Photovoltaic system power quality monitoring and fault diagnosis method based on deep learning

Abstract

The invention relates to the technical field of photovoltaic system fault diagnosis, and provides a photovoltaic system power quality monitoring and fault diagnosis method based on deep learning. The method comprises the following specific steps: collecting three-phase electric power parameters of a photovoltaic system, and preprocessing the collected three-phase electric power parameter data to construct a training data set; constructing a bidirectional long-short-term memory network Bi-LSTM model, training the model by using the preprocessed training data set, and predicting the fault type of the photovoltaic module; inputting three-phase power parameters collected in real time into the trained Bi-LSTM model, and outputting the fault type and fault probability distribution of the photovoltaic module; comparing a model prediction result with an actual experiment result, and judging whether the recognition accuracy meets a preset requirement or not; and if the accuracy is lower than the requirement, the parameter optimization model is iteratively updated until the accuracy of the model output structure meets the preset requirement.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic system fault diagnosis, in particular to a photovoltaic system power quality monitoring and fault diagnosis method based on deep learning. BACKGROUND

[0002] Grid-connected photovoltaic system refers to a system that converts direct current generated by solar photovoltaic components into alternating current through an inverter and connects it to the public power grid. The presence of the inverter will generate harmonics in the grid-connected process of the photovoltaic system. The inverter contains a large number of power electronic components, and the nonlinear characteristics of these components will cause waveform distortion, thereby generating harmonics. In addition, the delay of inverter switching speed will also affect the overall dynamic performance output of the grid system, generating small-range harmonics. In the case of large changes in weather (such as irradiance, temperature), the fluctuation range of harmonics will also become large.

[0003] In order to reduce the harm caused by harmonics, it is necessary to monitor the power quality of the photovoltaic system and reduce the impact of harmonics through a series of means such as installing harmonic filters, reducing nonlinear loads, and improving power factors. The aforementioned methods generally require the installation of power quality monitors to monitor power quality in real time and take appropriate action after harmonics occur. The response speed of the device directly affects the stability of the power grid. Therefore, there is an urgent need for a method to diagnose and predict possible faults in the photovoltaic system to improve the response speed of the corresponding device. SUMMARY

[0004] Embodiments of the present application provide a hemodialysis patient care management system, method, computer readable medium and electronic device, which can at least partially solve the problem of ensuring the privacy and security of patient information and physiological data during data transmission.

[0005] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the present application.

[0006] According to one aspect of the present application, a photovoltaic system power quality monitoring and fault diagnosis method based on deep learning is provided, comprising the following specific steps:

[0007] S1: Power quality data acquisition and preprocessing; real-time acquisition of three-phase electric power parameters of the photovoltaic system through a three-phase electric parameter measurement module, and preprocessing of the acquired three-phase electric power parameter data to construct a training data set;

[0008] S2: Construction and training of deep learning model; a bidirectional long short-term memory network Bi-LSTM model is constructed, and the model is trained using the preprocessed training data set to predict the fault type of the photovoltaic component;

[0009] S3: Real-time monitoring and fault diagnosis: input the real-time collected three-phase electric power parameters into the trained Bi-LSTM model, and output the fault type and fault probability distribution of the photovoltaic module;

[0010] S4: Model verification; compare the model prediction results with the actual experimental results to determine whether the recognition accuracy meets the preset requirements; if not, optimize the parameters and update the model iteratively until the accuracy of the model output structure meets the preset requirements.

[0011] In the present application, based on the foregoing scheme, the ZDT-3561 flat plate three-phase power quality analyzer is used in step S1 to collect the three-phase electric power parameters of the photovoltaic system; the three-phase electric power parameters include voltage, current and power factor.

[0012] In the present application, based on the foregoing scheme, the windowed interpolation FFT algorithm is used to analyze the harmonic of the current and voltage signals, and the total harmonic distortion value of the voltage, the total harmonic distortion value of the current, and the time series data set containing the harmonic current effective value, voltage distortion rate, power factor, and active power are calculated.

