Photovoltaic power generation fault risk prediction method and system based on long short-term memory network

Abstract

The present invention provides a photovoltaic power generation fault risk prediction method and system based on a long short-term memory network, which relates to the field of data processing. The system comprises: a data acquisition module, which is used to collect environmental status information of photovoltaic panels, battery status information of photovoltaic panels, inverter status information of inverters, bracket status information of bracket structures and line status information of connecting lines; a power generation prediction module, which is used to generate photovoltaic power generation prediction power of photovoltaic power generation equipment based on environmental status information of photovoltaic panels through a long short-term memory network; and a risk prediction module, which is used to predict photovoltaic power generation fault risks based on photovoltaic power generation prediction power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of photovoltaic power generation equipment. The system has the advantage of realizing automatic and comprehensive prediction of photovoltaic power generation fault risks.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a photovoltaic power generation failure risk prediction method and system based on a long short-term memory network. Background Art

[0002] With the advancement of science and technology, people pay more and more attention to the development and use of renewable energy. Solar photovoltaic power generation has the characteristics of clean, zero emission and never-ending, and has become the first choice to replace traditional fossil energy. Photovoltaic power generation equipment is a device that converts solar energy into electrical energy using the photoelectric effect. It is usually composed of photovoltaic panels, inverters, support structures, connecting lines, etc. Fault diagnosis of photovoltaic power generation equipment is one of the key links in the maintenance and management of photovoltaic power generation equipment. With the widespread application and popularization of photovoltaic power generation technology, fault diagnosis has become a necessary means to ensure the efficient operation of photovoltaic systems and long-term stable power generation.

[0003] In the existing technology, the electricity generated by photovoltaic power generation equipment is generally stored in an electricity storage device. The power generation of the photovoltaic power generation equipment is also monitored by monitoring the changes in the stored electricity. However, when the stored electricity changes abnormally, the cause of the abnormality cannot be quickly monitored.

[0004] Therefore, it is necessary to provide a photovoltaic power generation failure risk prediction method and system based on long short-term memory network to achieve automatic and comprehensive prediction of photovoltaic power generation failure risks. Summary of the invention

[0005] The present invention provides a photovoltaic power generation fault risk prediction system based on a long short-term memory network, comprising: a data acquisition module, including a power generation acquisition unit, an environmental data acquisition unit, a battery data acquisition unit, an inverter data acquisition unit, a bracket data acquisition unit and a line data acquisition unit, wherein the power generation acquisition unit is used to acquire real-time photovoltaic power generation of a photovoltaic power generation device, the environmental data acquisition unit is used to acquire environmental status information of a photovoltaic battery panel, the battery data acquisition unit is used to acquire battery status information of a photovoltaic battery panel, the inverter data acquisition unit is used to acquire inverter status information of an inverter, the bracket data acquisition unit is used to acquire bracket status information of a bracket structure, and the line data acquisition unit is used to acquire line status information of a connecting line; a power generation prediction module, used to generate photovoltaic power generation prediction power of the photovoltaic power generation device based on the environmental status information of the photovoltaic battery panel acquired by the environmental data acquisition unit through a long short-term memory network; and a risk prediction module, used to predict photovoltaic power generation fault risks based on the photovoltaic power generation prediction power, real-time photovoltaic power generation power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation device.

[0006] Furthermore, the battery data acquisition unit includes a light sensing device, a temperature sensing device and an infrared image acquisition device, wherein the light sensing device includes a plurality of light sensors, and the plurality of light sensors are respectively arranged at a plurality of positions of the photovoltaic panel; the temperature sensing device includes a plurality of temperature sensors, and the plurality of temperature sensors are respectively arranged at a plurality of positions of the photovoltaic panel; the infrared image acquisition device includes a posture adjustment component and an infrared image acquisition component arranged on the posture adjustment component.

[0007] Furthermore, the battery data acquisition unit acquires battery status information of the photovoltaic panel, including: for each of the light sensors, based on the light intensity information of the panel acquired by the light sensor at multiple consecutive time points, calculating the shading risk value of the light sensor; based on the shading risk value of each of the light sensors, determining the target temperature sensor from the multiple temperature sensors; for each of the target temperature sensors, based on the temperature information of the panel acquired by the target temperature sensor at multiple consecutive time points, calculating the hot spot risk value of the target temperature sensor; based on the hot spot risk value of each of the target temperature sensors, determining the image acquisition posture; the infrared image acquisition device acquires infrared thermal imaging of the panel based on the image acquisition posture.

