A photovoltaic device control method and system

Abstract

This invention discloses a photovoltaic (PV) equipment control method and system, relating to the field of PV power control technology. The PV equipment control method acquires the operating sequence data of the PV inverter and the thermal imaging time-series data of the PV surface of the target PV equipment; it performs correlation and integration processing on the operating sequence data of the PV inverter of the target PV equipment to extract the inverter power quality assessment value; based on a pre-trained thermal imaging power quality mapping model, it performs power mapping processing on the PV surface thermal imaging time-series data to obtain the PV power quality mapping assessment value of the target PV equipment; and combines this with the inverter power quality assessment value for comprehensive analysis to obtain the grid-connected power quality assessment value of the target PV equipment. This invention performs power quality control processing on the target PV equipment through the grid-connected power quality assessment value, thereby achieving collaborative perception of PV power quality and ensuring that the grid-connected power quality assessment value can fully reflect the power quality status of the entire link, thus guaranteeing the stable adaptation of the PV equipment to the power grid.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic power control technology, specifically to a photovoltaic equipment control method and system. Background Technology

[0002] With the rapid expansion of photovoltaic (PV) installations, the power quality of PV equipment connected to the grid has become one of the core factors affecting the safety and stability of the power grid and the operating efficiency of PV systems. As the source of power generation, PV modules are prone to thermal anomalies during operation, which are quite harmful. When a module experiences a thermal anomaly, it can not only cause a sharp drop in local power generation efficiency and a shortened module lifespan, but also trigger a series of power quality problems: PV cells in hot spots will transform from power generation units to energy consumption units, causing drastic fluctuations in module output power, which in turn can lead to unstable fluctuations in grid current, increase grid line losses, and even cause relay protection devices to malfunction. However, most current PV management solutions do not pay enough attention to module thermal anomalies, making it difficult to accurately predict the power quality problems caused by thermal anomalies.

[0003] Furthermore, as the core conversion hub between photovoltaic modules and the power grid, the inverter determines the final quality level of grid-connected power. Electrical quality disturbances caused by abnormal thermal states of the modules will directly affect the input side of the inverter, increasing the difficulty of inverter regulation. If the inverter fails to detect abnormal thermal states of the modules in time and adjust the regulation strategy accordingly, it is very easy to amplify electrical quality problems and form a vicious cycle.

[0004] The limitations of existing technologies include at least the following problems: Existing technologies struggle to achieve coordinated perception of photovoltaic power quality, leading to delayed responses in complex operating conditions, which in turn affects grid connection safety and power generation efficiency. Specifically, existing methods struggle to detect power quality degradation trends caused by hot spots and uneven aging of modules. For example, when hidden hot spots appear in localized areas of photovoltaic panels, existing methods struggle to identify the harmonic distortion risks they cause in the early stages, which can easily lead to power quality exceeding standards at the grid connection point or even protection tripping. Furthermore, existing technologies assess inverter operating status and module temperature status separately, lacking a comprehensive analysis of electrothermal coupling effects. When environmental irradiance fluctuates drastically, it is difficult to coordinate and optimize inverter modulation, resulting in additional efficiency losses and impacting grid-connected power quality. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a photovoltaic equipment control method and system, which solves the problem that existing technologies are unable to coordinately sense photovoltaic power quality and that the disconnect between the status of components and inverters affects the grid-connected power quality.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a photovoltaic equipment control method, comprising the following steps: acquiring photovoltaic inverter operation sequence data and photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment; performing correlation and integration processing on the photovoltaic inverter operation sequence data of the target photovoltaic equipment to extract the inverter power quality evaluation value of the target photovoltaic equipment; performing power mapping processing on the photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment based on a pre-trained thermal imaging power quality mapping model to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment, and performing comprehensive analysis in conjunction with the inverter power quality evaluation value to obtain the grid-connected power quality evaluation value of the target photovoltaic equipment; and performing power quality control processing on the target photovoltaic equipment based on the grid-connected power quality evaluation value.

[0007] Furthermore, the photovoltaic inverter runtime timing data includes power distribution value, inverter modulation ratio, voltage deviation value, phase deviation value, IGBT junction temperature deviation value, and inverter efficiency value at each time point. The specific steps for extracting the inverter power quality evaluation value of the target photovoltaic equipment are as follows: read the photovoltaic inverter runtime timing data of the target photovoltaic equipment and perform preprocessing; classify the preprocessed photovoltaic inverter runtime timing data of the target photovoltaic equipment to obtain the inverter control timing subset and inverter state timing subset of the target photovoltaic equipment; perform correlation and integration processing on the inverter control timing subset and inverter state timing subset of the target photovoltaic equipment respectively to obtain the adjustment adaptation evaluation value and operating condition steady-state evaluation value of the target photovoltaic equipment at each time point; perform collaborative evaluation processing based on the adjustment adaptation evaluation value and operating condition steady-state evaluation value of the target photovoltaic equipment at each time point to extract the inverter power quality evaluation value of the target photovoltaic equipment.

[0008] Furthermore, the specific steps to obtain the adjustment and adaptation evaluation value of the target photovoltaic equipment at each time point are as follows: Based on the inverter control time series subset of the target photovoltaic equipment, construct the control covariance matrix of the target photovoltaic equipment; based on the control covariance matrix of the target photovoltaic equipment, analyze the control coefficient set of the target photovoltaic equipment; based on the control coefficient set and the inverter control time series subset of the target photovoltaic equipment, extract the adjustment and adaptation evaluation value of the target photovoltaic equipment at each time point.

[0009] Furthermore, the specific steps of the collaborative evaluation process are as follows: Sample entropy analysis is performed on the adjustment and adaptation evaluation value and the steady-state evaluation value of the target photovoltaic equipment at each time point to obtain the inverter coefficient set of the target photovoltaic equipment; based on the inverter coefficient set of the target photovoltaic equipment and the adjustment and adaptation evaluation value and the steady-state evaluation value of each time point, the data are input into the preset inverter collaborative evaluation model to extract the corresponding power quality evaluation value at each time point; a moving average processing is performed on the power quality evaluation value of the target photovoltaic equipment at each time point to obtain the inverter power quality evaluation value of the target photovoltaic equipment.

[0010] Furthermore, the photovoltaic surface thermal imaging time series data includes several frames of photovoltaic surface thermal imaging data. Specifically, the photovoltaic surface thermal imaging data consists of the temperature value and corresponding two-dimensional coordinates of each pixel in the thermal imaging. The thermal image electro-mass mapping model includes an input layer, an electro-mass mapping layer, an LSTM layer, and an output layer.

[0011] Further, the specific steps to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic device are as follows: input the photovoltaic surface thermal imaging time series data of the target photovoltaic device into the pre-trained thermal imaging power quality mapping model, extract the power quality mapping dataset of the target photovoltaic device, including hot spot harmonic tendency evaluation value, distortion correlation evaluation value, and voltage flicker evaluation value; based on the power quality mapping dataset of the target photovoltaic device, extract the photovoltaic power quality mapping evaluation value of the target photovoltaic device.

