A solar power generation system capable of detecting fires in the connection panel and inverter sensor failures

Abstract

The present invention relates to a photovoltaic power generation system capable of detecting fires in a junction box and inverter sensor failures, which detects fires through precise detection by utilizing a combination of image analysis and current and acoustic sensor data to detect minute arcs or sparks occurring inside a junction box, and also detects failure of a voltage sensor mounted inside an inverter that detects voltage, and enables continuous MPPT to be performed even if the voltage sensor fails, thereby improving power generation efficiency by continuously producing power during the maintenance of the inverter or replacement of the voltage sensor. The system comprises a plurality of strings composed of multiple photovoltaic modules that generate electricity using solar energy connected in series or parallel, a junction box that collects DC power produced from each of the strings, an inverter that receives DC power transmitted from the junction box, tracks MPPT using a boost converter equipped with voltage and voltage sensors for voltage and current measurement, converts the DC power into AC power for output, and a control server that receives and monitors fire and power generation information through communication with the junction box and inverter. It also includes an arc and spark detection unit configured inside the junction box that detects arcs or sparks by utilizing a combination of image analysis and current and acoustic sensor data, and a component configured inside the inverter It is characterized by comprising a sensor failure and control unit that detects when a voltage sensor detecting voltage fails and controls the continuous performance of MPPT even if the voltage sensor fails.

Description

Technology Field

[0001] The present invention relates to a photovoltaic power generation system, and more specifically, to a photovoltaic power generation system capable of detecting fire in a junction box and inverter sensor failure, which detects fire through precise detection by utilizing a combination of image analysis and current and acoustic sensor data to detect minute arcs or sparks occurring inside the junction box, and also detects failure of a voltage sensor mounted inside the inverter that detects voltage, and enables continuous MPPT to be performed even if the voltage sensor fails, thereby allowing power to be continuously produced during the maintenance of the inverter or replacement of the voltage sensor to improve power generation efficiency. Background Technology

[0002] Generally, a photovoltaic power generation system includes solar cell modules that convert light energy into direct current (DC) electrical energy, inverters that convert DC power into alternating current (AC) power, and junction boxes that connect the solar cell modules to the inverter and transmit the DC power produced by the solar cell modules to the inverter. The AC power converted by the inverter is stepped up and transmitted to a separate electrical grid system or supplied to a load for use.

[0003] Meanwhile, the inverter used in a solar power generation system basically consists of a boost converter that amplifies the voltage of the PV panel and an inverter that converts DC power into AC power.

[0004] That is, FIG. 1 is a schematic circuit diagram showing the interior of an inverter of a general photovoltaic power generation system. As shown in FIG. 1, it is equipped with a boost converter (2) that receives DC power produced from a PV array (1), amplifies the voltage, and outputs it, an inverter (3) that receives DC power output from the boost converter (2), converts it into AC power, and outputs it, and a current sensor (4) that detects current and a voltage sensor (5) that detects voltage at the input terminal of the boost converter (2).

[0005] Here, the PV array (1) is configured by connecting solar cells that convert incident solar energy into electrical energy in a series form, and is configured in the form of a PV module in which a plurality of PV arrays (1) are combined according to the amount of power required by the AC grid, so that DC power can be generated for each PV array (1).

[0006] In addition, the amount of energy of the incident sunlight and the amount of sunlight according to the angle of inclination with the sun change from moment to moment, and when multiple PV arrays (1) are configured in the system, the sunlight environment at the location where each PV array (1) is installed may differ, so the PV array (10) outputs irregular DC power.

[0007] The above boost converter (2) is connected to the output terminal of the PV array (1) with a voltage sensor (5) and boosts the low voltage transmitted from the PV array (1) to a level greater than a predetermined size and outputs it.

[0008] That is, the irregular DC power output from the PV array (1) is stabilized through the boost converter (2) and boosted to an appropriate size for output.

[0009] This boost converter (2) operates to receive a PWM control signal from the outside and boost the voltage to a set voltage of a certain magnitude. More specifically, it is configured to include a first switching element (Q1), an inductor (L1), and a diode (D1). When the first switching element (Q1) is turned ON according to the PWM control signal, the power transmitted from the PV array (1) is stored in the inductor (L1). When the first switching element (Q1) is turned OFF, the power stored in the inductor (L1) is applied to a capacitor (Cd) configured at the output terminal of the boost converter (2).

[0010] The inverter (3) is connected to the output terminal of the boost converter (2) and is equipped with a plurality of switching elements (Q2~Q7) to switch according to an external PWM control signal, thereby converting the power output from the boost converter (2) into a form and frequency of power to be supplied to the load (AC Grid) and supplying it to the load.

[0011] That is, the inverter (3) is connected to the output terminal of the boost converter (2) and is equipped with a plurality of switching elements (Q2~Q7) and is a component that converts the power output from the boost converter (2) into a form and frequency of power to be supplied to the power receiving unit by switching operation according to an external PWM control signal, wherein the number of each switching element (Q2~Q7) is determined according to the configuration pattern (three-phase switching method or single-phase switching method), and each switching element (Q2~Q7) is controlled by a pulse width modulation method at a given switching frequency to convert DC power into three-phase or single-phase AC power.

[0012] Accordingly, by considering the characteristics of the PV array (1) that can obtain maximum power by adjusting the load of the power receiving section, the maximum power operating point data that allows the PV array (1) to deliver maximum power according to the variation of the load (AC Grid) can be extracted through the MPPT algorithm.

[0013] At this time, programming data for controlling the size of the AC Grid by PWM controlling the boost converter (2) and inverter (3) according to the extracted data of the maximum power operating point is stored in advance, and the boost converter (2) and inverter (3) can be appropriately controlled by reading out PWM control data corresponding to the extracted data of the maximum power operating point.