[0013] In the present application, based on the foregoing scheme, in the bidirectional long short-term memory network Bi-LSTM model constructed in step S2, the time series data set is used as the training data set; the input layer receives the time series data set, the hidden layer contains at least 3 LSTM units, and the output layer is connected to a fully connected network; the grid search method is used to optimize the model hyperparameters, and the learning rate is set to 0.001-0.01, the batch size is set to 32-128, and the LSTM unit number is set to 64-256 to minimize the prediction error.

[0014] In the present application, based on the foregoing scheme, the training data set in step S2 includes normal working condition time series data set and fault working condition time series data set; the fault working condition time series data set includes PID attenuation fault, hot spot fault and loose connection fault.

[0015] In the present application, based on the foregoing scheme, the PID attenuation fault is manifested as a decrease of ≥10% in the open circuit voltage of the photovoltaic module;

[0016] The hot spot fault is manifested as an abnormal increase in local current and a temperature difference of ≥15%;

[0017] The loose connection fault is manifested as a sudden increase of ≥5% in the current harmonic distortion rate.

[0018] In the present application, based on the foregoing scheme, in step S3, an alarm threshold is preset, and if the predicted fault probability exceeds the alarm threshold, an alarm signal is generated and the fault type and fault location are located.

[0019] In the present application, based on the foregoing scheme, comprising: using confusion matrix to evaluate fault classification accuracy in step S4, the calculation formula is:

[0020]

[0021] Wherein, TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

[0022] The second aspect of the present application provides a computer readable medium having a computer program stored thereon, the computer program being executed by a processor to implement the above-mentioned deep learning-based photovoltaic system power quality monitoring and fault diagnosis method.

[0023] The third aspect of the present application provides an electronic device, comprising:

[0024] One or more processors;

[0025] A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned deep learning-based photovoltaic system power quality monitoring and fault diagnosis method.

[0026] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the deep learning-based photovoltaic system power quality monitoring and fault diagnosis method provided in the various optional implementation manners.

[0027] In the technical scheme of the present application, by monitoring the three-phase electric energy parameters in the system operation process, a bidirectional long short-term memory network Bi-LSTM model is constructed, a training set and a verification set are constructed using actual operation parameters to train the model, and the change trend of each electric energy parameter in a certain time before the fault occurs is analyzed. Combined with the operation law of the measured data, the probability of fault position and fault type can be predicted. When the probability exceeds the threshold value, an alarm is given, and at the same time, a suitable method is selected according to the fault reason to eliminate the fault, which greatly improves the response speed of the fault and improves the stability of the photovoltaic grid-connected system. At the same time, the model has a self-optimization function and is optimized regularly to ensure the accuracy of the prediction results.

[0028] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0029] The accompanying drawings, which are incorporated herein and constitute part of this specification, illustrate embodiments consistent with the application and, together with the description, further serve to explain the principles of the application. It is to be understood that the drawings are only schematic, and that they do not necessarily correspond to the actual relative sizes of the components. It is also to be understood that the examples depicted in the drawings are only some embodiments of the application and that many other embodiments of the application can be made without departing from the scope thereof.

[0030] Figure 1 A flow chart of a deep learning based photovoltaic system power quality monitoring and fault diagnosis method in an embodiment of the present application is schematically shown.

[0031] Figure 2 A schematic diagram of a deep learning based photovoltaic system power quality monitoring and fault diagnosis system in an embodiment of the present application is schematically shown.

[0032] Figure 3 A structural schematic diagram of a computer system of an electronic device suitable for implementing embodiments of the present application is shown. DETAILED DESCRIPTION

[0033] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the implementations set forth herein; rather, these implementations are provided as non-limiting examples so that this disclosure will be thorough and complete, and will fully convey the scope of the example implementations to those skilled in the art. Like reference numerals may refer to like elements throughout the description.

[0034] Moreover, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the application. One skilled in the relevant art will recognize, however, that the

[0035] The block diagrams in the drawings show only the functionality of the embodiments and do not imply that the functions must be implemented in a particular order. For example, the functions can be implemented in reverse order or in stages. Also, the block diagrams do not necessarily show all of the functions of the embodiments.