[0012] Further, the specific steps for extracting the electrical quality mapping dataset of the target photovoltaic device are as follows: In the input layer, each frame of photovoltaic surface thermal imaging data of the target photovoltaic device is received and data cleaning is performed; in the electrical quality mapping layer, thermoelectric correlation feature processing is performed on each frame of photovoltaic surface thermal imaging data of the target photovoltaic device after data cleaning, and the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is extracted; in the LSTM layer, temporal correlation processing is performed on the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device to obtain the temporal fused electrical quality feature vector of the target photovoltaic device; in the output layer, based on the temporal fused electrical quality feature vector of the target photovoltaic device, the electrical quality mapping dataset of the target photovoltaic device is output.

[0013] Further, the specific steps for extracting the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device are as follows: perform thermal state feature extraction processing on each frame of photovoltaic surface thermal imaging data of the target photovoltaic device to obtain the thermal state sensing feature vector of the corresponding frame of photovoltaic surface thermal imaging; and perform mapping processing on the thermal state sensing feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device to obtain the electrical quality mapping feature vector of the corresponding frame of photovoltaic surface thermal imaging.

[0014] Further, the specific steps to obtain the grid-connected power quality assessment value of the target photovoltaic equipment are as follows: read the inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment and perform standardization processing; input the standardized inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment into the preset comprehensive power quality assessment model, and extract the grid-connected power quality assessment value of the target photovoltaic equipment.

[0015] A photovoltaic (PV) equipment control system includes: a data acquisition module for acquiring the operating sequence data of the PV inverter and the time-series data of PV surface thermal imaging of the target PV equipment; an inverter power quality analysis module for performing correlation and integration processing on the operating sequence data of the PV inverter of the target PV equipment to extract the inverter power quality evaluation value of the target PV equipment; a thermal imaging power quality mapping integration module for performing power mapping processing on the PV surface thermal imaging time-series data of the target PV equipment based on a pre-trained thermal imaging power quality mapping model to obtain the PV power quality mapping evaluation value of the target PV equipment, and performing comprehensive analysis in combination with the inverter power quality evaluation value to obtain the grid-connected power quality evaluation value of the target PV equipment; and a power quality control module for performing power quality control processing on the target PV equipment based on the grid-connected power quality evaluation value.

[0016] The present invention has the following beneficial effects: (1) The photovoltaic equipment control method constructs a full-link power quality assessment system from the module end to the inverter end by collecting the photovoltaic inverter operation sequence data and photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment. First, the inverter operation sequence data is classified and integrated into control and state subsets. After covariance matrix analysis, sample entropy processing and inverter collaborative assessment model calculation, the inverter power quality assessment value is accurately extracted. Then, through the thermal imaging power quality mapping model with LSTM layer, the thermal imaging data is subjected to thermoelectric correlation feature extraction and time sequence fusion to obtain the photovoltaic power quality mapping assessment value. Finally, it is input into the comprehensive power quality assessment model to achieve deep fusion of the two types of assessment values, thereby comprehensively covering the core power quality influencing factors of module thermal state and inverter operation, realizing the collaborative perception of photovoltaic power quality, and enabling the grid-connected power quality assessment value to fully reflect the power quality status of the entire link, so as to ensure the stable adaptation of photovoltaic equipment and grid.

[0017] (2) The photovoltaic equipment control method achieves accurate characterization of inverter operating state electrical quality through the design of inverter electrical quality assessment process. Specifically, the photovoltaic inverter operating sequence data is first preprocessed and classified into inverter regulation time sequence subset and inverter state time sequence subset to ensure the relevance of data dimensions. Then, a regulation covariance matrix is ​​constructed based on the regulation subset to accurately extract the regulation adaptation assessment value at each time point, and the steady-state assessment value of the operating condition is obtained simultaneously. Then, the inherent laws of the data are mined through sample entropy analysis, and the collaborative calculation is performed in combination with the inverter collaborative assessment model. Finally, the random noise is effectively filtered through moving average processing so that the inverter electrical quality assessment value can accurately reflect the inverter's regulation adaptation capability and operating condition stability level.

[0018] (3) The photovoltaic equipment control method realizes the accurate conversion of photovoltaic module thermal data to electrical parameters by constructing a thermal imaging electrical quality mapping model. It receives each frame of photovoltaic surface thermal imaging data and completes data cleaning to remove invalid interference information. Then, it obtains thermal sensing feature vectors through thermal feature extraction and processing, and establishes the correlation between thermal state and electrical quality through electrical quality mapping transformation to generate electrical quality mapping feature vectors. Finally, it uses the temporal correlation capability of the LSTM layer to fuse the feature vectors of multiple frames and outputs the electrical quality mapping dataset. It not only realizes the structured transformation of unstructured thermal imaging data, but also accurately quantifies the impact of thermal anomalies such as hot spots and distortion on power quality, providing comprehensive module-side electrical quality data for full-link electrical quality assessment, so as to accurately characterize the electrical quality risks related to module thermal state.

[0019] (4) The photovoltaic equipment control method system, through modular closed-loop architecture design, realizes efficient coordination of the entire process of photovoltaic equipment power quality, significantly improving the timeliness of control response. The data acquisition module acquires the inverter running sequence data and photovoltaic surface thermal imaging sequence data, providing comprehensive input for subsequent analysis. The inverter power quality analysis module and the thermal imaging power quality mapping integration module have clear division of labor, focusing on the power quality assessment of the inverter end and the module end respectively, and then generating grid-connected power quality assessment values ​​through comprehensive analysis to ensure the comprehensiveness of the assessment results. The power quality control module directly executes control processing based on the assessment results, forming a complete closed loop, thereby avoiding control disconnect caused by data flow lag or information fragmentation, thus helping photovoltaic equipment to continuously and stably optimize power quality under diverse operating conditions.

[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0021] Figure 1 This is a flowchart of a photovoltaic equipment control method according to the present invention.

[0022] Figure 2 This is a flowchart illustrating the specific steps involved in extracting the inverter power quality evaluation value of a target photovoltaic device using a photovoltaic device control method according to the present invention.

[0023] Figure 3 This is a system block diagram of a photovoltaic equipment control method according to the present invention. Detailed Implementation

[0024] Please see Figure 1This invention provides a technical solution: a photovoltaic equipment control method, comprising the following steps: acquiring photovoltaic inverter operation sequence data and photovoltaic surface thermal imaging time sequence data of a target photovoltaic equipment within a set period; performing correlation and integration processing on the photovoltaic inverter operation sequence data of the target photovoltaic equipment to extract the inverter power quality evaluation value of the target photovoltaic equipment; performing power mapping processing on the photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment based on a pre-trained thermal imaging power quality mapping model to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment, and performing comprehensive analysis in conjunction with the inverter power quality evaluation value to obtain the grid-connected power quality evaluation value of the target photovoltaic equipment; and performing power quality control processing on the target photovoltaic equipment based on the grid-connected power quality evaluation value, specifically as follows: Determine whether the grid-connected power quality assessment value of the target photovoltaic equipment is higher than the preset grid-connected power quality assessment threshold. If it is higher than the preset grid-connected power quality assessment threshold, no control is performed. If it is not higher than the preset grid-connected power quality assessment threshold, control is performed, such as adjusting the inverter's PWM modulation strategy or harmonic filtering parameters to enhance the suppression capability of core harmonics such as the 3rd, 5th, and 7th orders, directly reducing the harmonic distortion rate on the grid-connected side; enabling the photovoltaic's built-in reactive power compensation module (such as a static var generator SVG), or adjusting its reactive power output to compensate for reactive power fluctuations on the grid side and stabilize the grid-connected voltage amplitude.

