Photovoltaic panel dust deposition detection and pollutant identification method based on unmanned aerial vehicle-mounted hyperspectrum

Abstract

The invention relates to an identification method, in particular to a photovoltaic panel dust deposition detection and pollutant identification method based on an unmanned aerial vehicle-mounted hyperspectrum. The method comprises the following steps: collecting spectral data; preprocessing the spectral data; extracting effective spectral features; change detection; judging whether pollutants exist or not; carrying out pollutant comparison; and evaluating ash deposition and giving an alarm. According to the invention, through combination of hyperspectral imaging and spectral variation analysis, automatic and non-contact monitoring of the accumulated dust and pollutants on the photovoltaic panel is realized, and tiny spectral change on the surface of the photovoltaic panel can be captured in real time and with high precision, so that the degree and the type of the accumulated dust and pollutants are accurately detected, the detection precision and the real-time performance are improved, and the detection efficiency is improved. And the maintenance cost is effectively reduced, unnecessary cleaning operation is avoided, and maintenance suggestions can be provided in time according to the detection result, so that the energy conversion efficiency of the photovoltaic panel is further improved.

Description

Technical Field

[0001] The present invention relates to an identification method, and in particular to a photovoltaic panel dust accumulation detection and pollutant identification method based on an unmanned aerial vehicle-borne hyperspectral image. Background Art

[0002] As the global demand for renewable energy continues to grow, photovoltaic power generation has been widely used as a clean and sustainable form of energy. However, photovoltaic modules are easily affected by environmental and meteorological factors during long-term operation, leading to problems such as dirt accumulation and panel damage, which in turn affects the power generation efficiency and safety of photovoltaic power stations.

[0003] Currently, the inspection of photovoltaic modules mainly relies on manual inspection, which is not only time-consuming and labor-intensive, but also has a long inspection cycle for the entire station. Hidden dangers cannot be eliminated in time, resulting in a continuous decline in power generation and economic losses. Summary of the invention

[0004] The purpose of the present invention is to solve the technical problems that the existing manual inspection method of photovoltaic components is not only time-consuming and labor-intensive, but also has a long inspection cycle for the entire station, and hidden dangers cannot be eliminated in time, resulting in a continuous decline in power generation and bringing economic losses. A method for detecting dust accumulation and identifying pollutants on photovoltaic panels based on drone-mounted hyperspectral images is provided.

[0005] The present invention uses a fully autonomous drone system, combined with the collaborative work of a drone farm (tower or ground-based), which not only provides charging, maintenance and storage functions for drones, but also performs task planning and scheduling, so that drones can autonomously perform inspection tasks without human intervention. Through drones equipped with hyperspectral imaging sensors, the system can perform non-contact detection of photovoltaic panels in real time and efficiently, and detect dirt and damage in a timely manner, thereby optimizing maintenance plans, reducing labor costs, and ensuring efficient operation and power generation safety of photovoltaic power stations.

[0006] To achieve the above purpose, the technical solution adopted by the present invention is:

[0007] A method for detecting dust accumulation and identifying pollutants on photovoltaic panels based on an unmanned aerial vehicle (UAV) hyperspectral image is provided, which is special in that it includes the following steps:

[0008] Step 1: Obtain spectral data of the photovoltaic panel to be tested through the hyperspectral imaging sensor carried by the drone;

[0009] Step 2: preprocessing the collected spectral data;

[0010] Step 3: Perform principal component analysis on the preprocessed spectral data to reduce the dimension and extract effective spectral features;

[0011] Step 4: By using the pre-built time series-based spectral change detection model, the extracted effective spectral features are compared with the historical reference features to identify the spectral changes caused by dust accumulation and pollutants, and determine whether there are pollutants; if so, the extracted effective spectral features are compared with the reference spectrum in the existing spectral library to identify the type of pollutants, and then execute step 5; if not, directly execute step 5;

[0012] Step 5: Input the extracted effective spectral features into a support vector machine, use the support vector machine to classify and quantify the dust accumulation and pollutant levels, and obtain the test results;

[0013] Step 6: Based on the detection results, evaluate the impact of dust accumulation and pollutants on the photovoltaic panel photoelectric conversion efficiency, generate dust accumulation and pollutant distribution maps and level reports, and issue a maintenance alarm when the impact of dust accumulation and / or pollutants exceeds a preset threshold.