[0036] The flow diagrams depicted herein are merely examples and that numerous variations and modifications thereto can be made that fall within the scope of the present application. For example, the order of the blocks can be rearranged, combined, separated into multiple blocks or placed into various other sequences. Also, many of the blocks can be called indirectly, and one or more blocks can be omitted, depending on the circumstances.

[0037] The implementation details of the technical solutions of the application are described in detail as follows:

[0038] Figure 1 A flowchart of a deep learning-based photovoltaic system power quality monitoring and fault diagnosis method according to an embodiment of the application is shown. Referring to Figure 1 The deep learning-based photovoltaic system power quality monitoring and fault diagnosis method includes at least steps S1 to S4, which are described in detail as follows:

[0039] A deep learning-based photovoltaic system power quality monitoring and fault diagnosis method includes the following specific steps:

[0040] S1: Power quality data acquisition and preprocessing; real-time acquisition of three-phase electric power parameters of the photovoltaic system through a three-phase electric parameter measurement module, and preprocessing of the acquired three-phase electric power parameter data to construct a training data set;

[0041] In step S1, a ZDT-3561 flat plate three-phase power quality analyzer is used to acquire the three-phase electric power parameters of the photovoltaic system; the three-phase electric power parameters include voltage, current and power factor. The active power, reactive power and apparent power can be calculated from the acquired data; the calculation formula is as follows:

[0042] Active power (P): P= ×U×I×cos(ϕ)

[0043] Reactive power (Q): Q= ×U×I×sin(ϕ)

[0044] Apparent power (S): S= ×U×I

[0045] Wherein, U is the line voltage, I is the line current, and ϕ is the power factor angle;

[0046] Then, a windowed interpolation FFT algorithm can be used to analyze the harmonics of the current and voltage signals, calculate the total harmonic distortion value of the voltage, the total harmonic distortion value of the current, and the time series data set containing the harmonic current effective value, voltage distortion rate, power factor, and active power.

[0047] The specific steps of the windowed interpolation FFT algorithm are as follows:

[0048] A Hanning window function is applied to the current and voltage signals to suppress spectral leakage;

[0049] A cubic spline interpolation is used to correct the spectral resolution to extract the effective value of the fundamental wave to the 50th harmonic;

[0050] The voltage THD and the current THD are calculated, and the formula is as follows:

[0051]

[0052] wherein V n and I n are the effective values of the nth harmonic, and V1 and I1 are the effective values of the fundamental wave.

[0053] When using this scheme, windowing is used to reduce spectral leakage and the fence effect in the FFT algorithm, and the original signal needs to be windowed. Commonly used window functions include the Hanning window, the Hamming window, the Blackman window, etc.; the FFT algorithm, i.e., the fast Fourier transform, is used to process the windowed signal to obtain the frequency spectrum information of the signal; interpolation is needed because the resolution of the FFT algorithm is limited by the number of sampling points and the sampling frequency, so the frequency spectrum needs to be interpolated to improve the resolution and accuracy of the frequency spectrum.

[0054] Based on the above-mentioned collected and calculated parameters, a time series dataset is constructed; the time series dataset includes multidimensional data, and each different data type is one-dimensional time series data.

[0055] S2: Construction and training of a deep learning model; a bidirectional long short-term memory network Bi-LSTM model is constructed, the model is trained using the preprocessed training dataset, and the fault type of the photovoltaic module is predicted;

[0056] In the bidirectional long short-term memory network Bi-LSTM model constructed in step S2, the time series dataset is used as the training dataset; the input layer receives the time series dataset, the hidden layer contains at least 3 LSTM units, and the output layer is connected to a fully connected network; the grid search method is used to optimize the model hyperparameters, the learning rate is set to 0.001-0.01, the batch size is set to 32-128, the number of LSTM units is set to 64-256, and the prediction error is minimized.

[0057] The training dataset in step S2 includes a normal working condition time series dataset and a fault working condition time series dataset; the fault working condition time series dataset includes three fault types of PID attenuation fault, hot spot fault, and loose wiring fault.