[0008] Furthermore, the bracket data acquisition unit includes a bracket sound acquisition device, a bracket vibration sensing device and a rotation angle acquisition device, wherein the rotation angle acquisition device is used to collect the angle of rotation of the photovoltaic panel driven by the bracket structure.

[0009] Further, the power generation prediction unit generates photovoltaic power generation prediction power based on the environmental status information of the photovoltaic panels collected by the environmental data collection unit through a long short-term memory network, including: generating multiple first photovoltaic power generation prediction powers based on the environmental status information of the photovoltaic panels collected by the environmental data collection unit through multiple power prediction models, wherein at least one of the power prediction models is constructed based on a long short-term memory network; calculating the difference discrete parameters of the multiple first photovoltaic power generation prediction powers; judging whether to perform prediction correction based on the difference discrete parameters; when it is determined to perform prediction correction, generating photovoltaic power generation prediction power based on the multiple first photovoltaic power generation prediction powers through the prediction correction model; when it is determined not to perform prediction correction, generating photovoltaic power generation prediction power based on the multiple first photovoltaic power generation prediction powers.

[0010] Furthermore, the risk prediction module performs photovoltaic power generation fault risk prediction based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment, including: calculating fault risk parameters based on the photovoltaic power generation predicted power and photovoltaic power generation real-time power of the photovoltaic power generation equipment; judging whether to perform photovoltaic power generation fault risk prediction according to the fault risk parameters; if it is determined to perform photovoltaic power generation fault risk prediction, calculating hot spot effect influence parameters based on the battery status information, and calculating sun following accuracy influence parameters based on the bracket status information and the battery status information; and calculating the photovoltaic power generation predicted power, hot spot effect influence parameters and the sun following accuracy influence parameters based on the photovoltaic power generation predicted power, hot spot effect influence parameters and the sun following accuracy influence parameters of the photovoltaic power generation equipment. influencing parameters, generating a second photovoltaic power generation predicted power of the photovoltaic power generation equipment; when the power difference between the second photovoltaic power generation predicted power of the photovoltaic power generation equipment and the photovoltaic power generation real-time power is greater than a preset power difference threshold, based on the inverter state information, an inverter fault risk prediction is performed, based on the line state information, a line fault risk prediction is performed, based on the hot spot effect influencing parameters, a hot spot fault prediction is performed based on the hot spot effect influencing parameters, and a solar following fault prediction is performed based on the solar following accuracy influencing parameters; when the power difference between the second photovoltaic power generation predicted power of the photovoltaic power generation equipment and the photovoltaic power generation real-time power is less than or equal to the preset power difference threshold, a hot spot fault prediction is performed based on the hot spot effect influencing parameters, and a solar following fault prediction is performed based on the solar following accuracy influencing parameters.

[0011] Furthermore, the risk prediction module calculates hot spot effect influencing parameters based on the battery status information, including: determining the hot spot risk area based on infrared thermal imaging of the battery panel through a clustering algorithm; for each hot spot risk area, determining the hot spot parameters of the hot spot risk area based on the infrared thermal imaging of the hot spot risk area; and calculating the hot spot effect influencing parameters based on the hot spot parameters of each hot spot risk area.

[0012] Furthermore, the risk prediction module calculates the parameters affecting the solar following accuracy based on the bracket status information and the battery status information, including: determining the parameters affecting the solar following accuracy based on the bracket status information, the ambient light intensity and the light intensity of the photovoltaic panel.

[0013] Furthermore, the risk prediction module performs inverter fault risk prediction based on the inverter status information, including: establishing an inverter fault risk prediction model; performing inverter fault risk prediction based on the inverter status information through the inverter fault risk prediction model; the risk prediction module performs line fault risk prediction based on the line status information, including: establishing a line fault risk prediction model; performing line fault risk prediction based on the line status information through the line fault risk prediction model.

[0014] The present invention provides a photovoltaic power generation fault risk prediction method based on a long short-term memory network, comprising: collecting real-time photovoltaic power generation of a photovoltaic power generation device; collecting environmental status information of a photovoltaic battery panel; collecting battery status information of the photovoltaic battery panel; collecting inverter status information of an inverter; collecting support status information of a support structure; collecting line status information of a connecting line; generating photovoltaic power generation prediction power of the photovoltaic power generation device based on the environmental status information of the photovoltaic battery panel collected by the environmental data collection unit through a long short-term memory network; and predicting photovoltaic power generation fault risks based on the photovoltaic power generation prediction power, real-time photovoltaic power generation, environmental status information, battery status information, inverter status information, support status information and line status information of the photovoltaic power generation device.