[0014] As described above, the inverter used in a conventional photovoltaic power generation system is basically composed of a boost converter that amplifies the voltage of the PV array and an inverter (3) that converts DC power into AC power, and a current sensor (4) and a voltage sensor (5) are configured at the input terminal of the boost converter (2) to perform MPPT so that the PV array (1) can always generate power at the maximum power point.

[0015] However, if the current sensor or voltage sensor detecting current or voltage in an inverter according to conventional technology fails, MPPT cannot be performed and the operation of the solar power inverter must be stopped.

[0016] Ultimately, there was a problem causing economic losses because power could not be generated until the sensors or the solar inverters themselves were repaired or replaced.

[0017] In addition, channels (composed of input terminals, fuses, diodes, etc.) connected to the output terminals of multiple solar cell strings are arranged in the photovoltaic junction box and connected to each other so that DC power from the solar cell side is collected and transmitted to the inverter through the output terminals.

[0018] Various electrical components (input terminals, fuses, diodes, and wiring) constituting multiple channels are arranged in a dense array within the junction box. Consequently, if an overcurrent is generated in relation to a specific solar cell string or if internal ignition occurs in a specific channel due to overheating or a short circuit, the fire can rapidly spread to the entire junction box. In particular, since photovoltaic power generation systems are installed on building rooftops or outdoors, fires in the junction box often spread to buildings or forests, causing massive damage; therefore, early prediction of fire occurrence is crucial. Furthermore, especially when the installation area of ​​solar cell modules is large, it is necessary to accurately identify the areas where fire is predicted to occur.

[0019] Furthermore, if a fire in a solar panel or junction box is detected and power production and supply from the solar panel continue, the fire spreads very rapidly from the point of ignition due to the adjacent arrangement of electrical components or cables within the junction box, making suppression difficult.

[0020] Considering that early extinguishing within the first minute from the point of ignition is generally very important when a fire occurs, there is an urgent need for a means to quickly suppress a fire in a connection box.

[0021] In this regard, conventional fire detection methods that confirm the occurrence of an arc through an arc detector have limitations, such as the difficulty in detecting cases where overheating or deterioration occurs without an arc, and the inability to analyze and identify the channel that is the direct cause of the fire from multiple channels, or to perform centralized fire prevention through selective power cutoff of the identified channel.

[0022] [Prior Art Literature]

[0023] (Patent Document 1) Korean Patent Publication No. 10-2021-0147192 (Published Dec. 07, 2021)

[0024] (Patent Document 2) Korean Registered Patent Publication No. 10-2521644 (Registered on April 10, 2023)

[0025] (Patent Document 3) Korean Registered Patent Publication No. 10-2409193 (Registered June 10, 2022)

[0026] (Patent Document 4) Korean Patent Publication No. 10-2454755 (Registered Oct. 11, 2022)

[0027] (Patent Document 5) Korean Registered Patent Publication No. 10-1870300 (Registered June 18, 2018)

[0028] (Patent Document 6) Korean Registered Patent Publication No. 10-1830308 (Registered Feb. 12, 2018) The problem to be solved

[0029] The present invention was devised to solve the above-mentioned problems and aims to provide a photovoltaic power generation system capable of detecting fires in junction boxes and inverter sensor failures, which detects fires by precisely detecting minute arcs or sparks occurring inside the junction box through the combined use of image analysis and current and acoustic sensor data, and also detects when a voltage sensor mounted inside the inverter fails, thereby enabling continuous MPPT to be performed even when the voltage sensor fails, so that power can be continuously produced to improve power generation efficiency during the maintenance of the inverter or replacement of the voltage sensor. means of solving the problem

[0030] A photovoltaic power generation system capable of detecting fire in a junction box and inverter sensor failure according to the present invention for achieving the above-mentioned purpose comprises: a plurality of strings configured by connecting a plurality of photovoltaic modules that generate electricity using solar energy in series or in parallel; a junction box that collects DC power produced from each of the strings; an inverter that receives DC power transmitted from the junction box, tracks MPPT using a boost converter equipped with voltage and voltage sensors for measuring voltage and current, converts it into AC power, and outputs it; a control server that receives and monitors fire and power generation information through communication with the junction box and the inverter; an arc and spark detection unit configured inside the junction box that detects arcs or sparks by combining image analysis and current and acoustic sensor data; and a sensor failure and control unit configured inside the inverter that detects when a voltage sensor detecting voltage fails and controls the system to continuously perform MPPT even if the voltage sensor fails.

[0031] Here, the arc and spark detection unit comprises: an image capturing unit positioned to capture the interior of the junction box at close range; a sensor unit that detects current and sound within the junction box; a frame buffer that receives image data captured by the image capturing unit and stores it in frame units for a predetermined time interval; an interest region generation unit that generates an interest region corresponding to an arc / spark candidate based on short-term luminescence or rapid brightness change among the images stored in the frame buffer; an enhancement detection unit that divides the interest region generated by the interest region generation unit into a plurality of sub-regions, performs bounding box-based object detection in each sub-region, and merges the detection results for each sub-region to enhance detection of an arc / spark as a small object in the interest region; a feature calculation unit that calculates a current score based on high frequency features from the current signal of the sensor unit and / or an acoustic score based on arc-related features from the acoustic signal; and a confirmation determination unit that determines whether an arc / spark is confirmed by fusing at least one of the image-based value calculated by the enhancement detection unit and the current score and / or acoustic score calculated by the feature calculation unit. It is characterized by comprising a control signal generation unit that generates and outputs a control signal for alarm, output restriction, or blocking according to the confirmation judgment result of the confirmation judgment unit.