[0025] The specific steps to obtain the grid-connected power quality assessment value of the target photovoltaic equipment are as follows: Read the inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment, and perform standardization processing (the tanh function can be used to normalize its value to between 0 and 1). The standardized inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment are input into the preset comprehensive power quality assessment model to extract the grid-connected power quality assessment value of the target photovoltaic equipment. The comprehensive power quality assessment model is as follows: ;in, The grid-connected power quality assessment value for the target photovoltaic equipment. The standardized inverter power quality evaluation value of the target photovoltaic equipment. The inverter power supply regulation coefficients are stored in the database. The standardized photovoltaic power quality mapping evaluation value of the target photovoltaic equipment. The photovoltaic mapping adjustment coefficients are stored in the database. , These are the collaboration coefficients stored in the database.

[0026] In the model , The acquisition steps are as follows: Obtain the historical inverter power quality assessment values ​​and historical photovoltaic power quality mapping assessment values ​​for several historical periods of the target photovoltaic equipment after standardization. Calculate the average historical inverter power quality assessment values ​​and the average historical photovoltaic power quality mapping assessment values, and sum them to obtain the power quality sum value. Ratio the average historical inverter power quality assessment values ​​and the average historical photovoltaic power quality mapping assessment values ​​to the power quality sum value, and use the corresponding results as the power quality sum value. , .

[0027] In the model The acquisition steps are as follows: Read the historical inverter power quality assessment values ​​and historical photovoltaic power quality mapping assessment values ​​for several historical periods of the target photovoltaic equipment after standardization. Based on the Pearson correlation coefficient, extract the correlation coefficient value between the two and use it as... .

[0028] In the model The results of the basic weights are corrected to constrain the deviation of "good single-end power quality" and "poor single-end power quality", reflecting the coordinated matching quality of the entire photovoltaic-to-inverter link. Utilizing the characteristics of the geometric mean short-board constraint, if the electrical quality at one end is extremely poor, even if the electrical quality at the other end is extremely good, the result will be significantly reduced—which aligns with the fact that photovoltaic-inverter is a series system, and a single-end short board will lower the overall electrical quality. If the difference in electrical quality between the two ends is too large, the result of this item will approach 0, and the correction effect will be weakened—which is in line with the operating logic that "incoordination of electrical quality between the two ends will lead to inverter output fluctuations and harmonic superposition."

[0029] Specifically, such as Figure 2 As shown, the photovoltaic inverter's runtime sequence data includes the (reactive) power distribution value, inverter modulation ratio, voltage deviation value, (PLL) phase deviation value, IGBT junction temperature deviation value, and inverter efficiency value at each time point. The specific steps for extracting the inverter power quality evaluation value of the target photovoltaic equipment are as follows: Read the operating sequence data of the photovoltaic inverter of the target photovoltaic equipment and preprocess it, such as using min-max normalization to remove units from the parameters of the photovoltaic inverter operating sequence data so that their values ​​are mapped between 0 and 1. Then, classify the preprocessed operating sequence data of the target photovoltaic equipment to obtain the inverter control timing subset and the inverter state timing subset, specifically: The preprocessed photovoltaic inverter runtime sequence data is split into active control operation and passive state feedback logic to ensure that both subsets retain the standardized parameter values ​​at each time point. Inverter control timing subset: Focusing on the core operating parameters of the inverter actively adjusting power quality, this subset includes pre-processed (reactive) power allocation values, inverter modulation ratio, and (PLL) phase deviation values. These parameters are all control commands or synchronous tracking parameters actively output by the inverter based on the grid-connected control strategy, which directly determine the direction and intensity of voltage, phase, and power adjustment. They are the input operating parameters for achieving power quality optimization, and their timing data reflects the changes in the inverter's control behavior at different time points. Inverter Status Timing Subset: Focusing on the power quality results and hardware operating status parameters after inverter regulation, this subset includes pre-processed voltage deviation values, IGBT junction temperature deviation values, and inverter efficiency values. These parameters are all passive feedback results after inverter regulation operations. Among them, the voltage deviation value directly reflects the power quality regulation effect, the IGBT junction temperature deviation value reflects the hardware load status, and the inverter efficiency value reflects the power conversion efficiency. Its timing data reflects the operating status of the inverter at different time points. The inverter control timing subset and inverter state timing subset of the target photovoltaic equipment are respectively correlated and integrated to obtain the adjustment adaptation evaluation value (used to characterize the degree of fit between the inverter's active control operation and the power quality requirements of the grid) and the steady-state evaluation value (used to characterize the stability of the power quality steady-state operation) of the target photovoltaic equipment at each time point. Based on the adjustment adaptation evaluation value and the steady-state evaluation value of the target photovoltaic equipment at each time point, a collaborative evaluation process is performed to extract the inverter power quality evaluation value of the target photovoltaic equipment.

[0030] The reactive power allocation value is the ratio of the reactive power output command value to the total power output command value of the photovoltaic inverter at this point in time. It can be obtained by receiving the grid connection voltage command value and the current measured output voltage value in real time through the inverter's built-in DSP, and by using the built-in PI regulation algorithm to instantly calculate the reactive power output command value that needs to be compensated, and then performing a ratio calculation with the total power output command value to obtain the power allocation value at the current point in time.

[0031] The inverter modulation ratio is the ratio of the amplitude of the AC fundamental voltage output by the inverter's PWM (Pulse Width Modulation) to the amplitude of the DC bus voltage at that point in time. It can be obtained by acquiring the instantaneous value of the DC bus voltage in real time through the DC bus voltage sensor, and the DSP can simultaneously obtain the target value of the AC fundamental voltage required by the power grid. The inverter modulation ratio is obtained by dividing the target value of the AC fundamental voltage by the instantaneous value of the DC bus voltage.

[0032] The voltage deviation value is the degree of difference between the instantaneous values ​​of the three-phase AC output voltage of the inverter at that time point. It can be obtained by synchronously collecting the instantaneous values ​​of the three-phase voltage at that time point through the three-phase voltage sensor built into the inverter, extracting the maximum instantaneous voltage value, the minimum instantaneous voltage value, and the instantaneous average voltage value, and performing ratio processing, i.e. (maximum instantaneous voltage value - minimum instantaneous voltage value) / instantaneous average voltage value, and using the result as the voltage deviation value at that time point.

