence are as follows;

[0090] If the current prediction is triggered, a fusion prediction model of a multi-motor linkage system in a coal mine underground is constructed, and the fusion prediction model includes a time series current prediction model, a time series working condition prediction model, and a motor group coupling model;

[0091] Specifically, a historical current sequence of each motor is obtained, all abnormal working conditions and three-phase stator current vectors in a premonitory collection period in the historical current sequence are removed, the historical current sequence is supplemented by an interpolation method, and the supplemented historical current sequence is normalized to obtain a time series current training set, the time series current training set in a monitoring window is intercepted to obtain a window current sequence;

[0092] Similarly, a historical working condition sequence of each motor is obtained, all abnormal working conditions and working condition data vectors in a premonitory collection period in the historical working condition sequence are removed, the historical working condition sequence is supplemented by an interpolation method, and the supplemented historical working condition sequence is normalized to obtain a time series working condition training set, the time series working condition training set in a monitoring window is intercepted to obtain a window working condition sequence;

[0093] A long short-term memory network (LSTM) is used to construct and train the time series current prediction model and the time series working condition prediction model;

[0094] The time series current prediction model is provided with an input layer, a hidden layer, and an output layer, the input layer receives the window current sequence, the hidden layer is provided with two LSTM units and a dropout layer to prevent overfitting, the output layer outputs a time series predicted current sequence of a prediction window through a fully connected layer, in the training process, a root mean square error is used as a loss function, an Adam optimizer is used, an initial learning rate, a decay strategy, and a training round are set, an early stopping method is used to avoid overfitting, the time series current training set is divided into a training set and a validation set at a ratio of 8:2, a root mean square error (RMSE) of the time series current prediction model on the validation set is calculated, and when the RMSE is less than a preset threshold, it is determined that the time series current prediction model is trained.

[0095] The prediction window is a time period with a fixed length starting from a current time point, and the prediction window slides with time.

[0096] The window current sequence is input into the trained time series current prediction model to obtain a time series predicted current sequence of the motor in the prediction window.

[0097] Similarly, the time series working condition prediction model is provided with an input layer, a hidden layer, and an output layer, the input layer receives the window working condition sequence, the output layer outputs a time series predicted working condition sequence of a prediction window through a fully connected layer, the time series working condition training set is divided into a training set and a validation set at a ratio of 8:2, a root mean square error (RMSE) of the time series working condition prediction model on the validation set is calculated, and when the RMSE is less than a preset threshold, it is determined that the time series working condition prediction model is trained.

[0098] The window condition sequence is input into the trained time sequence condition prediction model to obtain a time sequence predicted condition sequence of the motor in the prediction window;

[0099] The basic information of each motor is obtained through motor nameplate parameters, design data and mechanical parameters, and the loss parameters of each motor are obtained through historical current sequence and historical condition sequence. Based on the motor dynamics equation and Kirchhoff's law, combined with the coupling effect of multi-motor, a motor group coupling model is established, which takes each motor condition data vector as input and outputs each motor three-phase stator current vector:

[0100] ;

[0101] The obtained time sequence predicted condition sequence is input into the motor group coupling model to obtain each motor physical predicted current sequence;

[0102] The prediction accuracy index is defined And , And The root mean square error of the time sequence current prediction model and the motor group coupling model in the time sequence current training set is respectively represented, and the weights of the time sequence current prediction model and the motor group coupling model are set And :

[0103] ;

[0104] ;

[0105] The fusion prediction model is weighted and fused with the prediction results of the time sequence current prediction model and the motor group coupling model based on the calculated weight, that is, the time sequence predicted current sequence and the physical predicted current sequence are weighted and fused through the calculated weight, to obtain the predicted current sequence of each motor, and the predicted current sequence includes the predicted three-phase stator current vector of the motor at each collection time in the prediction window ;

[0106] ;

[0107] Wherein represents the current time, represents the Lth sampling time after the current time;

[0108] It should be noted that the role of this step is to train the time series prediction model using historical normal data when no current anomaly is found, and to establish a motor group coupling model combining motor basic information and loss parameters. The current is predicted by two models and the results are fused. The time series prediction model captures the time series characteristics of motor parameters, and the coupling model considers the interaction of multiple motors. The fusion prediction improves the accuracy, predicts the current change in advance, provides data support for coal mine motor operation scheduling and fault prevention, and dynamically allocates weights based on prediction accuracy. The fusion method enhances the reliability and adaptability of the prediction result;

[0109] S4: Analyze the historical current sequence in the monitoring window and the predicted current sequence obtained by prediction, analyze the current anomaly of the motor, and give an early warning;

[0110] Based on the window current sequence and the obtained predicted current sequence, the three-phase stator current vector at each collection time in the monitoring window and the prediction window is obtained and the predicted three-phase stator current vector , the current effective value and the predicted current effective value , the formula is:

[0111] ;

[0112] ;

[0113] Set the current anomaly threshold. If any collection time satisfies the current effective value or the predicted current effective value is greater than the set current anomaly threshold, it is determined that the current overload anomaly occurs at the collection time. If the current overload anomaly occurs at not less than KL consecutive collection times, it is determined that the corresponding motor has a current overload anomaly.

