CN114487816A - Abnormity detection method for photovoltaic tube type equipment conveying system - Google Patents

Abstract

The invention aims to disclose an abnormality detection method for a photovoltaic tube type equipment conveying system, which comprises the following steps: collecting operation big data of a transmission system; preprocessing the operation big data to construct a training data set for predicting the current value I of the motor; training the training data set through a machine learning algorithm to obtain a motor current value I prediction model; collecting real-time running current value R of motor_i(ii) a Inputting the real-time operation data of the transmission system into the prediction model, and mapping out the real-time predicted current value P of the motor_i(ii) a Setting a safety deviation delta of a predicted current value of the motor; if P_i‑R_iIf the value is greater than delta, the condition that the photovoltaic tube type equipment conveying system is abnormal is judged, and in the abnormal detection process, the upper limit current value, the lower limit current value and the actual current value change along with the change, so that the operation curve of the actual current is wrapped by the upper limit current value curve and the lower limit current value curve, and each time the motor is started, and the maximum current value and the minimum current value of the motor are calculated according to the actual current value, so that the motor is started, and the maximum current value and the minimum current value of the motor are calculated according to the actual current value and the actual current value of the motorIntelligent anomaly detection at one time.

Description

Abnormal detection method of photovoltaic tubular equipment conveying system

technical field

The invention relates to the technical field of photovoltaic tubular equipment control, in particular to an abnormality detection method of a photovoltaic tubular equipment transmission system.

Background technique

Photovoltaic production equipment (tubular processing technology) conveying system is used for handling semi-finished cells. This system exists in multiple process steps of a complete set of photovoltaic tubular equipment, such as cell diffusion, cell coating and other tubular processes. Conveyor systems are required that transport semi-finished cells from location A to location B. In the specific process of transporting semi-finished cells, the semi-finished cells are placed in the graphite boat of the handling tool, and the semi-finished cells need to be transported by means of a motor as a driving force, and the semi-finished cells are transported by means of conveyor belts. state.

In a relatively ideal state, the transmission system can transport the semi-finished cells according to the established control program and path, but in the actual operation process, the transmission system does not run at a uniform speed, and there are static, startup, acceleration, uniform speed, deceleration, etc. In a variety of working conditions, the conveyor system also has transport actions in the horizontal direction and the height direction. In the case of multiple factors, the conveyor system may encounter abnormalities, such as loose conveyor belts, foreign objects stuck in the conveyor belt, and two parallel conveyor belts are misplaced, etc. .

In the existing technical solution, in order to detect the abnormality encountered by the photovoltaic tubular equipment transmission system in time, various motors of the transmission system are abnormally detected. The specific method is to set a fixed threshold value for the motor current. The threshold value is generally the upper limit value. When the actual current value of the motor exceeds the fixed threshold value, it is determined that the motor is abnormal, so as to determine that the transmission system is abnormal; however, the actual operation of the motor in the actual transmission system is complex, and there are static, In this process, the actual current value of a certain motor varies greatly in different states. The practice of setting a fixed threshold is a one-size-fits-all approach. It usually selects the maximum operating state of the transmission system or the The minimum margin is used as the threshold. In tubular equipment, since the current value of the motor is very different in the running state, using the marginal value as the threshold will result in under-reporting because the threshold is too high in the low-current operating state, or the normal high current False alarms are generated during operation, so the fixed threshold method has been unable to be popularized and used, which leads to the lack of effective abnormal alarm methods in the existing photovoltaic tube equipment transmission system. Serious losses such as damage to components, splintered cells, downtime, etc.

In view of this, it is necessary to improve the abnormal detection method of the photovoltaic tubular equipment conveying system in the prior art to solve the above problems.

SUMMARY OF THE INVENTION

The purpose of the present invention is to disclose an abnormality detection method for a photovoltaic tubular equipment transmission system, which constructs operation big data based on the data of the operation physical characteristics of the transmission system, and uses a machine learning algorithm to analyze the training data set formed by the operation big data of the transmission system. Carry out training to obtain a prediction model, and realize the intelligentization of abnormal detection of photovoltaic tubular equipment transmission system.