[0033] The phase deviation value (PLL) is the instantaneous absolute difference between the phase of the three-phase voltage fundamental wave (50Hz) of the grid tracked in real time by the inverter's phase-locked loop (PLL) at that point in time and the phase of the three-phase AC voltage fundamental wave output by the inverter itself. It can be obtained in real time by the inverter's built-in three-phase voltage sensor, PLL and digital signal processor (DSP) according to the following steps: synchronously acquire the instantaneous analog signal of the grid three-phase voltage at that point in time, convert it into a digital signal after the signal conditioning circuit and transmit it to the PLL, and extract the grid voltage fundamental wave phase through the synchronous rotating coordinate system algorithm; at the same time, receive the voltage phase command (i.e. the inverter output voltage fundamental wave target phase) synchronously output by the inverter PWM, and then the DSP calculates it instantaneously according to the formula phase deviation value = |grid voltage fundamental wave phase - inverter output voltage fundamental wave phase| to obtain the (PLL) phase deviation value at that point in time.

[0034] The IGBT junction temperature deviation value is the real-time operating temperature deviation of the inverter's IGBT chip at that point in time. It can sense the junction temperature change in real time through the built-in temperature sensor of the IGBT, transmit the temperature signal to the DSP temperature acquisition, and instantly convert it through the built-in calibration algorithm to obtain the junction temperature value at the current time point. Then, it performs difference processing (taking the average value) with the IGBT rated temperature in the inverter equipment manual stored in the database (i.e., the average of its IGBT rated temperature range), and uses the result as the IGBT junction temperature deviation value.

[0035] The inverter efficiency value is the ratio of the instantaneous output power of the inverter's AC side to the instantaneous input power of the DC side at this point in time. It can be obtained by synchronously collecting instantaneous voltage and current data from the built-in voltage / current sensors on the DC and AC sides of the inverter, instantaneously calculating the input and output power using the power calculation algorithm built into the DSP, and then obtaining the inverter efficiency value at the current point in time through ratio calculation.

[0036] The specific steps to obtain the adjustment and adaptation evaluation value of the target photovoltaic equipment at each time point are as follows: Based on the inverter control time sequence subset of the target photovoltaic equipment, construct the control covariance matrix of the target photovoltaic equipment. Specifically, calculate the arithmetic mean of the standardized values ​​of each parameter at all time points in the full time series, namely the power distribution mean, the inverter modulation ratio mean, and the phase deviation mean. Then, center the three standardized parameter values ​​at each time point. That is, for the j-th parameter at the i-th time point (j=1 corresponds to the power distribution value, j=2 corresponds to the inverter modulation ratio value, and j=3 corresponds to the phase deviation value), subtract the arithmetic mean of the parameter from the standardized value of the parameter at that time point to obtain the centered parameter value. Arrange them in the original time sequence to form a centered time series matrix. Each row of the matrix corresponds to a time point, and each row contains three elements, which are the centered values ​​of the three parameters at that time point. The covariance between each pair of parameters is calculated according to the covariance calculation rules, resulting in a 3×3 control covariance matrix. In this matrix, the elements on the diagonal correspond to the variances of the three parameters respectively, and the elements off the diagonal correspond to the covariances between two different parameters. Based on the regulation covariance matrix of the target photovoltaic equipment, the regulation coefficient set of the target photovoltaic equipment is analyzed. Specifically, the regulation covariance matrix is ​​decomposed into eigenvalues ​​to obtain three eigenvalues ​​(sorted from largest to smallest, and each eigenvalue corresponds to a principal component, which is a linear combination of the original three parameters, namely power distribution value, inverter modulation ratio, and phase deviation value) and three corresponding unit eigenvectors (each element of each eigenvector corresponds to a linear combination coefficient of the original parameters). For each principal component, its variance contribution rate is equal to the eigenvalue corresponding to that principal component divided by the sum of the three eigenvalues. For each principal component, the sum of the absolute values ​​of all elements in its eigenvector is calculated; and the contribution ratio of each original parameter in that principal component is calculated, which is the absolute value of the element corresponding to that original parameter in the eigenvector divided by the sum of the absolute values ​​of all elements in that eigenvector; the variance contribution rate of the original parameter allocated from that principal component is calculated, which is the variance contribution rate of that principal component multiplied by the corresponding contribution ratio; the variance contribution rates of the three principal components allocated to the same original parameter are summed to obtain the final variance contribution rate corresponding to that original parameter, and this is used as the corresponding adjustment coefficient, namely the power allocation adjustment coefficient, modulation ratio adjustment coefficient, and phase deviation adjustment coefficient, to form a set of control coefficients; Based on the control coefficient set and inverter control timing subset of the target photovoltaic equipment, the adjustment adaptation evaluation value of the target photovoltaic equipment at each time point is extracted. Specifically, for the power allocation value and inverter modulation ratio value of the target photovoltaic equipment at each time point, the average power allocation value and the average inverter modulation ratio value are extracted respectively. The power allocation value and inverter modulation ratio value at each time point are then compared with the average power allocation value and the average inverter modulation ratio value (absolute value is taken) to obtain the power allocation deviation value and inverter modulation ratio deviation value at each time point. Combined with the phase deviation value at the corresponding time point, the power allocation adjustment coefficient, modulation ratio adjustment coefficient, and phase deviation adjustment coefficient are weighted and fused to obtain the adjustment adaptation evaluation value at each time point. It should be noted that the phase deviation value needs to be transformed using the reciprocal suppression mapping function f(x)=1 / (1+x) during the weighted fusion process.

[0037] The specific steps for obtaining the steady-state evaluation value of the target photovoltaic equipment at each time point are as follows: Based on the time-series subset of the inverter state of the target photovoltaic equipment, extract the voltage deviation adjustment coefficient, junction temperature deviation adjustment coefficient, and inverter efficiency value adjustment coefficient of the target photovoltaic equipment (the acquisition logic is consistent with the control coefficient set). The voltage deviation value, IGBT junction temperature deviation value, and inverter efficiency value at each time point are weighted and fused with the voltage deviation adjustment coefficient, junction temperature deviation adjustment coefficient, and inverter efficiency value adjustment coefficient to obtain the steady-state evaluation value of the operating condition at each time point. It should be noted that the voltage deviation value and IGBT junction temperature deviation value need to be transformed using the reciprocal suppression mapping function f(x)=1 / (1+x) during the weighted fusion process.

[0038] It should be noted that the adjustment and adaptation evaluation values ​​and the steady-state evaluation values ​​are both normalized, mapping their values ​​to the same range.

[0039] The specific steps of the collaborative evaluation process are as follows: For each time point of the target photovoltaic equipment, the adjustment and adaptation evaluation value and the steady-state evaluation value are analyzed using sample entropy analysis to obtain the inverter coefficient set of the target photovoltaic equipment (including the adaptation weight coefficient and the steady-state weight coefficient). Specifically, based on the adjustment and adaptation evaluation value and the steady-state evaluation value at each time point, the corresponding information entropy value is extracted. Then, the corresponding information entropy value is transformed using the reciprocal suppression mapping function f(x)=1 / (1+x), such as 1 / (1+information entropy value of the adjustment and adaptation evaluation value), and summed to obtain the information entropy sum. The transformed information entropy value is then compared with the information entropy sum to obtain the weight coefficients corresponding to each parameter, namely the adaptation weight coefficient and the steady-state weight coefficient. The inverter coefficient set of the target photovoltaic equipment, as well as the adjustment and adaptation evaluation value and the steady-state evaluation value of the operating condition at each time point, are input into the preset inverter collaborative evaluation model to extract the power quality evaluation value at the corresponding time point. The power quality evaluation value of the target photovoltaic equipment at each time point is processed by moving average (the moving window can be 10-20 time points, and the moving time step can be 1 time point) to obtain the inverter power quality evaluation value of the target photovoltaic equipment.