[0058] The characteristics of the three fault types are as follows:

[0059] The PID attenuation fault is manifested as a decrease in the open-circuit voltage of the photovoltaic module of ≥10%;

[0060] The hot spot fault is manifested as an abnormal increase in local current and a temperature difference of ≥15%;

[0061] The loose wiring fault is manifested as a sudden increase in current harmonic distortion rate of ≥5%.

[0062] In the conventional scheme, the fault type is determined by detecting the changes of the above key parameters, which inevitably has the problem of slow response. The innovation of the present scheme is to combine the monitoring data with the bidirectional long short-term memory network model, and predict the probability of failure based on the preceding monitoring data to take corresponding protective measures in advance.

[0063] S3: Real-time monitoring and fault diagnosis: input the real-time collected three-phase electric power parameters into the trained Bi-LSTM model, and output the fault type and fault probability distribution of the photovoltaic module;

[0064] In the present application, based on the foregoing scheme, an alarm threshold is preset in step S3, and if the predicted failure probability exceeds the alarm threshold, an alarm signal is generated and the fault type and fault location are located.

[0065] Further, the alarm threshold can be set to 85%, and when the predicted result output by the model is higher than 85%, an alarm signal is generated and the component with the fault is accurately located, and the appropriate way is selected according to the fault type to eliminate the influence of the fault.

[0066] S4: Model verification; compare the model prediction result with the actual experimental result to determine whether the recognition accuracy meets the preset requirement; if not, optimize the parameters and update the model iteratively until the accuracy of the model output structure meets the preset requirement.

[0067] In step S4, the confusion matrix is used to evaluate the fault classification accuracy, and the calculation formula is:

[0068]

[0069] Among them, TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

[0070] Example 1

[0071] The present scheme will be described in detail using a specific case.

[0072] A 10MW distributed photovoltaic system arranged in a certain province in North China was tested, which included 1200 components, and the operating environment temperature was-20-40 degrees Celsius.

[0073] S1: Power quality data acquisition and preprocessing; the three-phase electric power parameters of the photovoltaic system are collected in real time by the three-phase electric parameter measurement module, and the collected three-phase electric power parameter data are preprocessed to construct a training data set;

[0074] Firstly, ZDT-3561 flat three-phase power quality analyzer is used to collect data of components arranged in the photovoltaic field; the device supports voltage measurement range 0-1000V (accuracy ±0.1%), current measurement range 0-100A (accuracy ±0.2%), sampling frequency 10kHz; the collected three-phase electric power parameters include voltage, current and power factor; the obtained data can be used to calculate active power, reactive power and apparent power; the calculation formula is as follows:

[0075] Active power (P): P= ×U×I×cos(ϕ)

[0076] Reactive power (Q): Q= ×U×I×sin(ϕ)

[0077] Apparent power (S): S= ×U×I

[0078] Wherein, U is line voltage, I is line current, and ϕ is power factor angle;

[0079] Then, windowed interpolation FFT algorithm can be used to analyze harmonics of current and voltage signals, and to calculate voltage total harmonic distortion value, current total harmonic distortion value, and time series data set containing harmonic current effective value, voltage distortion rate, power factor and active power.

[0080] The specific steps of windowed interpolation FFT algorithm are as follows:

[0081] Hanning window function is applied to current and voltage signals to suppress spectrum leakage;

[0082] Cubic spline interpolation is used to correct spectrum resolution to extract effective value of fundamental wave to 50th harmonic;

[0083] Voltage THD and current THD are calculated, and the formula is as follows:

[0084]

[0085] Wherein, V n and I n are nth harmonic effective value, V1 and I1 are fundamental wave effective value.

[0086] When using the scheme, windowing is used to reduce spectral leakage and the fence effect in the FFT algorithm, and the original signal needs to be windowed. Commonly used window functions include Hanning window, Hamming window, Blackman window, etc.; the FFT algorithm is a fast Fourier transform, which is used to process the windowed signal to obtain the frequency spectrum information of the signal; interpolation is due to the limited resolution of the FFT algorithm by the number of sampling points and the sampling frequency, so the frequency spectrum needs to be interpolated to improve the resolution and accuracy of the frequency spectrum.