In order to achieve the above purpose of the invention, the present invention provides an abnormality detection method for a photovoltaic tubular equipment transmission system, comprising the following steps:

Collect the operation big data of the conveying system;

Preprocessing the operating big data to construct a training data set for predicting the motor current value I;

The training data set is trained by the machine learning algorithm, and the prediction model of the motor current value I is obtained;

Collect the real-time running current value R _i of the motor;

Input the real-time operation data of the transmission system into the prediction model, and map out the real-time predicted current value P _i of the motor;

Set the safety deviation δ of the predicted current value of the motor;

If |P _i -R _i |>δ, it is determined that the photovoltaic tubular equipment transmission system is abnormal.

As a further improvement of the present invention, the preprocessing includes at least one of de-null value, de-duplication value or smooth noise data.

As a further improvement of the present invention, the smoothed noise data is to smooth the motor current noise through a Gaussian algorithm.

As a further improvement of the present invention, when the motor current is smoothed, the smoothing is performed in stages according to the running state of the motor.

As a further improvement of the present invention, the running state of the motor includes static state, startup, acceleration, equalization speed, and deceleration, and the static state is static unloading or static load.

As a further improvement of the present invention, it also includes the following steps: removing null values and duplicate values in the real-time operation data of the transmission system; removing transmission system failure data in the operation big data of the transmission system.

As a further improvement of the present invention, the machine learning algorithm is a random forest algorithm.

As a further improvement of the present invention, the big operating data and the real-time operating data are the actual position, set position, actual speed, set speed, load state, and position induction of the motor.

As a further improvement of the present invention, the operating big data and the real-time operating data are motor idle, enable signal, actual position, set position, actual speed, set speed, load status, and station induction.

As a further improvement of the present invention, the real-time operation data of the transmission system is incorporated into the training data set to obtain a new prediction model.

Compared with the prior art, the beneficial effects of the present invention are:

The operation big data is constructed based on the data of the operation physical characteristics of the conveying system, the training data set formed by the operation big data of the conveying system is trained by the machine learning algorithm, and the prediction model is obtained; then the real-time operation data of the conveying system is input into the prediction model, the real-time predicted current value of the motor is mapped; a safety deviation is set for the predicted current value of the motor. When the absolute value of the difference between the real-time predicted current value and the real-time actual current value exceeds the safety safety deviation, it is determined that the transmission system is abnormal; The actual current value of the motor changes when the motor accelerates, decelerates, and the load changes. The set current safety deviation will cause the real-time current to have an upper limit current value and a lower limit current value, and the upper limit current value and the lower limit current value are different from the actual current. The value changes with the change, so that the actual current running curve is wrapped by the upper limit current value curve and the lower limit current value curve, which is equivalent to setting the corresponding safety deviation at every moment of the motor operation, so as to realize the intelligentization of the motor at every moment. In the abnormality detection, compared with the existing fixed threshold method, in the process of abnormality detection, the present invention is equivalent to setting a corresponding safety deviation at every moment of motor operation, thereby realizing intelligent abnormality detection of the motor at every moment.

Description of drawings

FIG. 1 is a flowchart of an abnormality detection method of a photovoltaic tubular equipment transmission system according to the present invention;

FIG. 2 is a comparison diagram of the current value of the current safety domain of the present invention and the actual current sampling;

3 is a data comparison diagram before and after the Gaussian average of the motor current of the present invention;

Fig. 4 is the motor current of the present invention staged smooth and unstaged smooth comparison diagram;

Fig. 5 is the test curve of the fitting degree of predicted current value of the present invention;

Fig. 6 is the flow chart of the predicted current value P _i of the present invention;

Fig. 7 is the decision-making flow chart of the host computer of the present invention.

Detailed ways

The present invention will be described in detail below with reference to the various embodiments shown in the accompanying drawings, but it should be noted that these embodiments do not limit the present invention. Equivalent transformations or substitutions all fall within the protection scope of the present invention.

The specific implementation process of the present invention will be described below through a plurality of embodiments.

Example 1:

Referring to FIG. 1 , this embodiment discloses a specific implementation of an abnormality detection method (hereinafter referred to as “method”) of a photovoltaic tubular equipment conveying system.

Referring to FIG. 1, in this embodiment, the method includes the following steps S1 to S7. In this embodiment, the training data set formed by the operation big data of the transmission system is trained by the machine learning algorithm, and the prediction is obtained. Model to realize intelligent abnormal detection of photovoltaic tubular equipment transmission system.