[0040] The inverter collaborative evaluation model is as follows: ;in, , , The first two are the target photovoltaic devices. Electrical quality assessment values, regulation and adaptation assessment values, and steady-state assessment values ​​at various time points. , The following are, in order, the adaptation weight coefficient and the steady-state weight coefficient of the target photovoltaic equipment. This is the safety margin stored in the database (in this implementation example, the value is 0.2 to avoid it being zero). 1, 2, 3, , This represents the total number of time points.

[0041] Inverter Collaborative Evaluation Model It is used to measure the degree of balance between the two indicators, reflecting the principle that the better the synergy between the indicators, the higher the electrical quality assessment value.

[0042] The specific steps for calculating the power quality assessment value of the target photovoltaic equipment at a certain time point are as follows. The available data includes: adjustment and adaptation assessment values ​​and steady-state assessment values ​​for the target photovoltaic equipment at five randomly selected time points, as shown in Table 1.

[0043] Adaptation weighting coefficient of target photovoltaic equipment Approximately 0.462; Steady-state weighting coefficient of target photovoltaic equipment Approximately 0.538; Substituting the data in Table 1 and the above coefficients into the preset inverter collaborative evaluation model, we get: The electrical quality assessment value of the target photovoltaic equipment at the first time point = (0.462×0.723+0.538×0.882)×((min(0.462×0.723, 0.538×0.882) / max(0.462×0.723, 0.538×0.882))+0.2)≈1.049; The electrical quality assessment value of the target photovoltaic equipment at the second time point = (0.462×0.718+0.538×0.867)×((min(0.462×0.718,0.538×0.867) / max(0.462×0.718,0.538×0.867))+0.2)≈1.037; The electrical quality assessment value of the target photovoltaic equipment at the third time point = (0.462×0.692+0.538×0.874)×((min(0.462×0.692, 0.538×0.874) / max(0.462×0.692, 0.538×0.874))+0.2)≈1.004; The electrical quality assessment value of the target photovoltaic equipment at the fourth time point = (0.462×0.706+0.538×0.901)×((min(0.462×0.706, 0.538×0.901) / max(0.462×0.706, 0.538×0.901))+0.2)≈1.020; The electrical quality assessment value of the target photovoltaic equipment at the fifth time point = (0.462×0.712+0.538×0.912)×((min(0.462×0.712,0.538×0.912) / max(0.462×0.712,0.538×0.912))+0.2)≈1.028.

[0044] In this implementation plan, the parameters are first standardized using Min-Max, and then the time-series subsets are split according to active control and passive state. This ensures data comparability and clarifies the functional positioning of different parameters. Secondly, the evaluation dimensions are comprehensive. By constructing a covariance matrix and decomposing eigenvalues, the variance contribution rate of each parameter is quantified, and the dual-dimensional evaluation values ​​of regulation adaptation and steady-state operation are accurately extracted. The former focuses on the fit between active control and grid demand, while the latter reflects the stability of the operating state, achieving comprehensive coverage of the control process and operating results. Finally, a reciprocal suppression mapping function is used to handle negative parameters, and weights are objectively allocated through sample entropy analysis. Combined with a collaborative model containing a balance correction factor, the dynamic fusion of the two evaluation values ​​is achieved. Finally, noise is filtered by moving average, highlighting the parameter synergy effect. This ensures that the inverter power quality evaluation value can truly reflect the power quality assurance capability of the inverter, providing highly reliable support for the comprehensive power quality analysis of the entire chain.

[0045] Specifically, the photovoltaic surface thermal imaging time series data includes several frames of photovoltaic surface thermal imaging data. The photovoltaic surface thermal imaging data specifically includes the temperature value of each pixel in the thermal imaging and its corresponding two-dimensional coordinates. The thermal image electro-mass mapping model includes an input layer, an electro-mass mapping layer, an LSTM layer, and an output layer.

[0046] The specific steps to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment are as follows: Input the time series data of the photovoltaic surface thermal imaging of the target photovoltaic equipment into the pre-trained thermal imaging power quality mapping model, extract the power quality mapping dataset of the target photovoltaic equipment, including the hot spot harmonic tendency evaluation value, distortion correlation evaluation value, and voltage flicker evaluation value; Based on the power quality mapping dataset of the target photovoltaic equipment, extract the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment, specifically: read the hot spot harmonic tendency evaluation value, distortion correlation evaluation value, and voltage flicker evaluation value of the target photovoltaic equipment, perform weighted summation processing, and transform the weighted summation processing result using the reciprocal suppression mapping function f(x)=1 / (1+x) to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment (used to characterize the overall excellence of grid-connected power quality after the photovoltaic surface thermal anomaly is transformed by thermal mapping).

[0047] It should be noted that the weighting coefficients corresponding to the hot spot harmonic tendency assessment value, distortion correlation assessment value, and voltage flicker assessment value can be obtained by acquiring historical hot spot harmonic tendency assessment values, historical distortion correlation assessment values, and historical voltage flicker assessment values ​​for several historical periods, and then using sample entropy analysis to obtain the weighting coefficients corresponding to the above parameters.

[0048] The specific steps for extracting the electrical quality mapping dataset of the target photovoltaic device are as follows: In the input layer (thermal imaging electrical quality mapping model), each frame of photovoltaic surface thermal imaging data of the target photovoltaic device is received and data cleaning is performed. Specifically, a physical temperature threshold range of -20℃ to 85℃ is set based on the actual working temperature characteristics of the photovoltaic module. Sensor noise or abnormal pixels that exceed this range are removed, and the average temperature of the effective pixels in its 8-neighborhood is used to fill the gap. A region mask matrix is ​​constructed in combination with the preset physical size of the photovoltaic device. Effective pixels covering the photovoltaic panel body are filtered by coordinate bounding boxes, and background interference such as brackets and ground is removed. A 3×3 neighborhood median filtering algorithm is used to replace isolated noise pixels with an absolute value of difference from the neighborhood median > 5℃ with the median, while retaining key feature edge information such as hot spots and temperature gradients. In the electro-mass mapping layer (of the thermal imaging electro-mass mapping model), thermo-electric correlation feature processing is performed on each frame of photovoltaic surface thermal imaging data of the target photovoltaic device after data cleaning and processing, and the electro-mass mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is extracted. In the LSTM layer (of the thermal imaging electrical quality mapping model), the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is processed temporally to obtain the temporally fused electrical quality feature vector of the target photovoltaic device. Specifically, the LSTM layer adaptively captures the temporal dependencies of electrical quality features between different frames through a forget gate, input gate, and output gate gating mechanism. For example, it captures the continuous fluctuation pattern of voltage flicker features and the temporal evolution logic of the distortion grid stability correlation features. At the same time, it filters redundant temporal noise and retains key dynamic change information. By dynamically updating the temporal dimension of the electrical quality mapping features of multiple frames, the correlation information between the features of each frame and the features of historical frames is embedded into the feature expression of the current frame. Finally, the output is a temporally fused electrical quality feature vector with the same length as the input sequence, such as: For the hot spot induced harmonic tendency features, distortion grid stability correlation features, and divergence voltage flicker features in the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging, a sliding window of 5 frames is used, with the sliding step size set to one frame. The above features are then processed by moving average to obtain the hot spot harmonic tendency features, distortion correlation features, and voltage flicker features. In the output layer (of the thermal image electrical quality mapping model), based on the time-series fused electrical quality feature vector of the target photovoltaic device, the electrical quality mapping dataset of the target photovoltaic device is output. Specifically, the hot spot harmonic tendency features, distortion correlation features, and voltage flicker features are activated by the Sigmoid function to obtain hot spot harmonic tendency evaluation values, distortion correlation evaluation values, and voltage flicker evaluation values ​​with results between 0 and 1.