[0014] Furthermore, in step 4, the spectrum change detection expression of the spectrum change detection model is:

[0015]

[0016] Where: d is the Euclidean distance corresponding to the extracted effective spectral features; x k is a historical reference feature; k is the effective spectral feature extracted; m is a constant;

[0017] By judging whether d exceeds a preset threshold δ, it is determined whether there is a pollutant.

[0018] Furthermore, in step 4, the extracted effective spectral features are compared with the reference spectra in the existing spectral library, and the spectral similarity is calculated by the following spectral angle matching formula:

[0019]

[0020] Where: x is the extracted effective spectral feature; y is the reference spectrum in the spectral library; θ is the spectral angle between the spectral feature and the reference spectrum;

[0021] The spectral angle θ between the spectral feature and the reference spectrum is compared with the set spectral angle size and pollutant type comparison table to identify the type of pollutant.

[0022] Furthermore, in step 5, the classification model of the support vector machine is:

[0023] f(x)=sign(w T x+b)

[0024] Where: f(x) is the classification model, w is the weight vector, and b is the bias term.

[0025] Furthermore, in step 6, the photovoltaic panel photoelectric conversion efficiency in the current state is determined based on the detection results, and then the change in the photovoltaic panel photoelectric conversion efficiency is calculated using the following formula to evaluate the impact of dust accumulation and pollutants:

[0026]

[0027] Where: η is the percentage of photovoltaic panel photoelectric conversion efficiency loss caused by dust accumulation and pollutants, P actual is the photovoltaic panel photoelectric conversion efficiency in a clean state, P ideal is the photovoltaic panel photoelectric conversion efficiency in the current state.

[0028] Furthermore, step 2 is specifically as follows:

[0029] The collected spectral data are subjected to radiation correction and normalization to achieve preprocessing.

[0030] Furthermore, in step 3, the following formula is used to perform principal component analysis on the preprocessed spectral data to reduce the dimension and extract effective spectral features:

[0031] X reduced =X norm W T

[0032] CW=WΛ

[0033]

[0034] Where: C is the covariance matrix; n is the number of samples; Λ is the diagonal matrix; W is the eigenvector matrix of principal component analysis; X norm is the main component spectral feature after projection; X reduced is the effective spectral feature extracted.

[0035] The beneficial effects of the present invention are:

[0036] The present invention realizes the automatic and non-contact monitoring of dust accumulation and pollutants on photovoltaic panels by combining hyperspectral imaging with spectral variation analysis. It can capture tiny spectral changes on the surface of photovoltaic panels in real time and with high precision, so as to accurately detect the degree and type of dust accumulation and pollutants. Compared with traditional manual inspection methods, the present invention not only improves the accuracy and real-time performance of detection, but also effectively reduces maintenance costs, avoids unnecessary cleaning operations, and can provide maintenance suggestions in a timely manner based on the detection results, thereby further improving the energy conversion efficiency of photovoltaic panels. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1It is a schematic diagram of the principle of the photovoltaic panel dust detection and pollutant identification method based on the unmanned aerial vehicle-mounted hyperspectral spectrum of the present invention;

[0038] Figure 2 It is a flow chart of the photovoltaic panel dust detection and pollutant identification method based on the unmanned aerial vehicle-borne hyperspectral method of the present invention. DETAILED DESCRIPTION

[0039] In order to make the purpose, advantages and features of the present invention clearer, the following is a further detailed description of a photovoltaic panel dust detection and pollutant identification method based on drone-mounted hyperspectral imaging proposed by the present invention in combination with the accompanying drawings and specific embodiments. The advantages and features of the present invention will be clearer according to the following specific implementation methods.