[0032] In addition, the sensor fault and control unit comprises: a fault threshold setting unit that sets a threshold value to determine whether the current sensor and voltage sensor are faulty; an MPPT execution unit that performs MPPT according to the current and voltage measured through the current sensor and voltage sensor; a control unit that receives the MPPT result performed through the MPPT execution unit and generates a switching control signal to control the boost converter by compensating for the difference between the target voltage and the voltage measured through the voltage sensor; a voltage calculation unit that receives the control signal through the control unit and calculates the voltage by subtracting the measured voltage from the target voltage to reduce the voltage error by PI control; a calculation value determination unit that receives the voltage value calculated through the voltage calculation unit and simultaneously calculates the voltage and current in the CCM and DCM to determine whether to operate in CCM or DCM depending on the current of the inductor constituting the boost converter; a sensor fault determination unit that receives the voltage value determined through the calculation value determination unit and determines whether the voltage sensor is faulty by comparing it with the threshold value set in the fault threshold setting unit; and, if a fault is determined based on the result of the sensor fault determination unit, the current It is characterized by including a control target conversion unit that calculates the current error used by changing to PI control by subtracting the measured current from the target current, performs PI control, and continuously performs MPPT. Effects of the invention

[0033] A solar power generation system capable of detecting connection box fire and inverter sensor failure according to an embodiment of the present invention has the following effects.

[0034] First, fire accidents can be prevented in advance by detecting fires through precise detection using a combination of video analysis and current and acoustic sensor data regarding minute arcs or sparks occurring inside the junction box.

[0035] Second, due to the image module and division-merging-based enhanced detection, the unit for arc and spark detection can be miniaturized while improving the detection rate.

[0036] Third, blocking reliability can be improved by suppressing image detection errors through current high frequency and / or acoustic sensor confirmation fusion.

[0037] Fourth, it can be operated in real-time embedded mode through selective reinforcement detection and computation control based on candidate ROI areas.

[0038] Fifth, accidents can be prevented proactively by storing frames and sensor features before and after an event.

[0039] Sixth, if the voltage sensor installed inside the inverter fails, it can detect this and continue to perform MPPT to generate power even if the voltage sensor fails, thereby eliminating economic losses that occur during the maintenance or replacement of the inverter. Brief explanation of the drawing

[0040] Figure 1 is a schematic circuit diagram showing the interior of an inverter of a typical photovoltaic power generation system. FIG. 2 is a schematic diagram showing a photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure according to the present invention. FIG. 3 is a configuration diagram showing the arc and spark detection unit in more detail in a photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure of FIG. 2. FIG. 4 is a configuration diagram showing the sensor failure and control unit in more detail in a photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure of FIG. 2. FIG. 5 is a circuit diagram showing a boost converter in a photovoltaic power generation system according to the present invention. Figure 6 is a graph showing the simulation results for verifying the PV voltage analysis in CCM mode. Figure 7 is a graph showing the simulation results for verifying PV voltage analysis in DCM mode. Figure 8 is a diagram showing the waveform of the change in calculated voltage according to the change in input voltage. Figure 9 is a diagram showing the waveform of the change in calculated voltage according to the enlarged input voltage change. Figure 10 is a graph showing the MPPT simulation results when using a voltage sensor. Figure 11 is an enlarged graph of a portion of Figure 10. Figure 12 is a graph showing the MPPT simulation results when continuous MPPT control was performed due to voltage sensor failure. Figure 13 is an enlarged graph of a portion of Figure 11 FIG. 14 is a flowchart illustrating a continuous control method using a sensor failure detection device of a photovoltaic power generation system according to the present invention. Specific details for implementing the invention

[0041] Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings so that those skilled in the art can easily implement them. It should be noted that the same reference numerals used to denote components in the accompanying drawings are used whenever possible when the same components are denoted in other drawings.

[0042] The present invention is capable of various modifications and may have various embodiments, and specific embodiments are illustrated in the drawings and described in detail in the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the present invention. In describing the present invention, detailed descriptions of related prior art are omitted if it is determined that such detailed descriptions may obscure the essence of the present invention.

[0043] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, terms such as “comprising” or “consisting of” are intended to indicate the presence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0044] FIG. 2 is a schematic diagram showing a photovoltaic power generation system capable of detecting a junction box fire and an inverter sensor failure according to the present invention, FIG. 3 is a more detailed diagram showing an arc and spark detection unit in the photovoltaic power generation system capable of detecting a junction box fire and an inverter sensor failure of FIG. 2, and FIG. 4 is a more detailed diagram showing a sensor failure and control unit in the photovoltaic power generation system capable of detecting a junction box fire and an inverter sensor failure of FIG. 2.

[0045] As illustrated in FIG. 2, the solar power generation system capable of detecting junction box fire and inverter sensor failure according to the present invention comprises: a plurality of strings (10) configured by connecting a plurality of solar modules that generate electricity using solar energy in series or in parallel; a junction box (20) that collects DC power produced from each string (10); an inverter (30) that receives DC power transmitted from the junction box (20), tracks MPPT using a boost converter equipped with voltage and voltage sensors for measuring voltage and current, converts it into AC power, and outputs it; a control server (40) that receives and monitors fire and power generation information through communication with the junction box (20) and the inverter (30); an arc and spark detection unit (100) configured inside the junction box (20) that detects arcs or sparks by combining image analysis and current and acoustic sensor data; and a sensor failure and a voltage sensor configured inside the inverter (20) that detects when a voltage sensor detecting voltage fails and controls the system to continuously perform MPPT even if the voltage sensor fails. It is made including a control unit (200).