[0049] The pre-training steps for the thermal imaging electrical mass mapping model are as follows: Sample data from multiple photovoltaic devices under different operating conditions were collected. Each sample included synchronously collected time-series thermal imaging data of the photovoltaic surface (including the temperature value of each pixel and two-dimensional coordinates), as well as the corresponding time-dimensional measured power quality data of the grid-connected system (including harmonic distortion rate, grid subsynchronous oscillation amplitude, voltage flicker value, etc.). The thermal imaging data was cleaned according to the temperature threshold of -20℃ to 85℃, the region mask matrix, and the 3×3 neighborhood median filtering rule. The measured power quality data of the grid-connected system were labeled with corresponding hot spot harmonic tendency labels (based on harmonic distortion rate quantization), distortion correlation labels (based on subsynchronous oscillation amplitude quantization), and voltage flicker labels (based on voltage flicker value quantization), forming a labeled pre-training dataset, which was divided into training set, validation set, and test set in a 7:2:1 ratio.

[0050] Model structure initialization: A thermal imaging electro-mass mapping model framework including an input layer, an electro-mass mapping layer, an LSTM layer, and an output layer is built. The input layer is configured with a data receiving interface adapted to the dimensional format of the cleaned thermal imaging data. The electro-mass mapping layer initializes thermal feature extraction operators (eight-neighbor connectivity determination, isotherm curvature calculation, temperature gradient vector field / divergence field calculation modules), and presets initial values ​​for mapping coefficients / weights of hot spot-induced harmonic tendency, distorted power grid stability correlation, and divergence voltage flicker features (area correction coefficient initial value 0.03, temperature gradient weight coefficient initial value 0.2, curvature variation coefficient weight 0.4, etc.). The LSTM layer initializes the weight matrices and bias terms of the forget gate, input gate, and output gate, sets the hidden layer dimension to adapt to the length of the electro-mass mapping feature vector, and sets the sliding window initial value to 5 frames with a step size of 1 frame. The output layer incorporates a Sigmoid activation function to adapt the output of evaluation values ​​in the 0-1 range.

[0051] Training set data preprocessing and feature transformation: Thermal features are extracted from each frame of cleaned thermal imaging data in the training set. Indicators such as total hot spot area, average hot spot temperature gradient, and curvature variation coefficient are calculated and integrated into thermal sensing feature vectors. Based on preset mapping coefficients / weights, the thermal sensing feature vectors are transformed into hot spot-induced harmonic tendency, distortion grid stability correlation, and divergence voltage flicker features to form power quality mapping feature vectors. The time-series power quality mapping feature vectors are reconstructed according to the LSTM layer sliding window rule.

[0052] Model Iterative Training: The reconstructed electrochemical mapping feature vector is input into the model, which is then processed by an LSTM layer to capture temporal dependencies and output a temporally fused electrochemical feature vector. The output layer is then activated by a Sigmoid layer to obtain the predicted evaluation value. Using the labeled ground truth as the benchmark, mean squared error (MSE) is used as the loss function to calculate the loss between the predicted evaluation value and the ground truth. The model parameters (including the mapping coefficients / weights of the electrochemical mapping layer, the gating weight matrix of the LSTM layer, etc.) are updated through backpropagation using the Adam optimizer. The number of iterations is set (e.g., 200 iterations). After each iteration, the model accuracy is evaluated on the validation set. If the loss on the validation set does not decrease for 10 consecutive iterations, training is stopped.

[0053] Model validation and parameter tuning: Input the test set data into the trained model, output hot spot harmonic tendency, distortion correlation, and voltage flicker evaluation values, and calculate the mean absolute error (MAE) with the true label; if the MAE exceeds the preset threshold (e.g., 0.05), adjust the preset coefficients / weights of the electro-mapping layer and the sliding window size of the LSTM layer, and re-iterate the training; if the MAE meets the threshold requirement, lock the model parameters.

[0054] Model solidification and storage: Save the trained thermal imaging electro-mass mapping model structure and optimal parameter set (including mapping coefficients / weights of the electro-mass mapping layer, LSTM layer gating parameters, output layer activation function parameters, etc.), and generate a pre-trained model file that can be directly called for subsequent electro-mass mapping evaluation of target photovoltaic equipment.