[0114] Otherwise, calculate the three-phase unbalance degree of the motor and the predicted three-phase unbalance degree , the formula is:

[0115] ;

[0116] ;

[0117] If any collection time satisfies the three-phase unbalance degree or the predicted three-phase unbalance degree is greater than the national standard threshold, it is determined that the three-phase unbalance anomaly occurs at the collection time. If the three-phase unbalance anomaly occurs at not less than KL consecutive collection times, it is determined that the corresponding motor has a three-phase unbalance anomaly.

[0118] If the motor has current overload anomaly or three-phase imbalance anomaly, an alarm signal is generated and the motor number and current anomaly type are sent to the administrator terminal, and the current anomaly type includes current overload anomaly and three-phase imbalance anomaly.

[0119] It should be noted that the role of this step is to judge whether the current is abnormal by calculating the current effective value, three-phase imbalance degree and other indicators, and if abnormal, timely alarm is generated, closed-loop monitoring and early warning is realized, multi-dimensional current abnormality judgment index ensures the comprehensiveness of abnormal identification, timely alarm mechanism enables the staff to respond quickly, reduces the loss of failure, and continuous monitoring ensures real-time control of the motor operation state;

[0120] The technical scheme of the embodiment of the application is: a sensor network is set, historical current sequences and historical working condition sequences of each motor are collected, abnormal working conditions and corresponding precursor working conditions are marked in the historical working condition sequences of each motor, a monitoring window is set, whether there is an abnormal working condition or a precursor working condition in the monitoring window is judged, if not, current prediction is triggered, a fusion prediction model is constructed, the fusion prediction model includes a time series current prediction model and a time series working condition prediction model trained based on the historical current sequences and the historical working condition sequences respectively, a motor group coupling model is established by acquiring basic information of each motor, each motor is predicted through the fusion prediction model, a predicted current sequence of each motor is obtained, the historical current sequence in the monitoring window and the predicted current sequence obtained by prediction are analyzed, current anomalies of the motor are analyzed and early warning is performed.

[0121] The basic principles, main features and advantages of the application are shown and described above. It should be understood by those skilled in the art that the application is not limited by the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the application, and various changes and improvements can be made without departing from the spirit and scope of the application, and these changes and improvements all fall within the scope of the claimed application. The scope of protection of the application is defined by the appended claims and their equivalents.

Claims

1. A current linkage early warning method for coal mine underground multi-motor coupling effect, characterized by: The application relates to a coal mine underground multi-motor linkage system current and working condition prediction method. A sensor network is arranged for the coal mine underground multi-motor linkage system to collect historical current sequences and historical working condition sequences of the motors; Abnormal working conditions and corresponding precursor working conditions are marked in the historical working condition sequences of the motors, a monitoring window is arranged, and it is judged whether abnormal working conditions or precursor working conditions exist in the monitoring window; if not, current prediction is triggered; The judgment method of whether abnormal working conditions or precursor working conditions exist is as follows: A monitoring window with a fixed time length and ending at the current time is arranged, and it is judged whether the monitoring window and a time period corresponding to any abnormal working condition coincide; if so, it is judged that abnormal working conditions exist in the monitoring window; If not, a plurality of precursor analysis time periods are arranged in the monitoring window, the precursor analysis time periods all end at the current time and have different time lengths, historical working condition sequences in all the precursor analysis time periods of the motor are intercepted, precursor analysis sequences of the motor are obtained and normalized, and normalized precursor analysis sequences are obtained; All the normalized precursor sequences of the precursor working conditions are obtained, and if the DTW distance between any normalized precursor analysis sequence in the monitoring window and any normalized precursor sequence is less than a distance threshold, it is judged that a precursor working condition exists in the monitoring window; If current prediction is triggered, a fusion prediction model is constructed, the fusion prediction model comprises a time series current prediction model and a time series working condition prediction model trained based on historical current sequences and historical working condition sequences respectively, and a motor group coupling model established based on motor basic information, and the motors are predicted through the fusion prediction model to obtain predicted current sequences of the motors; The historical current sequences in the monitoring window and the predicted current sequences obtained through prediction are analyzed, current abnormalities of the motors are analyzed, and early warning is performed.

2. The current linkage warning method for the coupling effect of multiple motors in a coal mine according to claim 1, characterized in that: The acquisition method of the historical current sequences and the historical working condition sequences is as follows: A sensor network is arranged for the coal mine underground multi-motor linkage system, motor parameters of the motors are collected in real time according to a set data collection frequency, time points of the sensor network are marked in real time according to the data collection frequency as collection time points, at each collection time point, three-phase stator current vectors and working condition data vectors of the motors are constructed; A time period between a time point when the sensor network starts to collect and the current time point is marked as a historical time period, for each motor, the three-phase stator current vectors and the working condition data vectors collected in the historical time period are integrated according to time sequences to obtain historical current sequences and historical working condition sequences of the motors.

3. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 2, characterized in that: The judgment method of the abnormal working conditions is as follows: Mean values and standard deviations of parameters of the working condition data vectors in the historical working condition sequences of the motors under normal working condition data ranges are obtained and calculated, dynamic abnormal threshold values are obtained through data processing, and if any parameter of the working condition data vector at any collection time point exceeds the set dynamic abnormal threshold value, the collection time point is given an abnormal time point mark including a motor number; For any motor number, if abnormal time point marks exist at continuous collection time points not less than KL, it is judged that the corresponding motor is in an abnormal working condition, the collection time points with continuous abnormal time point marks are defined as an abnormal working condition of the corresponding motor, and adjacent abnormal working conditions of the same motor with a time distance less than a preset time threshold are fused into one abnormal working condition.

4. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 2, characterized in that: The judgment manner of the precursor working condition is: Obtain all precursor sequence groups and contained normalized precursor sequences, and perform data processing on the number of normalized precursor sequences in the same precursor sequence group and the total number of abnormal sequences in the corresponding same type abnormal working condition to obtain a stability ratio; If the DTW distance between any two normalized precursor sequences not belonging to the same precursor sequence group meets the similarity standard, the two precursor sequence groups are classified into the same similarity set, the number of precursor sequence groups in the similarity set where the precursor sequence groups are located is obtained, and data processing is performed on the total number of all precursor sequence groups to obtain a similarity ratio; If the stability ratio is greater than a preset stability standard and the similarity ratio is less than a preset similarity standard, it is judged that the precursor sequence corresponding to the normalized precursor sequence in the precursor sequence group is a precursor working condition.

5. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 4, characterized in that: The acquisition manner of the precursor sequence group is: In the historical working condition sequence of the motor, abnormal sequences of the motor are obtained and normalized respectively to obtain normalized abnormal sequences and contained normalized working condition data vectors, and the DTW distance between the normalized abnormal sequences corresponding to any two abnormal sequences of the same motor is calculated, and if it is less than a distance threshold, the two abnormal sequences are the same type of abnormal working condition; In the historical working condition sequence of the motor, precursor sequences of the motor are obtained and normalized respectively to obtain normalized precursor sequences, and in the same type of abnormal working condition, the DTW distance between the normalized precursor sequences corresponding to any two abnormal sequences is calculated, and if it is less than a distance threshold, it is classified into the same precursor sequence group.

6. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 2, characterized in that: The acquisition manner of the predicted current sequence is: Obtain a time series current training set, calculate the root mean square error of the time series current prediction model and the motor group coupling model in predicting the time series current training set, set weights for the time series current prediction model and the motor group coupling model according to the root mean square error, Fuse the prediction results of the time series current prediction model and the motor group coupling model based on the set weights, that is, weight the time series predicted current sequence and the physical predicted current sequence by the calculated weights, to obtain the predicted current sequence of each motor.

7. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 6, characterized in that: The acquisition manner of the time series predicted current sequence is: All working condition data vectors and three-phase stator current vectors in all abnormal working conditions and precursor working conditions in the historical working condition sequence and the historical current sequence are removed respectively, the historical working condition sequence and the historical current sequence are supplemented by interpolation method and normalized to obtain a time series working condition training set and a time series current training set, a monitoring window is obtained and window working condition sequences and window current sequences are obtained by cutting; Using a long short-term memory network, using the time series working condition training set and the time series current training set, a time series working condition prediction model and a time series current prediction model are constructed and trained respectively, and combined with the window working condition sequence and the window current sequence, the time series predicted working condition sequence and the time series predicted current sequence of the motor are obtained.

8. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 7, characterized in that: The acquisition manner of the physical predicted current sequence is: The basic information of each motor is acquired, and the loss parameters of each motor are obtained through historical current sequence and historical working condition sequence statistics. Based on the motor dynamics equation and Kirchhoff's law, combined with the multi-motor coupling effect, a motor group coupling model is established, with the input being the working condition data vector of each motor and the output being the three-phase stator current vector of each motor. The time series prediction working condition sequence of each motor is input into the motor group coupling model to obtain the physical prediction current sequence of each motor.

9. The current linkage warning method for multi-motor coupling effect in underground coal mine according to claim 7, characterized in that: The current anomaly of the motor includes: The window current sequence and the prediction current sequence are obtained, the window current sequence includes a three-phase stator current vector, and the prediction current sequence includes a predicted three-phase stator current vector. The current effective value and the predicted current effective value are calculated through data processing. If the current effective value or the predicted current effective value satisfies the current overload abnormal condition at continuous not less than KL collection time points, it is determined that the corresponding motor has a current overload abnormality. Otherwise, the three-phase unbalance degree and the predicted three-phase unbalance degree of the motor are calculated. If the three-phase unbalance degree or the predicted three-phase unbalance degree is greater than the national standard critical value at continuous not less than KL collection time points, it is determined that the corresponding motor has a three-phase unbalance abnormality.

Images