[0015] Compared with the prior art, the photovoltaic power generation fault risk prediction method and system based on long short-term memory network provided by the present invention has at least the following beneficial effects:

[0016] 1. By collecting and analyzing the status information of key components such as batteries, inverters, brackets and lines, potential faults can be discovered in time, early warnings can be given and maintenance measures can be taken to avoid greater impact on the system caused by faults. Comprehensive risk assessment based on multi-source data can more accurately judge the overall health of the photovoltaic power generation system, provide decision support for operation and maintenance personnel, and reduce system downtime and maintenance costs. The data acquisition module supports remote monitoring functions, and operation and maintenance personnel can understand the system operation status without having to visit the site, greatly improving operation and maintenance efficiency. Combined with data analysis technology, the system can automatically diagnose the type and cause of the fault, provide detailed maintenance guidance for operation and maintenance personnel, and shorten fault handling time.

[0017] 2. By analyzing the occlusion risk value of the light sensor, determine the target temperature sensor that may detect the temperature change caused by the hot spot effect. Further, analyze the hot spot risk value of the target temperature sensor to determine the location where the hot spot effect may occur and determine the image acquisition posture, so that the infrared image acquisition device can capture the complete infrared imaging of the area where the hot spot effect occurs on the photovoltaic panel, thereby improving the accuracy of subsequent photovoltaic power generation failure risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0019] Figure 1 It is a module schematic diagram of a photovoltaic power generation fault risk prediction system based on a long short-term memory network according to some embodiments of this specification;

[0020] Figure 2 is a schematic diagram of a process for collecting battery status information of a photovoltaic panel according to some embodiments of this specification;

[0021] Figure 3 It is a schematic diagram of a process for predicting photovoltaic power generation failure risks according to some embodiments of this specification;

[0022] Figure 4 It is a flow chart of a photovoltaic power generation fault risk prediction method based on a long short-term memory network according to some embodiments of this specification. DETAILED DESCRIPTION

[0023] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly introduces the drawings required for describing the embodiments.

[0024] Figure 1 is a module schematic diagram of a photovoltaic power generation fault risk prediction system based on a long short-term memory network according to some embodiments of this specification, such as Figure 1 As shown, the photovoltaic power generation fault risk prediction system based on the long short-term memory network can include a data acquisition module, a power generation prediction module and a risk prediction module.

[0025] The data acquisition module may include a power generation acquisition unit, an environmental data acquisition unit, a battery data acquisition unit, an inverter data acquisition unit, a bracket data acquisition unit and a line data acquisition unit.

[0026] Among them, the power generation collection unit is used to collect the real-time photovoltaic power generation power of the photovoltaic power generation equipment.

[0027] The environmental data acquisition unit is used to collect environmental status information of the photovoltaic panel (for example, environmental temperature, environmental humidity, environmental wind speed, environmental light intensity, etc.).

[0028] The battery data acquisition unit is used to collect battery status information of the photovoltaic panel.

[0029] In some embodiments, the battery data acquisition unit includes a light sensing device, a temperature sensing device and an infrared image acquisition device, wherein the light sensing device includes a plurality of light sensors, and the plurality of light sensors are respectively arranged at a plurality of positions of the photovoltaic panel; the temperature sensing device includes a plurality of temperature sensors, and the plurality of temperature sensors are respectively arranged at a plurality of positions of the photovoltaic panel; and the infrared image acquisition device includes a posture adjustment component and an infrared image acquisition component arranged on the posture adjustment component.

[0030] It is understandable that the light sensing device, the temperature sensing device and the infrared image acquisition device should be installed in a position that will not block the effective light receiving area of ​​the photovoltaic panel to ensure the normal power generation efficiency of the photovoltaic panel.

[0031] Figure 2 is a schematic diagram of a process for collecting battery status information of a photovoltaic panel according to some embodiments of this specification, such as Figure 2 As shown, the battery data acquisition unit collects battery status information of the photovoltaic panel, including:

[0032] For each light sensor, the shielding risk value of the light sensor is calculated based on the light intensity information of the solar panel collected by the light sensor at multiple consecutive time points;

[0033] determining a target temperature sensor from a plurality of temperature sensors based on the shading risk value of each light sensor;

[0034] For each target temperature sensor, based on the solar panel temperature information collected by the target temperature sensor at multiple consecutive time points, a hot spot risk value of the target temperature sensor is calculated;

[0035] Determine the image acquisition posture based on the hot spot risk value of each target temperature sensor;

[0036] The infrared image acquisition device acquires infrared thermal images of the solar panel based on the image acquisition posture.