[0055] The specific steps for extracting the electrochemical mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device are as follows: Perform thermal feature extraction processing on each frame of photovoltaic surface thermal imaging data of the target photovoltaic device to obtain the corresponding frame's thermal sensing feature vector, specifically: Using an eight-neighbor connectivity rule, regions with at least five consecutive connected pixels exceeding 60 degrees Celsius are selected from the effective area and identified as hotspot regions. The total number of hotspots (i.e., hotspot area) is also counted. For each identified hotspot region, the center pixel with the highest temperature and the edge pixel closest to the center on the hotspot boundary are located, and their physical coordinates and temperature values ​​are recorded. The physical straight-line distance between the center pixel and the edge pixel is calculated by measuring the coordinate differences along the x and y axes and using Euclidean distance. The temperature gradient of a single hotspot is obtained by dividing the temperature difference between the center pixel and the edge pixel by this physical straight-line distance. The arithmetic mean of the temperature gradients of all hotspots is then taken as the average temperature gradient of the hotspot. The highest, average, and lowest temperatures are statistically analyzed from the effective area. Based on these, three key isotherms are extracted: the isotherm corresponding to the highest temperature minus 10 degrees Celsius, the isotherm corresponding to the average temperature, and the isotherm corresponding to the lowest temperature plus 10 degrees Celsius. Each isotherm consists of a set of connected pixels whose pixel temperature falls within ±0.5 degrees Celsius of the target temperature. For each isotherm, multiple sliding analysis windows are formed using three consecutive adjacent pixels. Taking the pixel in the middle of the window as the reference, the coordinate difference between the pixel and its adjacent preceding and following pixels is calculated to obtain two adjacent tangent vectors. The change in the angle between the two tangent vectors within each sliding window is calculated using the vector angle formula combined with the inverse cosine function, which represents the local curvature of that window. The standard deviation and mean of the local curvature values ​​of all sliding windows are statistically analyzed, and the ratio of the two is the curvature variation coefficient. At the same time, the number of discontinuous breakpoints in each isotherm is counted, and the break density is obtained by dividing the number of breakpoints by the total length of the isotherm (i.e., the total number of pixels). The pairwise spacing between the three isotherms is calculated (i.e., for each isotherm, the geometric center coordinates are calculated based on the two-dimensional coordinates of all its pixels, and then the straight-line distance between the geometric centers of the three sets of isotherms is calculated using the Euclidean distance formula), and the ratio of the maximum spacing to the minimum spacing is taken to obtain the spacing non-uniformity. Discrete difference operations are performed on the temperature and corresponding coordinates of all pixels within the effective area to obtain the temperature change rate of each pixel in the x-axis and y-axis directions. This constructs the temperature gradient vector field of each pixel. For a single pixel, the temperature difference between itself and its adjacent right pixel in the x-axis direction is taken and divided by the coordinate distance between the two pixels (determined based on the Euclidean distance formula) to obtain the temperature change rate of the pixel in the x-axis direction. Similarly, the temperature difference between itself and its adjacent lower pixel in the y-axis direction is taken and divided by the physical coordinate distance between the two pixels to obtain the temperature change rate of the pixel in the y-axis direction. The temperature change rates in these two directions constitute the temperature gradient vector of the pixel. The divergence field is obtained by discretizing the temperature gradient vector field using a second-order difference operation. Specifically, for the temperature change rate of each pixel along the x-axis, the difference between its x-axis adjacent pixels is calculated and divided by the coordinate spacing; for the temperature change rate of the pixel along the y-axis, the difference between its y-axis adjacent pixels is calculated and divided by the coordinate spacing; the two results are added together to obtain the divergence value of the pixel. Pixels with positive divergence values ​​represent heat source regions where heat diffuses outward, and pixels with negative divergence values ​​represent heat sink regions where heat converges inward. The proportion of positive divergence pixels in the divergence field to the total number of pixels in the effective region is the positive divergence region proportion, which is used to characterize the spatial proportion of the heat source region; the proportion of negative divergence pixels to the total number of pixels in the effective region is the negative divergence region proportion, which is used to characterize the spatial proportion of the heat sink region; the absolute value of the divergence values ​​of all effective pixels is calculated, and then the arithmetic mean of these absolute values ​​is taken to obtain the mean absolute divergence value. The total area of ​​hot spots, the average temperature gradient of hot spots, the coefficient of curvature variation, the fracture density, the spacing non-uniformity, the proportion of positive divergence regions, the proportion of negative divergence regions, and the mean absolute value of divergence are integrated in a fixed order to form a thermal sensing feature vector. The thermal sensing feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is mapped to obtain the electrical quality mapping feature vector of the corresponding frame of photovoltaic surface thermal imaging, specifically as follows: The total area and average temperature gradient of the hot spot in the thermal sensing feature vector are called up, and two preset coefficients are set: area correction coefficient (value range 0.01-0.05, determined according to the rated power calibration of the photovoltaic module) and temperature gradient weighting coefficient (value range 0.1-0.3, determined according to the thermal conduction characteristics of the module). The total area of ​​the hot spot is multiplied by the area correction coefficient to obtain the quantitative contribution value of the hot spot area. Then, the average temperature gradient of the hot spot is multiplied by the temperature gradient weighting coefficient to obtain the quantitative contribution value of the hot spot temperature gradient. This value is then multiplied by the quantitative contribution value of the hot spot area to obtain the hot spot-induced harmonic tendency characteristic, which characterizes the overall emission tendency of the hot spot to the core subharmonic, reflecting the harmonic pollution risk when the photovoltaic equipment is connected to the grid. The larger the value, the higher the comprehensive emission intensity of the specific subharmonic induced by the hot spot, and the more serious the harmonic distortion problem in power quality. The curvature variation coefficient, fracture density, and spacing non-uniformity in the thermal sensing feature vector are called and assigned preset weight coefficients, with the curvature variation coefficient having a weight of 0.4, fracture density having a weight of 0.3, and spacing non-uniformity having a weight of 0.2. The sum of the three weight coefficients is one. This weight allocation is determined based on the experimental calibration of the significance of each indicator's impact on power grid stability. The value of each indicator is multiplied by the corresponding weight coefficient, and the three product results are added together to obtain the distortion power grid stability correlation feature, which is used to characterize the risk of subsynchronous oscillation and small-signal stability margin of the power grid. The larger the value, the higher the risk of subsynchronous oscillation, the lower the small-signal stability margin, and the worse the power supply stability in power quality. The method calls upon the proportions of positive and negative divergence regions and the mean absolute value of divergence from the thermal sensing feature vector. First, it calculates the difference between the proportions of positive and negative divergence regions and multiplies this difference by a preset coefficient of 0.6. Simultaneously, it multiplies the mean absolute value of divergence by a preset coefficient of 0.4 and then adds the two products together to obtain the divergence voltage flicker feature. This accurately maps the voltage flicker risk caused by power fluctuations. The larger the value, the higher the severity of voltage flicker and the worse the voltage stability in power quality. The characteristics of hot spot-induced harmonic tendency, distortion grid stability correlation, and divergence voltage flicker are integrated in a fixed order to form a power quality mapping feature vector.

[0056] In this implementation scheme, abnormal pixels are removed by physical temperature thresholding, background interference is filtered by region masking matrix, and isolated noise is corrected by median filtering. While ensuring data purity, key thermal features such as hot spots and temperature gradients are fully preserved, providing high-quality input for subsequent mapping. Secondly, the thermal feature extraction is comprehensive and accurate. Hot spot regions are determined based on eight-neighbor connectivity. Combined with multi-dimensional analysis such as isotherm curvature, temperature gradient vector field, and divergence field, core features such as hot spot area, temperature gradient, and heat diffusion state are quantified, comprehensively capturing the spatial distribution and evolution of component thermal anomalies. Finally, through preset coefficient calibration and weight allocation, thermal features are accurately converted into quantitative electrical quality parameters such as hot spot harmonic tendency. The temporal dependency is captured by LSTM layer, and the weights are determined by sample entropy analysis and the evaluation value is calculated by combining the suppression mapping function. This not only realizes the effective conversion of unstructured thermal imaging data into structured electrical quality assessment, but also dynamically reflects the degree of impact of thermal anomalies on grid-connected power quality, providing accurate component-side data support for full-link electrical quality assessment.

[0057] Please see Figure 3 This invention provides a technical solution: a photovoltaic equipment control method system, comprising: a data acquisition module for acquiring photovoltaic inverter operation sequence data and photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment; an inverter power quality analysis module for performing correlation and integration processing on the photovoltaic inverter operation sequence data of the target photovoltaic equipment to extract the inverter power quality evaluation value of the target photovoltaic equipment; a thermal imaging power quality mapping integration module for performing power mapping processing on the photovoltaic surface thermal imaging time sequence data of the target photovoltaic equipment based on a pre-trained thermal imaging power quality mapping model to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment, and performing comprehensive analysis in conjunction with the inverter power quality evaluation value to obtain the grid-connected power quality evaluation value of the target photovoltaic equipment; and a power quality control module for performing power quality control processing on the target photovoltaic equipment based on the grid-connected power quality evaluation value.