[0087] Based on the above-mentioned collected and calculated parameters, a time series data set is constructed; the time series data set includes multi-dimensional data, and each different data type is one-dimensional time series data. The sampling interval is set to 1 minute, and the running data of the photovoltaic field is continuously collected for 30 days to establish a time series data set with sufficient data.

[0088] S2: Construction and training of a deep learning model; a bidirectional long short-term memory network Bi-LSTM model is constructed, and the model is trained using the preprocessed training data set to predict the fault type of the photovoltaic component;

[0089] In the bidirectional long short-term memory network Bi-LSTM model constructed in step S2, the time series data set is used as the training data set, and specifically:

[0090] The network architecture of the model includes an input layer, a hidden layer, and an output layer;

[0091] The input layer receives 15-dimensional time series data including (THD V , THD I , PF, P, Q, V1-V5, I1-I5); to reduce the dimension of the input layer, only the effective voltages and currents of the fundamental wave to the 5th harmonic are taken, and in the case of meeting the calculation amount, the data amount of the harmonic effective voltage and current can also be appropriately increased or reduced.

[0092] The hidden layer is set to 3 layers of bidirectional LSTM, with 128 units per layer and a Dropout rate of 0.2;

[0093] The output layer is used for positioning and classifying the fault type and outputting the fault probability.

[0094] The output layer is connected to a fully connected network; the grid search method is used to optimize the model hyperparameters, and the learning rate (0.001, 0.005, 0.01), batch size (32, 64, 128), and LSTM unit number (64, 128, 256) are set to minimize the prediction error. The optimal parameter combination is a learning rate of 0.005, a batch size of 64, and an LSTM unit number of 128;

[0095] The training data set in step S2 includes a normal working condition time series data set and a fault working condition time series data set; the fault working condition time series data set includes three fault types of PID attenuation fault, hot spot fault and loose connection fault. The following three common fault types are studied in detail:

[0096] The characteristics of the three fault types are as follows:

[0097] The PID attenuation fault is manifested as a decrease of ≥10% in the open circuit voltage of the photovoltaic module;

[0098] The hot spot fault is manifested as an abnormal increase in local current and a temperature difference of ≥15%;

[0099] The loose connection fault is manifested as a sudden increase of ≥5% in the current harmonic distortion rate.

[0100] PID effect, also known as potential-induced degradation, refers to the phenomenon of power attenuation of a solar cell under a certain external voltage for a long time; the main causes are: 1. High voltage: with the large-scale application of photovoltaic systems, the system voltage is getting higher and higher, and usually multiple cell modules need to be connected in series to reach the maximum power point tracking working voltage of the inverter, resulting in higher open circuit voltage and working voltage.

[0101] 2. Packaging problem: during the packaging process of the module, due to the incomplete insulation of EVA material, especially in a humid environment, water vapor can enter the inside of the module through silicone or backboard, resulting in a leakage current phenomenon of the packaging material to the battery insulation system.

[0102] 3. Ion drift: in a humid environment, the ester bond of EVA will decompose to produce freely movable acetic acid, which will react with the alkali on the surface of the glass to produce sodium ions. Under the action of an external electric field, sodium ions will move to the surface of the battery sheet and accumulate in the anti-reflective layer, thereby causing the occurrence of PID phenomenon.

[0103] In actual application, the existence of harmonics will affect the stability of the system voltage and indirectly affect the PID effect, which is manifested as a decrease of ≥10% in the open circuit voltage of the photovoltaic module.

[0104] Hot spot fault refers to local overheating on the surface of a photovoltaic module, which is caused by many factors; for example, shading caused by bird droppings, dust and other shading objects; cell faults, diode faults and other hot spot defects. Due to the influence of harmonics on the stability of system current and voltage, it indirectly affects the heat dissipation performance of the photovoltaic module, thereby increasing the risk of hot spot fault; when hot spot fault occurs, it is manifested as an abnormal increase in local current and a temperature difference of ≥15%.

[0105] Loose connection failure refers to the failure caused by the loose connection of cables, connectors and other components in the photovoltaic system. The main causes include:

[0106] 1. Installation problem: the cables, connectors and other components are not fastened in place during installation.