[0046] Here, the arc and spark detection unit (100) comprises an image capturing unit (110) positioned to capture the inside of the connection board (20) at close range, a sensor unit (120) that detects current and sound inside the connection board (20), a frame buffer (130) that receives image data captured by the image capturing unit (110) and stores it in frames for a predetermined time interval, an area of ​​interest generation unit (140) that generates an area of ​​interest (ROI) corresponding to an arc / spark candidate based on short-term luminescence or rapid change in brightness among the images stored in the frame buffer (130), an enhanced detection unit (150) that divides the area of ​​interest generated by the area of ​​interest generation unit (140) into a plurality of sub-areas, performs bounding box-based object detection in each sub-area, merges the detection results for each sub-area, and enhances the detection of arcs / sparks as small objects in the area of ​​interest, and a current score based on high frequency features from the current signal of the sensor unit (120) and / or arc-related features from the sound signal It comprises a feature quantity calculation unit (160) that calculates a sound score based on the image, a confirmation determination unit (170) that determines whether an arc / spark is confirmed by fusing at least one of the image-based value calculated from the reinforcement detection unit (150) and the current score and / or sound score calculated from the feature quantity calculation unit (160), and a control signal generation unit (180) that generates and outputs a control signal for alarm, output restriction, or blocking according to the confirmation determination result of the confirmation determination unit (170).

[0047] The above image capturing unit (110) obtains image information by performing close-range imaging of terminals, fuses, and circuit breaker areas inside the connection board (20), and the sensor unit (120) detects a current signal to calculate high-frequency feature quantity (band power / spike / kurtosis, etc.) and / or detects an acoustic signal to calculate acoustic feature quantity (MFCC / mel spectrum / high-frequency energy, etc.) respectively.

[0048] The above frame buffer (130) preserves frames before and after the event to secure evidence of a short period of time.

[0049] Additionally, the region of interest generation unit (140) derives a region of interest (ROI) for spark candidates from the entire frame using the inter-frame luminance difference and high-frequency components, and the enhancement detection unit (150) divides the region of interest (ROI) into multiple sub-regions (tiles / patches) to perform object detection and merges the results.

[0050] In the above-mentioned reinforcement detection unit (150), sub-regions are divided so as to overlap each other, and the merging includes at least one of non-maximum suppression (NMS), weighted merging, or stochastic merging.

[0051] The above-mentioned enhanced detection unit (150) adaptively determines the sub-area size and / or overlap ratio and / or detection threshold according to at least one of reflectance, flicker index, noise index or blur index.

[0052] The frame buffer (130) provides an event interval including frames before and after the candidate detection point, and the reinforcement detection unit (150) selectively performs reinforcement detection on the event interval.

[0053] The above feature quantity calculation unit (160) calculates at least one of band power, spike frequency, kurtosis, or spectral flatness by frequency band from a current signal, and calculates at least one of Mel spectrogram, MFCC, high-frequency energy, or click / burst event frequency from an acoustic signal.

[0054] The above confirmation determination unit (170) includes confirmation gate logic that determines arc / spark confirmation when the image-based spark score is greater than or equal to a first threshold and at least one of the current-based score or acoustic-based score is greater than or equal to a second threshold, and determines confirmation only when the confirmation gate condition persists for a predetermined time or longer, or applies hysteresis to the confirmation determination.

[0055] The control signal generation unit (180) includes a NORMAL / WATCH / WARNING / DANGER state machine, generates an output limit signal in the WARNING state, generates a blocking signal or a blocking recommendation signal in the DANGER state, is configured so that a separate microcontroller performs final blocking for safety interlock, and provides the output limit signal and / or blocking recommendation signal to the microcontroller.

[0056] The above reinforcement detection unit (150) configures a priority queue based on risk scores for multiple regions of interest, performs reinforcement detection on selected regions of interest to satisfy frame delay limits or computation time budgets, and stores or transmits an evidence package including frames before and after the event, reinforcement detection results (bounding boxes), and current / acoustic features or scores when an event occurs.

[0057] The above image module is configured to control exposure, shutter, or gain according to internal lighting conditions of the photovoltaic power generation device, or to emphasize short-duration light-emitting sparks including flicker-suppressing lighting.

[0058] A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that the confirmation judgment unit (170) adjusts a confirmation threshold or fusion weight based on a reliability index of image input or sensor input, and limits DANGER judgment when reliability is below a standard.

[0059] The above-mentioned reinforcement detection unit (150) further performs object detection for smoke or flame candidates, and the above-mentioned confirmation determination unit (170) further fuses temperature characteristic quantities and / or smoke / gas characteristic quantities to determine fire confirmation.

[0060] The above-mentioned region of interest generation unit (140) or enhancement detection unit (150) performs flicker removal or background modeling to suppress false positives caused by reflection, flicker, or dust / smoke.

[0061] It may further include a predictive maintenance analysis unit that accumulates and stores time-series data of the image-based spark score, current score, or acoustic score, and uses an artificial intelligence model to quantify and calculate a health index or failure probability indicating the deterioration state inside the connection box.

[0062] The above confirmation judgment unit (170) learns the correlation between the frequency and intensity of detected arc / spark occurrences and power generation data, predicts the decrease in power generation efficiency at a specific point in time in the future, and generates this as numerical data.

[0063] In the present invention, “division-merge based small object enhanced detection” means a process of dividing a region of interest (ROI) into multiple sub-regions (overlapping between sub-regions is possible), performing object detection for each sub-region (bounding box-based object detection in this embodiment), and merging the sub-region detection results to improve the detection performance of small objects, such as sparks, in the entire region of interest (ROI).