[0040] The photovoltaic panel dust accumulation detection and pollutant identification method based on unmanned aerial vehicle hyperspectral imaging provided by the present invention is as follows: Figure 2 As shown, the specific implementation includes the following steps:

[0041] 1. See Figure 1 , the spectral data of photovoltaic panels is obtained through the hyperspectral imaging sensor carried by the drone, capturing its reflection characteristics in the visible light to near-infrared band. Hyperspectral data can be represented as a three-dimensional matrix:

[0042] H(i,j,λ)

[0043] Where H(i,j,λ) represents the spectral reflectance at wavelength λ at pixel position (i,j), i and j are the spatial dimensions of the image, and λ is the number of spectral bands.

[0044] 2. Perform radiation correction and normalization on the collected spectral data to eliminate the influence of ambient light changes on spectral characteristics. The processed spectral data is expressed as:

[0045]

[0046] Among them, R norm (λ) represents the normalized spectral data, R(λ) represents the original spectral data, and R dark (λ) is the dark current response, R ref (λ) is the reference spectrum data of the white plate.

[0047] 3. Perform principal component analysis (PCA) on the original hyperspectral data to reduce the dimension and extract effective features. The mathematical representation of PCA is as follows:

[0048] X reduced =X norm W T

[0049] Among them, X reducedis the effective spectral feature extracted, W is the eigenvector matrix of PCA, X norm is the main component spectral feature after projection.

[0050] The PCA calculation process includes:

[0051] Compute the covariance matrix:

[0052]

[0053] Where C is the covariance matrix and n is the number of samples.

[0054] Solve for eigenvalues ​​and eigenvectors:

[0055] CW=WΛ

[0056] Among them, Λ is a diagonal matrix containing the eigenvalues ​​of the covariance matrix, and W is the corresponding eigenvector matrix.

[0057] PCA can be used to remove redundant information and retain only the most valuable spectral features for pollutant identification.

[0058] 4. Through the pre-built time series-based spectral change detection model, the current spectral features are compared with the historical reference features to identify spectral changes caused by dust accumulation and pollutants. The mathematical expression of change detection is:

[0059]

[0060] Among them, d is the Euclidean distance (norm) corresponding to the extracted effective spectral features, x k is the historical reference feature, y k is the effective spectral feature extracted, and m is a constant.

[0061] When d exceeds the preset threshold δ, it is judged that there is a pollutant; otherwise, it is judged that there is no pollutant.

[0062] If there are pollutants, the extracted effective spectral features are compared with the reference spectra in the existing spectral library, and the type of pollutants is determined by calculating the spectral similarity through spectral angle matching (SAM). The mathematical expression of spectral angle matching is:

[0063]

[0064] Among them, x is the spectral feature currently collected, y is the reference spectrum in the spectral library, and θ is the spectral angle between the two.

[0065] The spectral angle θ between the spectral feature and the reference spectrum is compared with the set spectral angle size and pollutant type comparison table to identify the type of pollutant.

[0066] 5. Input the extracted effective spectral features into the support vector machine (SVM), and use the support vector machine (SVM) to classify and quantify the degree of dust accumulation and pollutants to obtain the detection results.

[0067] The SVM classification model can be expressed as:

[0068] f(x)=sign(w T x+b)

[0069] Among them, f(x) is the classification model, w is the weight vector, and b is the bias term.

[0070] 6. The impact of dust accumulation and pollutants can be assessed by calculating the change in photoelectric conversion efficiency, defined as follows:

[0071]

[0072] Where η represents the efficiency loss percentage due to dust accumulation, P actual is the photoelectric conversion efficiency in the clean state, P ideal The photoelectric conversion efficiency of the photovoltaic panel in the current state is determined based on the detection results.