Step S1, collecting the operation big data of the transmission system. Specifically, in the process of transporting semi-finished cells from position A to position B, the photovoltaic tubular equipment conveying system needs to use directions along the X, Y and Z axes (X and Z represent the horizontal direction, Y represents the vertical direction) The coordination of several motors in motion, such as motor X1, motor X2, motor X3, motor Y1, motor Y2, motor Y3, motor Z1, motor Z2, motor Z3, etc., through the collection of large data of the operation of the transmission system, especially the transmission system is not abnormal The operation big data under the circumstance, specifically, the operation big data of the transmission system includes the attribute data of the actual position, the set position, the actual speed, the set speed, the load state, the station induction and so on. The fixed position is the loading position information, the actual speed and the set speed are the loading speed information, and the load status and station induction are the loading weight information. The collected operating big data of the conveying system can represent various working conditions encountered by the conveying system.

Step S2, preprocessing the operating big data to construct a training data set for predicting the motor current value I. Specifically, preprocessing is performed on the motor's running big data, and the preprocessing includes at least one of de-null value, de-duplication value or smooth noise data. As shown in Table 1, the motor's running data before and after de-null value The data sheet, Table 2 is the big data sheet of the operation of the motor before and after the deduplication value.

Table 1 Big data table of the operation of the motor before and after removing the null value

Table 2 Big data table of motor operation before and after deduplication

It needs to be further explained that the servo system existing in the transmission system controls the current value of the motor to make the load run to the specified position at the specified speed. The load of the motor is also constantly changing during the operation of the transmission system, which requires the servo system to repeatedly adjust the current value. The load is in the process of transmission, and the load is transferred and handed over. When the load is between the two motors, the motor runs. The state is constantly changing, and the current value of the motor fluctuates greatly, that is, the current value has a large noise. In order to reduce the noise of the current value, the current is smoothed by the Gaussian algorithm, and the comparison chart of the current value before and after smoothing is shown in Figure 3.

The specific method of smoothing the current Gaussian algorithm is to take the point value x of the current as the input, and output G(x) with the weight of the current. The specific formula is as follows:

Multiply G(x) by the point value x of the current to get the smoothed current value.

When the motor current is smoothed, the smoothing is performed in stages according to the running state of the motor. The running state of the motor includes static state, startup, acceleration, average speed, and deceleration, and the static state is static unloading or static load. That is to say, according to the different operating states of the motor, the current value fluctuates during the changing process of the motor operating state. In order to reduce the influence of the current noise, the current noise of each operating state of the motor is smoothed, and the data before and after the Gaussian average of the motor current is processed. See Figure 3 for the comparison diagram; during the start or acceleration stage of the motor, the current value fluctuates greatly, see Figure 3; the smoothing is performed in stages, and the inflection points of different motor operation stages are used as the dividing points, such as inflection point 1 and inflection point 2. Smoothing is performed in stages, so that the smoothing effect is better and closer to the actual situation, so as to avoid noise reduction and smoothing at the inflection point when it is not smoothed in stages, see Figure 4 for details. The actual physical scene, which provides the basis for the abnormal detection of the motor current at each subsequent moment.

It should be further explained that, in order to ensure the accuracy of constructing the training data set for predicting the motor current value I, in addition to the above-mentioned preprocessing process, it is also necessary to analyze the fault data of the transmission system in the operation big data of the motor. It is removed to provide an accurate big data basis for the subsequent construction of predictive models.

In step S3, the training data set is trained by the machine learning algorithm, and the prediction model of the motor current value I is obtained. Specifically, the machine learning algorithm is the random forest algorithm, which is one of the most powerful supervised learning algorithms, which takes into account the ability to solve regression problems and classification problems; In the regression problem, the output of each decision tree is averaged to obtain the final regression result; the characteristic of the random forest algorithm is that the larger the number of decision trees, the more robust the random forest algorithm is. The stronger it is, the higher the accuracy; therefore, the more big data on the operation of the motor is collected by the training data set, the higher the prediction reliability of the prediction model will be.

Step S4, collecting the real-time operating current value R _i of the transmission system. Specifically, the real-time running current value R _i of the transmission system will deviate when the transmission system is abnormal _{. Therefore, the current value R i is} the key detection data of each motor, but the current The value _Ri is constantly changing at every moment, especially during the start-up, acceleration and deceleration stages of the motor. Real-time detection of the abnormality of the current value _Ri is _crucial to ensuring the stable operation of the entire conveying system. important role.