[0064] sign meaning note I_t(x,y) Frame (luminance / gray) at time t Fixed camera D_t(x,y) Frame difference (positive luminescence component) Spark Candidate R_k k-th ROI (Region of Interest) Candidate area T_{k,m} m-th sub-area (tile / patch) within ROI Divided area s_det Object detector (bounding box) score Maximum or average S_V Video-based Spark score (0~1) Sigmoid S_I Current HF-based score (0~1) Band power S_A Acoustic-based score (0~1) MFCC / Mel etc. ConfirmArc Arc / Spark Confirmation (Boolean) Gate

[0065] Here, since the spark may appear as a short inter-frame emission, a candidate region of interest (ROI) is generated using frame difference. At this time, the generation of the candidate region of interest is as shown in Equation 1 below.

[0066]

[0067] Here, tau_D is the candidate mask threshold, and m is the ROI expansion margin. The region of interest generation unit (140) can suppress false positives for reflections / flickers in the first step by using area / shape / persistence conditions (e.g., A_min <= A <= A_max, occurring within 1~L frames).

[0068] In addition, an object detector is applied for the detection of small objects based on the above-mentioned segmentation-merging method. The object detector is configured to output a bounding box, and to avoid missing small sparks, the ROI is divided into multiple sub-regions to perform high-resolution detection and then merged. At this time, the ROI segmentation and merging are performed as shown in Equation 2 below.

[0069]

[0070] In this embodiment, the object detector outputs a spark candidate box and a score (s_det) in each sub-region, and calculates a representative score (s_det,max or s_det,avg) at the ROI level after merging.

[0071] In addition, the image-based spark score s_v includes false positive suppression, and the image score is calculated by combining "brightness (luminescence)", "small area", "short duration", and "object detector score". This can be expressed by Equation 3.

[0072]

[0073] Here, a1~a4 are weights, and E_ref / A_ref are normalization constants. The U / L term is used to deduct points for continuous emission (lighting / flicker).

[0074] In addition, the current high-frequency based score S_I is calculated by determining the bandpower or spike / kurtosis of the high-frequency band from the current signal i(t). This is represented by Equation 4 as follows. That is, Equation 4 is the current HF score.

[0075]

[0076] Here, mu_I and sigma_I are reference statistics in the normal range (site or equipment-specific calibration), and K_tilde is a kurtosis-based auxiliary term of the bandpass current (optional).

[0077] In addition, the acoustic-based score S_A is configured to generate a Mel spectrogram or MFCC from the acoustic / ultrasonic signal a(t), and for a lightweight classifier to output an arc probability. This can be expressed by Equation 5 as follows.

[0078]

[0079] In addition, the confirmation judgment unit (170) fuses the image-based spark score and the current / acoustic score to calculate the final arc score and the confirmation gate. This can be expressed by Equation 6 as follows. That is, Equation 6 is the fusion, confirmation gate, and time duration condition.

[0080]

[0081] Here, (theta_V, theta_I, theta_A) is the threshold value, and N is the number of samples corresponding to W seconds. K is the minimum number of confirmation occurrences within W seconds and is used to suppress one-off noise.

[0082] In addition, the control signal generator, which generates and outputs state machines and control signals, performs alarms / output restrictions / blocking (or blocking recommendations) according to the state machine.

[0083] Meanwhile, WATCH is when S_V or S_ARC exceeds the observation threshold (additional monitoring), WARNING is when ConfirmArc occurs intermittently or S_ARC persists for a certain period of time (output limit recommendation or notification), and DANGE is when ConfirmArc is confirmed (block recommendation / block) satisfying the time duration condition of formula group 6.

[0084] And ANGER can apply a manual return policy after latch for safety.

[0085] Meanwhile, the range of parameters is an exemplary range intended to enhance feasibility and can be adjusted according to the field environment.

[0086] parameters sign Recommended range (e.g.) note Tile (sub-area) size p 256~640 pixels Emphasize small objects, select based on embedded performance Overlap ratio o / p 0.10~0.30 Merge stability vs. computational cost trade-off ROI expansion margin m 5~30 pixels Prevention of detection omissions Buffer length W_buf 1~3 s Securing post-war frame evidence Candidate mask threshold tau_D Device calibration Settings based on flicker / noise Image threshold theta_V 0.65~0.85 Upward trend based on environmental reliability Current / Acoustic Threshold theta_I / theta_A 0.60~0.90 Adjust according to sensor quality Confirmation window W 2~10 s Suitable for short arc repetition detection Minimum number of verifications K 2~5 Severe noise suppression

[0087] The sensor fault and control unit (200) comprises a voltage and current measuring unit (210) for measuring voltage and current, a fault threshold setting unit (220) for setting a threshold value to determine whether there is a fault in the voltage and current measuring unit (210), an MPPT execution unit (230) for receiving the voltage and current measured through the voltage and current measuring unit (210), calculating power, and performing MPPT, a control unit (240) for receiving the MPPT result performed through the MPPT execution unit (230), compensating for the difference between the target voltage and the voltage measured through the voltage sensor, and generating a switching control signal through PI control to control the boost converter, a voltage calculation unit (250) for receiving the switching control signal through the control unit (240), converting the reference of the PI control into current and automatically calculating the voltage data being measured when a fault occurs in the voltage sensor, and receiving the voltage value calculated through the voltage calculation unit (250), of the inductor constituting the boost converter It comprises a calculation value determination unit (260) that determines whether to operate in CCM or DCM depending on the current and simultaneously calculates the voltage and current in the CCM and DCM; a sensor fault determination unit (270) that receives the voltage value determined through the calculation value determination unit (260) and determines whether the voltage sensor is faulty by comparing it with a threshold value set in the fault threshold setting unit (220); and a control target conversion unit (280) that, if a fault is determined based on the result of the sensor fault determination unit (270), changes to current PI control and calculates the current error used by subtracting the measured current from the target current to perform PI control and continues to perform MPPT.