[0073] Finally, dust and contaminant distribution maps and level reports are generated, and maintenance alerts are issued when the impact of dust and / or contaminants exceeds preset thresholds.

Claims

1. A method for detecting dust accumulation and identifying pollutants on photovoltaic panels based on unmanned aerial vehicle hyperspectral imaging, characterized in that: The following steps are involved: Step 1: Obtain spectral data of the photovoltaic panel to be tested through the hyperspectral imaging sensor carried by the drone; Step 2: preprocessing the collected spectral data; Step 3: Perform principal component analysis on the preprocessed spectral data to reduce the dimension and extract effective spectral features; Step 4: By using the pre-built time series-based spectral change detection model, the extracted effective spectral features are compared with the historical reference features to identify the spectral changes caused by dust accumulation and pollutants, and determine whether there are pollutants; if so, the extracted effective spectral features are compared with the reference spectrum in the existing spectral library to identify the type of pollutants, and then execute step 5; if not, directly execute step 5; Step 5: Input the extracted effective spectral features into a support vector machine, use the support vector machine to classify and quantify the dust accumulation and pollutant levels, and obtain the test results; Step 6: Based on the detection results, evaluate the impact of dust accumulation and pollutants on the photovoltaic panel photoelectric conversion efficiency, generate dust accumulation and pollutant distribution maps and level reports, and issue a maintenance alarm when the impact of dust accumulation and / or pollutants exceeds a preset threshold.

2. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle hyperspectral according to claim 1 is characterized in that ； In step 4, the spectrum change detection expression of the spectrum change detection model is: Where: d is the Euclidean distance corresponding to the extracted effective spectral features; x k is a historical reference feature; k is the effective spectral feature extracted; m is a constant; By judging whether d exceeds a preset threshold δ, it is determined whether there is a pollutant.

3. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle hyperspectral according to claim 2 is characterized by: In step 4, the extracted effective spectral features are compared with the reference spectra in the existing spectral library, and the spectral similarity is calculated using the following spectral angle matching formula: Where: x is the extracted effective spectral feature; y is the reference spectrum in the spectral library; θ is the spectral angle between the spectral feature and the reference spectrum; The spectral angle θ between the spectral feature and the reference spectrum is compared with the set spectral angle size and pollutant type comparison table to identify the type of pollutant.

4. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle-mounted hyperspectral according to any one of claims 1 to 3 is characterized in that: In step 5, the classification model of the support vector machine is: f(x)=sign(w T x+b) Where: f(x) is the classification model, w is the weight vector, and b is the bias term.

5. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle hyperspectral according to claim 4 is characterized by: In step 6, the photovoltaic panel photoelectric conversion efficiency in the current state is determined based on the test results, and then the change in the photovoltaic panel photoelectric conversion efficiency is calculated using the following formula to evaluate the impact of dust accumulation and pollutants: Where: η is the percentage of photovoltaic panel photoelectric conversion efficiency loss caused by dust accumulation and pollutants, P actual is the photovoltaic panel photoelectric conversion efficiency in a clean state, P ideal is the photovoltaic panel photoelectric conversion efficiency in the current state.

6. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle hyperspectral according to claim 1 is characterized in that: Step 2 is as follows: The collected spectral data are subjected to radiation correction and normalization to achieve preprocessing.

7. The photovoltaic panel dust detection and pollutant identification method based on unmanned aerial vehicle hyperspectral according to claim 6 is characterized by: In step 3, the following formula is used to perform principal component analysis on the preprocessed spectral data to reduce the dimension and extract effective spectral features: X reduced =X norm W T CW=WΛ Where: C is the covariance matrix; n is the number of samples; Λ is the diagonal matrix; W is the eigenvector matrix of principal component analysis; X norm is the main component spectral feature after projection; X reduced is the effective spectral feature extracted.

Images