[0037] Specifically, the occlusion risk value of the light sensor can be calculated based on the following formula:

[0038]

[0039] Among them, R (i,1) is the occlusion risk value of the i-th illumination sensor, I (i,t) is the light intensity sensed by the i-th light sensor at the t-th time point, I (E,t) is the ambient light intensity sensed at the t-th time point, T is the total number of sampled time points, ε1 is a preset parameter, and ε1>0, ε1 can be obtained based on experiments, and ε1 is used to characterize the standard deviation between the light intensity sensed by the unobstructed light sensor and the ambient light intensity, R(0,1) is the preset hot spot risk value.

[0040] A temperature sensor whose shading risk value is greater than a preset shading risk value threshold may be used as a target temperature sensor.

[0041] The hot spot risk value of the target temperature sensor can be calculated based on the following formula:

[0042]

[0043] T (P,t) =f(T (E,t) )

[0044] Among them, R (i,2) is the hot spot risk value of the i-th target temperature sensor, T (i,t) is the panel temperature sensed by the i-th target temperature sensor at the t-th time point, T (P,t) is the predicted temperature of the solar panel sensed at time point t, T (E,t) is the ambient temperature sensed at the t-th time point, f() is the temperature mapping function, which can be a nonlinear function, ε2 is a preset parameter, and ε2>0, ε2 can be obtained based on experiments, and ε2 is used to characterize the standard deviation between the temperature sensed by the temperature sensor without hot spot effect and the predicted solar panel temperature, R (0,2) is the preset hot spot risk value.

[0045] The target temperature sensor whose hot spot risk value is greater than the preset hot spot risk value threshold can be marked as a risk temperature sensor, and a marking sequence is generated according to the marking information of each temperature sensor, wherein the temperature sensor marked as a risk is represented by "1", and the temperature sensor not marked as a risk is represented by "0". As an example only, the marking sequence is (1, 1, 1, 0, 0, 0), which means that the first temperature sensor, the second temperature sensor, and the third temperature sensor are marked as risk temperature sensors. As another example, the marking sequence is (0, 1, 1, 1, 0, 0), which means that the second temperature sensor, the third temperature sensor, and the fourth sensor are marked as risk temperature sensors.

[0046] The data acquisition module can predetermine multiple groups of sample labeling sequences and the optimal image acquisition posture corresponding to each group of sample labeling sequences through experiments.

[0047] Calculate the sequence similarity between the current label sequence and the sample label sequence, take the sample label sequence with the largest sequence similarity and a sequence similarity greater than a preset sequence similarity threshold as the target sample label sequence, and obtain the optimal image acquisition posture corresponding to the target sample label sequence as the current image acquisition posture.

[0048] For example, sequence similarity can be calculated based on the following formula:

[0049]

[0050] Among them, S (C,j) is the sequence similarity between the current tag sequence and the jth sample tag sequence, σ1 is a preset parameter, and σ1>0, V (C,n) is the value of the nth element of the current tag sequence, V (j,n) The value of the nth element of the sequence marking the jth sample.

[0051] The inverter data acquisition unit is used to acquire inverter status information of the inverter, wherein the inverter status information may include status information of multiple components inside the inverter (eg, temperature, current, voltage, vibration, sound, etc.).

[0052] The support data acquisition unit is used to acquire support status information of the support structure.

[0053] In some embodiments, the bracket data collection unit includes a bracket sound collection device, a bracket vibration sensing device and a rotation angle collection device, wherein the rotation angle collection device is used to collect the angle of rotation of the photovoltaic panel driven by the bracket structure.

[0054] The line data acquisition unit is used to collect line status information of the connection line, such as line current, line voltage, line temperature, etc.

[0055] The power generation prediction module can be used to generate the photovoltaic power generation prediction power of the photovoltaic power generation equipment based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through the long short-term memory network.