[0088] Here, the fault boundary setting unit (220) is a part that sets reference values ​​for voltage and current, and based on these reference values, the sensor fault determination unit (270) can determine whether the sensor is faulty.

[0089] Meanwhile, the reference value set in the fault boundary setting unit (220) is determined using the absolute value of the voltage measured through the voltage and current detection unit (210) in the sensor fault determination unit (270) and the voltage calculated by the voltage calculation unit (250).

[0090] That is, the above absolute value is a preset voltage fault boundary value. If it is greater than or equal to, it can be determined that the voltage sensor is faulty. Accordingly, the above fault boundary value must be set first, and it can be determined as shown in Equation 1 below using the efficiency of the boost converter among solar power inverters.

[0091]

[0092] Here, the above mathematical formula 7 is the fault boundary value It is a formula for setting, represents the maximum voltage of the PV arrays connected to the input terminal of the solar power inverter. represents the minimum efficiency of the boost converter used in the solar power inverter. That is, sensor failure can be determined by setting a fault boundary value for the voltage sensor in advance using the above mathematical formula 7.

[0093] The above voltage and current measuring unit (210) measures voltage and current to calculate power and is used as an element for performing MPPT. Since the boost converter performs MPPT according to voltage and current, it continuously measures voltage and current.

[0094] The above MPPT execution unit (230) is performed in a boost converter and must control the solar panel to a voltage at which it can always supply maximum power, and this is called Maximum Power Point Tracking (MPPT). Meanwhile, the above MPPT generally uses change values ​​of voltage, current, and power, measures voltage and current, and continuously tracks the voltage point of the maximum power point.

[0095] The above control unit (240) amplifies the input voltage and outputs it through an appropriate On / Off cycle of a power semiconductor device configured in a boost converter. The MOSFET, which is the power semiconductor device used here, uses an On / Off signal created by a control amount.

[0096] In other words, an appropriate amount of control is required to match the target value, and generally, it is responsible for performing control to make the error of the difference between the target voltage and the voltage measured by the sensor 0[V].

[0097] The above voltage calculation unit (250) is necessary to calculate the voltage even without a voltage sensor in order to use the sensor failure detection and continuous control method of the solar power generation inverter, and thus it is necessary to analyze the operation of the boost converter separately.

[0098] The above-mentioned calculation value determination unit (260) receives the voltage value calculated through the above-mentioned voltage calculation unit (250) and calculates and determines the voltage and current in the above-mentioned CCM and DCM in order to determine whether to operate in CCM or DCM depending on the current of the inductor (L1) constituting the boost converter.

[0099] The sensor fault determination unit (270) receives a set calculation value from the calculation value determination unit (260) and compares it with a fault boundary value set in the fault boundary setting unit (220) to determine whether the sensor is faulty.

[0100] The above control target conversion unit (280) calculates the voltage when it is determined through the sensor fault determination unit (270) that a sensor fault has occurred, thereby enabling the inverter to continue MPPT operation.

[0101] Meanwhile, the PV voltage analysis in the inverter's CCM mode is explained as follows.

[0102] FIG. 3 is a circuit diagram showing a boost converter in a photovoltaic power generation system according to the present invention.

[0103] As shown in FIG. 5, a boost converter (31) is provided between the output terminal of the PV array (10) and the input terminal of the inverter (30) to stabilize the irregular DC power (Vin) output from the PV array (10) and boost it to an appropriate size and output it to the inverter (30).

[0104] Here, the boost converter (31) operates to receive a PWM control signal from the outside and boost the voltage to a set voltage of a certain magnitude. More specifically, it is configured to include a first switching element (Q1), an inductor (L1), and a diode (D1). When the first switching element (Q1) is turned ON according to the PWM control signal, the power transmitted from the PV array (10) is stored in the inductor (L1). When the first switching element (Q1) is turned OFF, the power stored in the inductor (L1) is applied to a capacitor (Cd) configured at the output terminal of the boost converter (300).

[0105] Advantages of using the above boost converter (31) include the ability to maintain a constant DC Link voltage according to variations in sunlight, a wide input range, and ease of inverter design and control when the DC Link voltage is maintained at a constant level. Since the above boost converter (300) can convert low voltage to high voltage, power generation is possible even in the morning and evening when solar radiation is low.

[0106] The above PV array (10) generates a high voltage through MPPT control even when solar irradiance is low. When the power generation is low, the DC Link is maintained at the grid voltage, and in this case, the inverter (30) is stopped. When solar irradiance is low in the morning and evening, if the output current control is misaligned due to sensor errors and control performance, the solar cell voltage may decrease rapidly.

[0107] At this time, the error of the above current sensor is usually 1%, but at low currents, circuit offset, magnetic offset, temperature deviation + product deviation accumulate, and the error at low currents increases.

[0108] Meanwhile, if the voltage of the PV array (10) becomes lower than the grid voltage, a large current is induced from the grid to the DC Link. At this time, depending on the configuration of the inverter (30), if there is no converter, the induced current is induced into the PV array (10), causing protection operation and failure.

[0109] The present invention relates to an efficient method of boosting voltage only when necessary according to the voltage range of the PV array (10) at different times, rather than boosting voltage by connecting a boost converter (31) to the input terminal of the inverter (30).

[0110] The present invention monitors the solar inverter during periods of low solar irradiance and low power generation to determine whether to operate the boost converter (31), and if necessary, operates only with its original function to provide efficient control, thereby actively operating according to the power generation amount (PV voltage) and time of day of the solar inverter.