[0107] 2. Long-term use: the connection is loose due to vibration, aging and other reasons during long-term use.

[0108] Loose connection failure can cause instability of system current and voltage, resulting in harmonics. In addition, the harmonics generated by the operation of the inverter can also exacerbate the risk of loose connection failure, as the harmonics can cause fluctuations in system current and voltage, increasing the possibility of loose and damaged connection components; in practical applications, the loose connection failure is manifested as a sudden increase in current harmonic distortion rate ≥5%.

[0109] In the conventional scheme, the fault type is determined by detecting the changes in the above key parameters, which inevitably has the problem of slow response; the innovation of this scheme is to combine the monitoring data with the bidirectional long and short term memory network model, and based on the previous monitoring data to predict the probability of failure, so as to take corresponding protective measures in advance. For example, the conventional monitoring method monitors that the current harmonic distortion rate increases by ≥5%, and then determines that the connection is loose; this scheme monitors the current harmonic distortion rate at the current time and in the previous period, when the harmonic distortion rate gradually increases and remains at about 3%, according to the growth rate of the harmonic distortion rate, the time when the harmonic distortion rate reaches 5% is predicted, the failure risk probability is estimated according to the change trend of the harmonic distortion rate, and the monitoring data in the future time period is compared to dynamically adjust the value of the failure risk probability; for example, the current time is t0, the failure occurs at t4 according to the current change trend (i.e. p4 is greater than 85%), the output result is (t1, p1); (t2, p2); (t3, p3); (t4, p4); (t n , p n ). t n is the nth time, p n is the probability of failure at that time; with continuous monitoring, the failure probability at t n is dynamically adjusted, when n<5 and p n >85%, an alarm is given.

[0110] In this scheme, the data with faults is labeled in advance; the data collected in step S1 is used to train the model; the normal working condition data is 30000, accounting for 70%; the fault data accounts for 30%, the proportion of different types of fault data can be adjusted according to the actual running failure probability of the current photovoltaic field, optionally, 1000 PID attenuation, 800 hot spots, 1200 loose connections.

[0111] The output result of the model is verified using the verification data set, and an accuracy higher than 95% is considered as the completion of model training; the fault type can be predicted by combining the above-mentioned indexes with the operation change rule of the electric energy parameters in a period of time before the fault occurs; the fault point can be determined by the position of the data collected by the three-phase electric parameter measurement module; the model predicts the fault type and fault point in the future T (T takes 1-5s) time according to the monitoring data of the current period; the accuracy of fault type prediction is negatively correlated with the time T.

[0112] S3: Real-time monitoring and fault diagnosis: input the real-time collected three-phase electric power parameters into the trained Bi-LSTM model, and output the fault type and fault probability distribution of the photovoltaic module;

[0113] An alarm threshold is preset in step S3, and if the predicted fault probability exceeds the alarm threshold, an alarm signal is generated and the fault type and fault position are located. Further, the alarm threshold can be set to 85%; when the predicted result output by the model is higher than 85%, an alarm signal is generated and the component with fault is accurately located, and the appropriate way is selected according to the fault type to eliminate the influence of the fault.

[0114] S4: Model verification; compare the model prediction result with the actual experimental result to determine whether the recognition accuracy meets the preset requirement; if not, optimize the parameters and update the model iteratively until the accuracy of the model output structure meets the preset requirement. The actual result can be calculated using the monitoring in step S1, and whether the fault occurs during the continuous operation of the equipment is observed to evaluate the accuracy of fault classification.

[0115] In an optional embodiment, the confusion matrix is used to evaluate the fault classification accuracy in step S4, and the calculation formula is:

[0116]

[0117] Wherein, TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

[0118] After the model runs for a period of time, the accuracy may decrease due to the influence of seasons and weather; when the fault classification accuracy is lower than 95%, the automatic optimization program is started; 10% of the latest data in the running data in the past month is extracted and added to the training set, the learning rate is adjusted to 0.003, the batch size is adjusted to 32, and the model is retrained until the accuracy of the model prediction result is higher than 95%.