[0058] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0059] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A photovoltaic equipment control method, characterized in that, Includes the following steps: Acquire the photovoltaic inverter's runtime timing data and the photovoltaic surface thermal imaging timing data of the target photovoltaic equipment; The operation sequence data of the photovoltaic inverter of the target photovoltaic equipment are correlated and integrated to extract the inverter power quality evaluation value of the target photovoltaic equipment. Based on the pre-trained thermal imaging power quality mapping model, the power quality mapping of the photovoltaic surface thermal imaging time series data of the target photovoltaic device is processed to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic device. Combined with the inverter power quality evaluation value, a comprehensive analysis is performed to obtain the grid-connected power quality evaluation value of the target photovoltaic device. Power quality control processing is performed on the target photovoltaic equipment based on the grid-connected power quality assessment value.

2. The photovoltaic equipment control method according to claim 1, characterized in that, The photovoltaic inverter runtime sequence data includes power distribution value, inverter modulation ratio, voltage deviation value, phase deviation value, IGBT junction temperature deviation value, and inverter efficiency value at each time point. The specific steps for extracting the inverter power quality evaluation value of the target photovoltaic equipment are as follows: Read the operating sequence data of the photovoltaic inverter of the target photovoltaic equipment and perform preprocessing; The preprocessed photovoltaic inverter runtime timing data of the target photovoltaic equipment is classified and processed to obtain the inverter control timing subset and inverter state timing subset of the target photovoltaic equipment. The inverter control timing subset and inverter state timing subset of the target photovoltaic equipment are respectively correlated and integrated to obtain the adjustment adaptation evaluation value and the steady-state evaluation value of the target photovoltaic equipment at each time point. Based on the adjustment and adaptation evaluation value and the steady-state evaluation value of the target photovoltaic equipment at each time point, a collaborative evaluation process is performed to extract the inverter power quality evaluation value of the target photovoltaic equipment.

3. The photovoltaic equipment control method according to claim 2, characterized in that, The specific steps to obtain the adjustment and adaptation evaluation value of the target photovoltaic equipment at each time point are as follows: Based on the inverter regulation time series subset of the target photovoltaic equipment, the regulation covariance matrix of the target photovoltaic equipment is constructed. Based on the regulation covariance matrix of the target photovoltaic equipment, the set of regulation coefficients of the target photovoltaic equipment is analyzed. Based on the control coefficient set of the target photovoltaic equipment and the inverter control timing subset, the adjustment adaptation evaluation value of the target photovoltaic equipment at each time point is extracted.

4. The photovoltaic equipment control method according to claim 3, characterized in that, The specific steps of the collaborative assessment process are as follows: Sample entropy analysis was performed on the adjustment and adaptation evaluation value and the steady-state evaluation value of the target photovoltaic equipment at each time point to obtain the inverter coefficient set of the target photovoltaic equipment. The inverter coefficient set of the target photovoltaic equipment, as well as the adjustment and adaptation evaluation value and the steady-state evaluation value of the operating condition at each time point, are input into the preset inverter collaborative evaluation model to extract the power quality evaluation value at the corresponding time point. The inverter power quality assessment value of the target photovoltaic equipment is obtained by performing a moving average on the power quality assessment value at each time point.

5. The photovoltaic equipment control method according to claim 1, characterized in that, The photovoltaic surface thermal imaging time series data includes several frames of photovoltaic surface thermal imaging data. Specifically, the photovoltaic surface thermal imaging data consists of the temperature value and corresponding two-dimensional coordinates of each pixel in the thermal imaging. The thermal image electro-mass mapping model includes an input layer, an electro-mass mapping layer, an LSTM layer, and an output layer.

6. The photovoltaic equipment control method according to claim 5, characterized in that, The specific steps to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment are as follows: The time-series thermal imaging data of the photovoltaic surface of the target photovoltaic device is input into the pre-trained thermal imaging electro-mass mapping model to extract the electro-mass mapping dataset of the target photovoltaic device, including hot spot harmonic tendency evaluation value, distortion correlation evaluation value, and voltage flicker evaluation value. Based on the target photovoltaic equipment's power quality mapping dataset, extract the photovoltaic power quality mapping evaluation value of the target photovoltaic equipment.

7. The photovoltaic equipment control method according to claim 6, characterized in that, The specific steps for extracting the power quality mapping dataset of the target photovoltaic equipment are as follows: In the input layer, each frame of photovoltaic surface thermal imaging data from the target photovoltaic device is received and the data is cleaned. In the electro-mass mapping layer, thermoelectric correlation feature processing is performed on each frame of photovoltaic surface thermal imaging data of the target photovoltaic device after data cleaning and processing, and the electro-mass mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is extracted. In the LSTM layer, the electrical quality mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is processed by temporal correlation to obtain the temporal fusion electrical quality feature vector of the target photovoltaic device; In the output layer, the electrical quality mapping dataset of the target photovoltaic device is output based on the time-series fused electrical quality feature vector of the target photovoltaic device.

8. The photovoltaic equipment control method according to claim 7, characterized in that, The specific steps for extracting the electro-mass mapping feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device are as follows: Thermal features are extracted from each frame of thermal imaging data of the target photovoltaic device to obtain the thermal sensing feature vector of the corresponding frame of thermal imaging of the photovoltaic surface. The thermal sensing feature vector of each frame of photovoltaic surface thermal imaging of the target photovoltaic device is mapped to obtain the electrical quality mapping feature vector of the corresponding frame of photovoltaic surface thermal imaging.

9. The photovoltaic equipment control method according to claim 1, characterized in that, The specific steps to obtain the grid-connected power quality assessment value of the target photovoltaic equipment are as follows: Read the inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment, and perform standardization processing; The standardized inverter power quality assessment value and photovoltaic power quality mapping assessment value of the target photovoltaic equipment are input into the preset comprehensive power quality assessment model to extract the grid-connected power quality assessment value of the target photovoltaic equipment.

10. A photovoltaic equipment control method system, employing the photovoltaic equipment control method according to any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire the operating sequence data of the photovoltaic inverter of the target photovoltaic equipment and the timing data of the photovoltaic surface thermal imaging. The inverter power quality analysis module is used to correlate and integrate the operating sequence data of the photovoltaic inverter of the target photovoltaic equipment and extract the inverter power quality evaluation value of the target photovoltaic equipment. The thermal imaging-electric quality mapping integration module is used to perform power mapping processing on the time series data of photovoltaic surface thermal imaging of the target photovoltaic device based on the pre-trained thermal imaging-electric quality mapping model, to obtain the photovoltaic power quality mapping evaluation value of the target photovoltaic device, and to perform comprehensive analysis in combination with the inverter power quality evaluation value to obtain the grid-connected power quality evaluation value of the target photovoltaic device. The power quality control module is used to perform power quality control processing on the target photovoltaic equipment based on the grid-connected power quality assessment value.

Images