[0111] That is, the boost converter operates only when the power output is low in the section with low solar irradiance, and when the power output exceeds a certain amount or the time period is reached, the DC Switch is reconnected and the boost converter (31) stops operating and operates in its original function, thereby increasing the power generation amount in all output sections. Therefore, the solar power generation system of the present invention increases the power generation amount compared to the existing system in all sections except when power generation is completely stopped.

[0112] Here, Vpv is the DC input voltage of the PV array, PWM is an external control signal that turns the first switching element (Q1) ON or OFF, L1 is the capacitance of the inductor, V Link is the voltage of the inverter's DC Link capacitor, and Ipv is the measured value of the current sensor.

[0113] The amount of power that the boost converter (31) must process changes in real time depending on the amount of solar radiation received by the PV array (10). In the case of the boost converter (31), the amount of current flowing through the inductor (L1) varies depending on its capacity.

[0114] If a continuous current of 0[A] or more flows through the inductor (L1), it is classified as Continuous Conduction Mode (CCM), and if the current of the charged inductor (L1) is completely discharged to 0[A], it is classified as Discontinuous Conduction Mode (DCM).

[0115] In the present invention, all analyses in CCM and DCM are performed to analyze the boost converter (31) of the inverter (30) under all conditions. Among these, the characteristics of the boost converter (31) in CCM are as follows in Equations 8 and 9.

[0116]

[0117]

[0118] When the boost converter (31) is operating in CCM mode, the current flowing through the inductor (L1) is always higher than 0[A]. Here, when the first switching element (Q1) is operating, it has the characteristics of Equation 8 above, and when the first switching element (Q1) is open, it has the characteristics of Equation 9.

[0119] Here, This refers to the current flowing through the inductor (L1) when the boost converter (31) is operating in CCM mode and the first switching element (Q1) is turned on. represents the switching period and is also the reciprocal of the switching frequency.

[0120] also, This refers to the current flowing through the inductor (L) when the boost converter (31) is operating in CCM mode and the first switching element (Q1) is turned off.

[0121] In addition, when the above boost converter (31) operates as a CCM, it is as shown in the following mathematical formula 10.

[0122]

[0123] Meanwhile, substitute mathematical equations 8 and 9 into mathematical equation 10 and V PV When rearranged for this, it is as shown in mathematical equation 11.

[0124]

[0125] The above mathematical equation 11 represents the voltage input / output relationship in CCM mode at the boost converter (31). Here, V PV represents the input voltage of the solar panel, and V Link represents the voltage of the inverter's DC Link capacitor. D represents the duty ratio, which is the rate at which power components are turned on per cycle.

[0126] If we look at mathematical equation 11 above, V LinK , if you know D V PVThe voltage can be determined mathematically. That is, due to the characteristics of the boost converter (31) and the PI controller, V LINK Since D can be known in real time, the input voltage of the boost converter (31) can be measured using mathematical formula 5.

[0127] A simulation was performed to verify the effect of PV voltage analysis in the above CCM mode. The circuit diagram used in the simulation is as shown in Figure 3, and the parameters of the components used were set as shown in Table 3 below.

[0128]

[0129] Here, Table 3 above shows the simulation parameters for verifying PV voltage analysis in CCM mode.

[0130] Figure 6 is a graph showing the results of a simulation for verifying PV voltage analysis in CCM mode, showing the results of a simulation performed to verify PV voltage analysis in CCM mode.

[0131] Here, the simulation was performed for 1[s], and the results at 0.7[s] were enlarged and analyzed.

[0132] V, the voltage measured after 0.7[s] PV It was measured as 450[V]. The voltage calculated by Equation 5 is It was calculated as 450[V].

[0133] That is, using mathematical equation 11 and the voltage sensor, V PV Even without directly measuring V PV It was verified that it can be calculated.

[0134] Meanwhile, the PV voltage in the DCM mode of the photovoltaic power generation system according to the present invention is analyzed as follows.

[0135] At this time, unlike CCM, the above DCM has the characteristic that the L current of the charged inductor (L1) is completely discharged to 0[A]. Accordingly, the voltage must be interpreted differently from CCM. Among these, the characteristics of DCM in the boost converter (300) are as shown in Equations 12 and 13 below.

[0136]

[0137]

[0138] When the boost converter shown in Fig. 5 is operating in DCM mode, the current charged in the inductor (L1) is always discharged to 0[A]. Here, when the MOSFET is operating, there is a characteristic as shown in Equation 12, and when the MOSFET is open, there is a characteristic as shown in Equation 13 only in the section where the current in the inductor (L1) is discharged to 0[A].

[0139] Here, substituting Equation 13 into Equation 12 and rearranging for D1 yields Equation 14.

[0140]

[0141] Here, the maximum current I of L flowing through the inductor (L1) peak Rewriting it gives the same as mathematical formula 15.

[0142]

[0143] Meanwhile, in the above DCM, the current charged in the inductor (L1) during D is completely discharged during D1. There exists a relationship in mathematical equation 16 as follows, which represents this amount of current as the average over one period T.

[0144]

[0145] Substitute the above mathematical formulas 14 and 15 into mathematical formula 16 and V PV When rearranged for this, it is as shown in mathematical equation 17.

[0146]

[0147] The above mathematical formula 17 represents a voltage calculation method in the DCM mode of the boost converter (300). Here, V PV represents the input voltage of the solar panel, and V LinK represents the voltage of the inverter's DC Link capacitor. D represents the duty ratio, which is the rate at which power components are turned on per cycle.

[0148] If we look at mathematical formula 17 as above, I PV , V LinK If you know , D, and T, then V PV The voltage can be determined mathematically. Due to the characteristics of the boost converter and PI controller, I PV , V LinK Since D and T can be known in real time, the input voltage in the DCM of the boost converter (300) can be measured using mathematical formula 17.