[0119] The application monitors three-phase electric energy parameters in the system operation process, constructs a bidirectional long and short term memory network Bi-LSTM model, uses actual operation parameters to construct a training set and a verification set to train the model, so that the model can predict the probability of fault position and fault type occurrence; when the probability exceeds the threshold value, an alarm is given, and at the same time, a suitable method is selected according to the fault cause to eliminate the fault, greatly improving the fault response speed and the stability of the photovoltaic grid-connected system; at the same time, the model has a self-optimization function and is optimized regularly to ensure the accuracy of the prediction result.

[0120] Embodiment 2

[0121] As shown in Figure 2 The embodiment discloses a photovoltaic system electric energy quality monitoring and fault diagnosis system based on deep learning, which comprises a three-phase electric parameter measurement module, a harmonic analysis module, a fault diagnosis module, an interactive display module and a storage module.

[0122] In the embodiment, the three-phase electric parameter measurement module preferably selects a ZDT-3561 flat plate type three-phase electric energy quality analyzer, which can collect data of components arranged in a photovoltaic field; the device supports voltage measurement range 0-1000V (accuracy ±0.1%), current measurement range 0-100A (accuracy ±0.2%), and sampling frequency 10kHz; the collected three-phase electric power parameters include voltage, current and power factor; the obtained data can be used to calculate active power, reactive power and apparent power, and to establish a time series data set.

[0123] The harmonic analysis module uses a windowed interpolation FFT algorithm to process the collected data to improve the harmonic detection accuracy.

[0124] The fault diagnosis module uses a trained deep learning model to predict the fault position and fault type of the photovoltaic component; at the same time, it has an automatic optimization function, and the latest operation data is collected to join the training set to optimize the model.

[0125] The interactive display module is used for manual input of control parameters and can display data in real time.

[0126] The storage module is used for storing operation data and establishing a log for subsequent retrieval of operation data.

[0127] Embodiment 3

[0128] Figure 3 The structure schematic diagram of the computer system of the electronic device suitable for realizing the embodiment of the application is shown.

[0129] It should be noted that the computer system of the electronic device in the embodiment is only an example, and should not bring any limitation to the function and use range of the embodiment of the application.

[0130] The computer system in this embodiment includes a central processing unit 401 which can perform various appropriate actions and processes in accordance with programs stored in a read only memory 402 or programs loaded from a storage section 408 into a random access memory 403, such as the deep learning-based photovoltaic system power quality monitoring and fault diagnosis method described in the above embodiments. Various programs and data required for system operation are also stored in the random access memory 403. The central processing unit 401, the read only memory 402, and the random access memory 403 are connected to each other through a bus 404. An input / output interface 405 is also connected to the bus 404.

[0131] The following components are connected to the input / output interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output interface 405 as necessary. A removable media 411 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is attached to the drive 410 as necessary, so that a computer program read therefrom is installed in the storage section 408 as necessary.

[0132] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program in accordance with embodiments of the present application. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program containing a computer program for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 409, and / or installed from the removable media 411. When the computer program is executed by the central processing unit 401, various functions defined in the system of the present application are performed.

[0133] It should be noted that the computer-readable medium in the embodiments of the present application can be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. The computer-readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read-Only Memory, CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus. In the present application, the computer-readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer-readable computer programs. Such a propagated data signal can take various forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium, which can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, device or apparatus. The computer programs contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, or the like, or any suitable combination of the above.

[0134] The flowcharts and block diagrams in the drawings illustrate the possible implementation architectures, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In the flowcharts or block diagrams, each block can represent a module, a program segment or a part of code containing one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders than that shown in the drawings. For example, two blocks that are shown in succession can actually be executed substantially in parallel, and they can also be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams or flowcharts, and the combination of blocks in the block diagrams or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0135] The units described in the embodiments of the present application can be implemented by software, or can be implemented by hardware, and the units described can also be arranged in a processor. In some cases, the names of the units do not constitute a limitation on the units themselves.

[0136] According to an aspect of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in the various optional implementation manners described above.