[0056] In some embodiments, the power generation prediction unit generates photovoltaic power generation prediction power based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through a long short-term memory network, including:

[0057] Generate multiple first photovoltaic power generation prediction powers based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through multiple power prediction models, wherein at least one power prediction model is constructed based on a long short-term memory network;

[0058] Calculating a plurality of difference discrete parameters of the first photovoltaic power generation predicted power;

[0059] Based on the difference discrete parameters, determine whether to make a forecast correction;

[0060] When it is determined to perform prediction correction, a prediction correction model is used to generate a photovoltaic power generation prediction power based on a plurality of first photovoltaic power generation prediction powers, wherein the prediction correction model may be a convolutional neural network (CNN) model;

[0061] When it is determined that no prediction correction is to be performed, the photovoltaic power generation prediction power is generated based on the plurality of first photovoltaic power generation prediction powers, for example, the average of the plurality of first photovoltaic power generation prediction powers is calculated as the photovoltaic power generation prediction power.

[0062] Specifically, long short term memory (LSTM) is a type of recurrent neural network (RNN). In practical applications, it is found that RNN has problems such as gradient disappearance, gradient explosion, and poor ability to store long-distance dependent information, so LSTM was introduced. LSTM is similar to RNN in terms of main structure. Its main improvement is that three gate structures are added to the hidden layer h, namely forget gate, input gate, and output gate. At the same time, a hidden state (cell state) is added. Its core formula includes:

[0063] f=σ(W f ·[h t-1 ,x t ]+b f )

[0064] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0065]

[0066] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0067] h t =o t *tanh(C t )

[0068] Among them, σ represents the sigmoid function, W f is the weight matrix of the forget gate, W i is the weight matrix of the input gate, W Cis the candidate unit state weight matrix, b f is the bias term of the forget gate, b i is the bias term of the input gate, b C is the bias term of the candidate unit state, f t 、i t , o t Respectively represent the forget gate, input gate and output gate, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, b f 、b i 、b C and b o are bias vectors, is the candidate unit state, C t-1 is the cell state at the previous time step, C t and h t Represent the cell state and hidden state respectively.

[0069] The power prediction model can also be a Bi-LSTM (Bidirectional Long Short Term Memory) neural network structure model. The Bi-LSTM neural network structure model is divided into two independent LSTMs. The input sequence is input into the two LSTM neural networks in forward and reverse order for feature extraction. The word vector formed by splicing the two output vectors (i.e., the extracted feature vectors) is used as the final feature expression of the word. The design concept of the Bi-LSTM model is to make the feature data obtained at time t have information between the past and the future at the same time.

[0070] The power prediction model can also be a BP (Back Propagation) network model. The BP network can learn and store a large number of input-output mode mapping relationships without revealing the mathematical equations describing such mapping relationships in advance. Its learning rule is to use the steepest descent method to continuously adjust the network weights and thresholds through back propagation to minimize the sum of square errors of the network. Its main characteristics are: the signal is forward propagated, while the error is reverse propagated.

[0071] The difference discrete parameter can be calculated based on the following formula:

[0072]

[0073] Among them, δ1 is the difference discrete parameter, P n The photovoltaic power generation prediction power generated by the nth power prediction model, N is the total number of power prediction models.

[0074] The risk prediction module can be used to predict photovoltaic power generation failure risks based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment.

[0075] Figure 3 is a schematic diagram of a process for predicting photovoltaic power generation failure risks according to some embodiments of this specification, such as Figure 3 As shown, in some embodiments, the risk prediction module performs photovoltaic power generation failure risk prediction based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment, including:

[0076] Calculate the fault risk parameters based on the predicted photovoltaic power generation and real-time photovoltaic power generation of photovoltaic power generation equipment;

[0077] Determining whether to perform photovoltaic power generation fault risk prediction according to the fault risk parameter, for example, when the fault risk parameter is greater than a preset fault risk parameter threshold, determining to perform photovoltaic power generation fault risk prediction;

[0078] If it is determined to perform photovoltaic power generation failure risk prediction, the hot spot effect influencing parameters are calculated based on the battery status information, and the sun following accuracy influencing parameters are calculated based on the bracket status information and the battery status information;

[0079] Based on the photovoltaic power generation predicted power, hot spot effect influencing parameters and sun following accuracy influencing parameters of the photovoltaic power generation equipment, a second photovoltaic power generation predicted power of the photovoltaic power generation equipment is generated. Specifically, the second photovoltaic power generation predicted power of the photovoltaic power generation equipment can be generated based on the photovoltaic power generation predicted power, hot spot effect influencing parameters and sun following accuracy influencing parameters of the photovoltaic power generation equipment through a power correction model, wherein the power correction model can be a long-term and short-term neural network model;