Step S5 , input the real-time operation data of the transmission system into the prediction model, and map out the real-time predicted current value P _i of the motor. Specifically, the real-time operation data of the transmission system includes attribute data of 6 aspects, such as actual position, set position, actual speed, set speed, load state, and station induction. For the null value in the real-time operation data of the motor After the repeated values are removed, the real-time operation data of the transmission system is input into the prediction model obtained in step S3, and the real-time predicted current value P _i of the motor is intelligently mapped according to the training results of the machine learning algorithm in the previous stage.

Step S6, setting the safety deviation δ of the predicted current value of the motor. Specifically, during the operation of the motor, the safety deviation δ of the predicted current value is set for the motor at each moment, so as to realize the abnormal detection of each time point in the operation of the motor during the entire operation of the motor.

Step S7, if |P _i -R _i |>δ, it is determined that the photovoltaic tubular equipment transmission system is abnormal. Specifically, referring to FIG. 7 , the predicted current value P _i , the real-time operating current value R _i , and the safety deviation δ of the predicted current value are all input to the host computer, and the host computer judges whether an abnormality occurs. The specific judgment process is as follows: when |P _i − When R _i |>δ, that is, the deviation between the actual current value R _i and the predicted current value P _i exceeds the safety deviation δ, it is judged as abnormal, otherwise, it is normal; the safety deviation δ is in different operation stages of the motor , different safety deviations can be set, and the size of the safety deviation δ can be set according to the safety level.

Referring to Fig. 5, Fig. 5 is the test curve of the predicted current value conformity of the present invention. The perfect prediction is an ideal curve in the AI prediction. The closer the final predicted current value is to the perfect prediction curve, the more perfect the prediction result is. The figure is the prediction result without the Gaussian algorithm smoothing on the current noise. It shows that a considerable part of the prediction results will deviate from the perfect prediction curve. The right picture in Figure 5 is the prediction result after the current noise is smoothed by the Gaussian algorithm. The fit of the perfect prediction curve of the prediction results is greatly improved.

The beneficial effects of this embodiment are: the training data set formed by the operation big data of the transmission system is trained by the machine learning algorithm to obtain a prediction model; then the real-time operation data of the transmission system is input into the prediction model, and the motor's real-time operation data is mapped out. Real-time predicted current value P _i ; set a safety deviation δ for the predicted current value of the motor, when the absolute value of the difference between the real-time predicted current value P _i and the real-time actual current value R _i exceeds the safety deviation, it is determined that the transmission system is abnormal; The actual current value of the motor changes when the motor accelerates, decelerates, and the load changes. The set current safety deviation will cause the real-time current to have an upper limit current value and a lower limit current value, and the upper limit current value and the lower limit current value are the same as the actual current. The value changes with the change, so that the actual current running curve is wrapped by the upper limit current value curve and the lower limit current value curve, the upper limit current value curve P _i =R _i +δ, the lower limit current value P _i =R _i -δ, which is equivalent to the motor A corresponding safety deviation is set at each moment of operation, so as to realize the intelligent abnormal detection of the motor at each moment. Compared with the existing fixed threshold method, the present invention in the abnormal detection process is equivalent to every time the motor is running. The corresponding safety deviation is set at one time, so as to realize the intelligent abnormal detection of the motor at each time; This embodiment only needs to provide 24h-48h of motor operation big data, so that a training data set can be formed and the requirements of the prediction model can be met, and then an abnormality detection method of the transmission system can be established.

Embodiment 2:

The difference from the first embodiment is that, in order to build a more accurate prediction model, the real-time operation data of the conveying system includes six aspects, such as actual position, set position, actual speed, set speed, load status, and station sensing. On the basis of the attribute data of the motor, add two aspects of attribute data such as motor idle and enable signal, that is, the training data set is composed of the attribute data of the above 8 aspects, and the training data set is trained by the machine learning algorithm. The prediction model is obtained; idle is Boolean data, representing the static or moving state of the load; the enable signal is also Boolean data, representing whether the motor is started or not started, these two motor attributes can describe the dynamic and static state of the load. When some motors in the transmission system are frequently started and stopped, by adding the idle and enable signals of the motor, the running state of the motor can be more comprehensively characterized, and the motor attribute data contained in the training data set can be more comprehensive, so as to make the construction of the motor more comprehensive. The motor current prediction model accurately predicts the current.