[0137] As another aspect, the present application also provides a computer readable medium, which can be included in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The computer readable medium carries one or more programs, which, when executed by the electronic device, enable the electronic device to implement the deep learning-based photovoltaic system power quality monitoring and fault diagnosis method described in the above embodiments.

[0138] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, such division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into several modules or units.

[0139] From the above description of the embodiments, those skilled in the art can easily understand that the example embodiments described herein can be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, U disk, mobile hard disk, etc.) or network, and includes several instructions to enable a computing device (which can be a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of the present application.

[0140] Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the features disclosed herein. It is intended that the application encompass any and all variations or modifications of the application which come within the scope of the overall disclosure and which fall within the generic principles or specific embodiments of the application. To the extent that there is a conflict or discrepancy between this detailed disclosure and the disclosure in the posted specification, including any documents incorporated by reference, the disclosed embodiments will control.

[0141] It is to be understood that the application is not limited to the precise construction here described and illustrated in the accompanying drawings, and that various modifications and changes can be effected therein by those skilled in the art without departing from the scope of the application. The scope of the application is limited only by the claims appended hereto.

Claims

1. A photovoltaic system power quality monitoring and fault diagnosis method based on deep learning, characterized in that: The specific steps include: S1: Power quality data acquisition and preprocessing: The three-phase power parameters of the photovoltaic system are collected in real time through the three-phase power parameter measurement module, and the collected three-phase power parameter data are preprocessed to construct a training data set; S2: Construction and training of deep learning models: Constructing a bidirectional long short-term memory (Bi-LSTM) model and training it using a preprocessed training dataset to predict PV module fault types and locations. S3: Real-time monitoring and fault diagnosis: The three-phase power parameters collected in real time are input into the trained Bi-LSTM model to output the fault type and fault probability distribution of the PV modules; S4: Model verification: Compare the model prediction results with the actual experimental results to determine whether the recognition accuracy meets the preset requirements; If it is lower than the requirement, the parameters are optimized and the model is iteratively updated until the accuracy of the model output structure meets the preset requirements.

2. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 1 is characterized in that: In step S1, a ZDT-3561 flat-panel three-phase power quality analyzer is used to collect three-phase power parameters of the photovoltaic system; the three-phase power parameters include voltage, current and power factor.

3. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 1 is characterized in that: The windowed interpolation FFT algorithm is used to perform harmonic analysis on the current and voltage signals, and the voltage and current total harmonic distortion values ​​are calculated, as well as a time series data set including the effective value of harmonic current, voltage distortion rate, power factor, and active power.

4. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 3 is characterized in that: In the bidirectional long short-term memory network Bi-LSTM model constructed in step S2, the time series dataset is used as the training dataset; the input layer receives the time series dataset, the hidden layer contains at least 3 layers of LSTM units, and the output layer is connected to a fully connected network; the model hyperparameters are optimized using the grid search method, with the learning rate set to 0.001-0.01, the batch size to 32-128, and the number of LSTM units to 64-256 to minimize the prediction error.

5. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 1 or 4, characterized in that: The training data set in step S2 includes a normal operating condition time series data set and a fault operating condition time series data set; the fault operating condition time series data set includes three fault types: PID attenuation fault, hot spot fault, and loose wiring fault.

6. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 5 is characterized in that: The PID attenuation fault is manifested as a drop in the open-circuit voltage of the PV module by ≥10%; Hot spot faults are manifested by abnormal increase in local current and temperature difference ≥15%; A loose connection fault is manifested as a sudden increase in the current harmonic distortion rate ≥5%.

7. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 1 is characterized in that: In step S3, an alarm threshold is pre-set. If the predicted fault probability exceeds the alarm threshold, an alarm signal is generated and the fault type and location are located.

8. The photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to claim 1, characterized in that: include: In step S4, the confusion matrix is ​​used to evaluate the fault classification accuracy, and the calculation formula is: Among them, TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

9. A computer-readable medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the photovoltaic system power quality monitoring and fault diagnosis method based on deep learning according to any one of claims 1 to 8 is implemented.

10. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the photovoltaic system power quality monitoring and fault diagnosis method based on deep learning as described in any one of claims 1 to 8.

Images