[0080] When the power difference between the second photovoltaic power generation prediction power and the photovoltaic power generation real-time power of the photovoltaic power generation equipment is greater than the preset power difference threshold, the inverter fault risk is predicted based on the inverter state information, the line fault risk is predicted based on the line state information, the hot spot fault is predicted based on the hot spot effect influencing parameters, and the sun following fault is predicted based on the sun following accuracy influencing parameters;

[0081] When the power difference between the second photovoltaic power generation predicted power and the photovoltaic power generation real-time power of the photovoltaic power generation equipment is less than or equal to the preset power difference threshold, hot spot fault prediction is performed based on the hot spot effect influencing parameters, and sun following fault prediction is performed based on the sun following accuracy influencing parameters.

[0082] Specifically, the fault risk parameter can be calculated based on the following formula:

[0083]

[0084] Among them, R3 is the failure risk parameter, P (C,t) is the real-time power of photovoltaic power generation at time t, P (P,t) is the predicted power of photovoltaic power generation at time point t, ε3 is a preset parameter, and ε3>0, ε3 can be obtained based on experiments, and ε3 is used to characterize the standard deviation between the predicted power of photovoltaic power generation without faults and the real-time power of photovoltaic power generation, R (0,3) is the preset failure risk parameter.

[0085] In some embodiments, the risk prediction module calculates the hot spot effect impact parameter based on the battery status information, including:

[0086] Determine the hot spot risk area based on the infrared thermal imaging of the solar panels through clustering algorithms;

[0087] For each hot spot risk area, based on the infrared thermal imaging of the hot spot risk area, determine the hot spot parameters of the hot spot risk area, wherein the hot spot parameters may include the area of ​​the hot spot risk area, the temperature mean value of the hot spot risk area, and the temperature dispersion parameter;

[0088] Based on the hot spot parameters of each hot spot risk area, the hot spot effect impact parameters are calculated.

[0089] Specifically, the K-means clustering algorithm can be used to cluster multiple pixels based on the RGB value of each pixel in the infrared thermal imaging of the solar panel to determine the hot spot risk area.

[0090] The area of ​​the hot spot risk area can be determined by an area determination model based on infrared thermal imaging of the hot spot risk area, wherein the area determination model can be a convolutional neural network model.

[0091] The temperature discrete parameters can be calculated based on the following formula:

[0092]

[0093] Among them, δ2 is the temperature discrete parameter of the hot spot risk area, RGB m is the RGB value of the mth pixel of the infrared thermal imaging of the hot spot risk area, and M is the total number of pixels included in the infrared thermal imaging of the hot spot risk area.

[0094] The hot spot effect influencing parameters can be calculated based on the following formula:

[0095]

[0096] Among them, ε1 is the hot spot effect parameter, α 11 and α 12 are all preset weights, and α 11 >0,α 12 >0,α 11 +α 12 =1, S C is the area of ​​the hot spot risk area, S o is the preset area, δ0 is the preset temperature discrete parameter, RGB (C,average) is the average temperature of the hot spot risk area, RGB (0,average) It is the average temperature of the preset hot spot risk area.

[0097] In some embodiments, the risk prediction module calculates the sun following accuracy influencing parameters based on the bracket status information and the battery status information, including:

[0098] Based on the bracket status information, ambient light intensity and light intensity of the photovoltaic panels, the parameters affecting the sun following accuracy are determined.

[0099] Specifically, the accuracy prediction model can be used to determine the parameters affecting the solar following accuracy based on the bracket status information, ambient light intensity and light intensity of the photovoltaic panels at multiple consecutive time points. The accuracy prediction model can be such that the larger the solar following accuracy affecting parameter, the smaller the solar following accuracy. The accuracy prediction model can be a long-term and short-term neural network model.

[0100] In some embodiments, the risk prediction module performs inverter fault risk prediction based on the inverter state information, including:

[0101] Establishing an inverter fault risk prediction model, wherein the inverter fault risk prediction model may be a long-term and short-term neural network model;

[0102] The inverter fault risk prediction model is used to predict the inverter fault risk based on the inverter status information.

[0103] Specifically, the inverter status information of the inverter at multiple consecutive time points can be input into the inverter fault risk prediction model. The inverter fault risk prediction model outputs the possible inverter faults of the inverter (for example, loose terminal blocks, low insulation impedance, low bus voltage, leakage current fault, DC overvoltage protection, etc.) and the risk probability corresponding to the inverter fault. The greater the risk probability, the higher the possibility that the inverter will have the inverter fault.