It should be further explained that, referring specifically to FIG. 6, FIG. 6 is a flow chart of obtaining the predicted current value P _i in real time in the present embodiment. The machine learning algorithm requires a large number of training data sets as input, that is, the more data sources of the training data set, the more , the higher the prediction accuracy of the obtained prediction model, therefore, the real-time operation data of the transmission system is merged into the training data set, that is, the real-time operation data of each transmission system will be stored in the training data Before storing the training data set, it will also undergo preprocessing such as removing null values, removing duplicate values, smoothing noise data, removing fault data, etc., for the machine learning algorithm to continue to learn, and then a new prediction model can be obtained. The updated prediction model predicts the motor current value. The advantage of this continuous iteration is that it helps to meet the needs of the prediction model with less initial data training set. The initial training data set in this embodiment only needs 24h-48h The operating data can meet the needs of the predictive model.

It should be noted that the current in the present invention refers to the ratio of the actual current of the motor to the rated current of the motor.

The series of detailed descriptions listed above are only specific descriptions for the feasible embodiments of the present invention, and they are not used to limit the protection scope of the present invention. Changes should all be included within the protection scope of the present invention.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, but that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the invention is to be defined by the appended claims rather than the foregoing description, which are therefore intended to fall within the scope of the claims. All changes within the meaning and scope of the equivalents of , are included in the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim.

In addition, it should be understood that although this specification is described in terms of embodiments, not each embodiment only includes an independent technical solution, and this description in the specification is only for the sake of clarity, and those skilled in the art should take the specification as a whole , the technical solutions in each embodiment can also be appropriately combined to form other implementations that can be understood by those skilled in the art.

Claims (10)

1. The anomaly detection method of the photovoltaic tubular equipment conveying system is characterized by comprising the following steps of:

collecting operation big data of a transmission system;

preprocessing the operation big data to construct a training data set for predicting the current value I of the motor;

training the training data set through a machine learning algorithm to obtain a prediction model of the motor current value I;

collecting real-time running current value R of motor_i；

Inputting the real-time operation data of the transmission system into the prediction model, and mapping out the real-time predicted current value P of the motor_i；

Setting a safety deviation delta of a predicted current value of the motor;

if P_i-R_iIf the value is greater than delta, the abnormity of the photovoltaic tube type equipment conveying system is judged.

2. The method of anomaly detection for a photovoltaic tubular equipment transport system of claim 1, wherein said preprocessing comprises at least one of a null value, a repetition value or a smooth noise data.

3. The method of claim 2, wherein the smoothing noise data is obtained by smoothing the motor current noise by a gaussian algorithm.

4. The abnormality detection method for a photovoltaic tube system according to claim 3, wherein smoothing is performed in stages according to an operation state of a motor when smoothing the motor current.

5. The method for detecting the abnormality of the transport system of the photovoltaic tubular equipment according to claim 4, wherein the operation state of the motor includes a static state, a starting state, an accelerating state, a speed equalizing state and a decelerating state, and the static state is a static unloading state or a static loading state.

6. The method for detecting the abnormality of the transport system of the photovoltaic tube type equipment according to claim 1, further comprising the steps of removing null values and repetition values in real-time operation data of the transport system; and removing fault data of the transmission system in the operation big data of the transmission system.

7. The method for anomaly detection of a photovoltaic tubular conveyor system according to any one of claims 1-6, wherein said machine learning algorithm is a random forest algorithm.

8. The method for detecting the abnormality of the transport system of the photovoltaic tubular equipment according to any one of claims 1 to 6, wherein the operation big data and the real-time operation data are an actual position, a set position, an actual speed, a set speed, a load state, and a station sensing of a motor.

9. The method for detecting the abnormality of the transport system of the photovoltaic tubular equipment according to any one of claims 1 to 6, wherein the operation big data and the real-time operation data are idle of a motor, an enable signal, an actual position, a set position, an actual speed, a set speed, a load state, a station sensing.

10. The method for detecting anomalies in a photovoltaic tubular equipment conveyor system according to any one of claims 1 to 6, characterized in that real-time operating data of the conveyor system is incorporated into the training data set to obtain a new predictive model.

Images