[0104] In some embodiments, the risk prediction module performs line fault risk prediction based on the line status information, including:

[0105] Establishing a line fault risk prediction model, wherein the line fault risk prediction model may be a long-term and short-term neural network model;

[0106] The line fault risk prediction model is used to predict line fault risks based on line status information.

[0107] Specifically, the line status information of the line at multiple consecutive time points can be input into the line fault risk prediction model, and the line fault risk prediction model outputs the possible line faults (for example, string circuit breaking, line aging, line insulation layer failure, etc.) and the risk probability corresponding to the line fault risk.

[0108] Figure 4 is a flow chart of a photovoltaic power generation fault risk prediction method based on a long short-term memory network according to some embodiments of this specification, such as Figure 4 As shown, the photovoltaic power generation failure risk prediction method based on long short-term memory network can include the following steps.

[0109] Step 410, collecting real-time photovoltaic power generation power of the photovoltaic power generation equipment;

[0110] Step 420, collecting environmental status information of the photovoltaic panels;

[0111] Step 430, collecting battery status information of the photovoltaic panel;

[0112] Step 440, collecting inverter status information of the inverter;

[0113] Step 450, collecting support status information of the support structure;

[0114] Step 460, collecting line status information of the connection line;

[0115] Step 470, generating a predicted photovoltaic power generation power of the photovoltaic power generation equipment based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through a long short-term memory network;

[0116] Step 480 , predicting photovoltaic power generation failure risks based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment.

[0117] The photovoltaic power generation fault risk prediction method based on long short-term memory network can be executed by the photovoltaic power generation fault risk prediction system based on long short-term memory network. For more descriptions of the photovoltaic power generation fault risk prediction method based on long short-term memory network, please refer to the relevant description of the photovoltaic power generation fault risk prediction system based on long short-term memory network, which will not be repeated here.

[0118] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A photovoltaic power generation fault risk prediction system based on long short-term memory network, applied to photovoltaic power generation equipment, characterized in that: include: The data acquisition module includes a power generation acquisition unit, an environmental data acquisition unit, a battery data acquisition unit, an inverter data acquisition unit, a bracket data acquisition unit and a line data acquisition unit, wherein the power generation acquisition unit is used to acquire the real-time power of photovoltaic power generation of the photovoltaic power generation equipment, the environmental data acquisition unit is used to acquire the environmental status information of the photovoltaic battery panel, the battery data acquisition unit is used to acquire the battery status information of the photovoltaic battery panel, the inverter data acquisition unit is used to acquire the inverter status information of the inverter, the bracket data acquisition unit is used to acquire the bracket status information of the bracket structure, and the line data acquisition unit is used to acquire the line status information of the connecting line; A power generation prediction module, used to generate a predicted photovoltaic power generation power of the photovoltaic power generation equipment based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through a long short-term memory network; The risk prediction module is used to predict the risk of photovoltaic power generation failure based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment, specifically including: Calculating a fault risk parameter based on the photovoltaic power generation prediction power and the photovoltaic power generation real-time power of the photovoltaic power generation equipment; Determining whether to perform photovoltaic power generation failure risk prediction according to the failure risk parameters; If it is determined to perform photovoltaic power generation failure risk prediction, then based on the battery status information, the hot spot effect influencing parameter is calculated, and based on the bracket status information and the battery status information, the sun following accuracy influencing parameter is calculated; Based on the photovoltaic power generation predicted power of the photovoltaic power generation equipment, the hot spot effect influencing parameter and the sun following accuracy influencing parameter, generating a second photovoltaic power generation predicted power of the photovoltaic power generation equipment; When the power difference between the second photovoltaic power generation predicted power of the photovoltaic power generation equipment and the photovoltaic power generation real-time power is greater than a preset power difference threshold, an inverter fault risk prediction is performed based on the inverter state information, a line fault risk prediction is performed based on the line state information, a hot spot fault prediction is performed based on the hot spot effect influencing parameter, and a sun following fault prediction is performed based on the sun following accuracy influencing parameter; When the power difference between the second photovoltaic power generation predicted power and the photovoltaic power generation real-time power of the photovoltaic power generation equipment is less than or equal to a preset power difference threshold, a hot spot fault prediction is performed based on the hot spot effect influencing parameters, and a sun following fault prediction is performed based on the sun following accuracy influencing parameters.

2. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to claim 1 is characterized in that: The battery data acquisition unit includes a light sensing device, a temperature sensing device and an infrared image acquisition device, wherein the light sensing device includes a plurality of light sensors, and the plurality of light sensors are respectively arranged at a plurality of positions of the photovoltaic panel; the temperature sensing device includes a plurality of temperature sensors, and the plurality of temperature sensors are respectively arranged at a plurality of positions of the photovoltaic panel; the infrared image acquisition device includes a posture adjustment component and an infrared image acquisition component arranged on the posture adjustment component.

3. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to claim 2 is characterized in that: The battery data acquisition unit acquires battery status information of the photovoltaic panel, including: For each of the light sensors, based on the light intensity information of the solar panel collected by the light sensor at multiple consecutive time points, calculating the shielding risk value of the light sensor; determining a target temperature sensor from the plurality of temperature sensors based on the shielding risk value of each of the light sensors; For each of the target temperature sensors, based on the solar panel temperature information collected by the target temperature sensor at multiple consecutive time points, calculating the hot spot risk value of the target temperature sensor; Determining an image acquisition posture based on a hot spot risk value of each of the target temperature sensors; The infrared image acquisition device acquires infrared thermal images of the solar panel based on the image acquisition posture.

4. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to any one of claims 1 to 3, characterized in that: The bracket data acquisition unit includes a bracket sound acquisition device, a bracket vibration sensing device and a rotation angle acquisition device, wherein the rotation angle acquisition device is used to acquire the angle of rotation of the photovoltaic panel driven by the bracket structure.

5. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to any one of claims 1 to 3, characterized in that: The power generation prediction module generates photovoltaic power generation prediction power based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through a long short-term memory network, including: Generate multiple first photovoltaic power generation prediction powers based on the environmental state information of the photovoltaic panels collected by the environmental data collection unit through multiple power prediction models, wherein at least one of the power prediction models is constructed based on a long short-term memory network; Calculating the difference discrete parameters of the plurality of first photovoltaic power generation predicted powers; Based on the difference discrete parameter, determining whether to perform prediction correction; When it is determined that the prediction correction is to be performed, generating the photovoltaic power generation prediction power based on the plurality of first photovoltaic power generation prediction powers by using the prediction correction model; When it is determined that the prediction correction is not to be performed, the photovoltaic power generation prediction power is generated based on the plurality of first photovoltaic power generation prediction powers.

6. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to claim 1 is characterized in that: The risk prediction module calculates hot spot effect impact parameters based on the battery status information, including: Determine the hot spot risk area based on the infrared thermal imaging of the solar panels through clustering algorithms; For each hot spot risk area, determining a hot spot parameter of the hot spot risk area based on infrared thermal imaging of the hot spot risk area; The hot spot effect influencing parameter is calculated based on the hot spot parameters of each of the hot spot risk areas.

7. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to claim 1 is characterized in that: The risk prediction module calculates the sun following accuracy influencing parameters based on the bracket status information and the battery status information, including: The sun following accuracy influencing parameter is determined based on the bracket status information, the ambient light intensity and the light intensity of the photovoltaic panel.

8. The photovoltaic power generation fault risk prediction system based on long short-term memory network according to claim 7 is characterized in that: The risk prediction module performs inverter fault risk prediction based on the inverter state information, including: Establish an inverter failure risk prediction model; Performing inverter fault risk prediction based on the inverter state information by using the inverter fault risk prediction model; The risk prediction module performs line fault risk prediction based on the line status information, including: Establish a line fault risk prediction model; The line fault risk prediction model is used to predict line fault risks based on the line status information.

9. A photovoltaic power generation fault risk prediction method based on long short-term memory network, characterized in that: The photovoltaic power generation fault risk prediction system based on the long short-term memory network applied to any one of claims 1 to 8 comprises: Collect the real-time photovoltaic power generation power of photovoltaic power generation equipment; Collect the environmental status information of the photovoltaic panels; Collect battery status information of photovoltaic panels; Collect inverter status information of the inverter; Collecting support status information of the support structure; Collecting line status information of the connection line; Generate the predicted photovoltaic power generation power of the photovoltaic power generation equipment based on the environmental status information of the photovoltaic panels through a long short-term memory network; Photovoltaic power generation failure risk prediction is performed based on the photovoltaic power generation predicted power, photovoltaic power generation real-time power, environmental status information, battery status information, inverter status information, bracket status information and line status information of the photovoltaic power generation equipment.

Images