[0149] Meanwhile, a simulation was performed to verify the effect of PV voltage analysis in the above DCM mode. The circuit diagram used in the simulation is as shown in Fig. 5, and the parameters of the components used were set as shown in Table 4 below.

[0150]

[0151] Figure 7 is a graph showing the simulation results for verifying PV voltage analysis in DCM mode.

[0152] That is, Figure 7 shows the results of a simulation performed to verify the PV voltage analysis in DCM mode. The simulation was performed for 1 [s], and the results at 0.7 [s] were magnified and analyzed.

[0153] V, the voltage measured after 0.7[s] PV It was measured as 450[V]. The voltage calculated by Equation 17 is It was calculated as 450.1[V].

[0154] That is, using mathematical formula 17, V PV Even without directly measuring VPV It was verified that it can be calculated.

[0155] The above calculation value determining unit (160) changes the amount of power that the boost converter (31) must process in real time according to the amount of solar radiation received by the PV array (10). In the case of the boost converter (31), the amount of current (L) flowing through the inductor (L) varies according to its capacity. If a continuous current of 0[A] or more flows through the inductor (L1), it operates in CCM mode, and if the charged current of the inductor (L1) is completely discharged to 0[A], it operates in DCM mode.

[0156] Therefore, the voltage and current in CCM and DCM must be calculated simultaneously and used appropriately. The results of the above calculations can be summarized as shown in Table 3 below.

[0157] In addition, the method of simultaneously using CCM and DCM analysis results in the boost converter of the photovoltaic power generation system according to the present invention is explained as follows.

[0158] The amount of power that the boost converter (31) must process changes in real time depending on the amount of solar radiation received by the PV array (10). In the case of the boost converter (31), the amount of current (L) flowing through the inductor (L1) varies according to its capacity. If the current (L) flowing through the inductor (L1) is a continuous current of 0[A] or more, it operates in CCM mode, and if the current (L) charged in the inductor (L1) is completely discharged to 0[A], it operates in DCM mode.

[0159] Accordingly, the voltage and current in CCM and DCM must be calculated simultaneously and used appropriately. The results of the above calculations are summarized in Table 5.

[0160]

[0161] Meanwhile, when operating in CCM mode, the result of Equation 17 cannot be used. Conversely, when operating in DCM mode, the result of Equation 11 cannot be used. However, due to the characteristics of the PV panel, CCM and DCM continuously change depending on the amount of solar irradiance. Accordingly, a method is needed to incorporate Equations 11 and 17 by reflecting the continuous changes in CCM and DCM.

[0162] Figure 8 is a diagram showing the waveform of the change in calculated voltage according to the change in input voltage, and Figure 9 is a diagram showing the waveform of the change in calculated voltage according to the enlarged change in input voltage.

[0163] That is, Fig. 8 shows the waveform of the change in the calculated voltage according to the change in the input voltage. The measured voltage V PV It was set to rise steadily from 550[V] to 650[V]. In this situation, the CCM voltage calculated by Equation 5 is and the DCM voltage calculated by Equation 17 is It is expressed as such. Looking at Fig. 8, it can be seen that the error of the voltage values ​​calculated between 2[s] and 3[s] changes.

[0164] In addition, Figure 9 shows the waveform of the change in calculated voltage according to the expanded input voltage change. Based on approximately 2.4 s, the state before is DCM, the state after 2.4 s is BCM which is the boundary current mode, and the state thereafter operates as CCM.

[0165] When it is DCM is normally V PV calculated. However It has an error and It has a larger value.

[0166] When it is CCM is normally V PV calculated. However It has an error and It has a larger value.

[0167] That is, V with BCM as the boundary PV This means that the appropriate value for is always a small value.

[0168] Accordingly, V considering DCM and CCM simultaneously through the condition as in mathematical equation 18 below PV can calculate.

[0169]

[0170] In order for the sensor fault detection unit (170) to calculate the voltage by replacing the voltage sensor and use it to perform continuous MPPT, a method is required to detect a fault when a problem occurs in the voltage sensor.

[0171] When the algorithm starts, it establishes a fault boundary to determine the fault. Subsequently, it continuously measures the voltage and current at the input of the PV panel. PI control is continuously performed, which allows D to be determined and V PV , I PV , V Link V using , D, T PV , I PV Calculate each. A voltage sensor failure can be determined by the following mathematical formula 19.

[0172]

[0173] As in mathematical equation 19 above, V is the fault boundary value. Fault It uses the absolute value of the measured voltage and the calculated voltage. This value is V, the preset voltage fault boundary value. Fault If it is greater than or equal to, it can be determined that the voltage sensor is faulty. Here, V Fault It can be determined as follows using the efficiency of the boost converter among solar power inverters.

[0174] The above control target conversion unit (180) cannot perform voltage PI control if a failure of the voltage sensor is detected. The error used for voltage PI control is given by the following mathematical formula 20.

[0175]

[0176] Voltage error used in voltage PI control is the target voltage The measured voltage at It is calculated by subtracting. However, due to a voltage sensor failure Since it cannot be used, it must be changed to the control of mathematical formula 20 as follows.

[0177]

[0178] Meanwhile, if the above voltage sensor fails, it must be switched to current PI control, and the current error used is the target current The measured current at Calculate by subtracting and perform PI control.

[0179]

[0180] Table 6 shows the parameters of the components required for the simulation performed to verify the MPPT fault persistence control scheme. The circuit diagram is identical to that in Fig. 5, and the initial V PV The voltage is 450[V], and the voltage V capable of producing maximum power. MPP It was set to 400[V].

[0181] Figure 10 is a graph showing the MPPT simulation results when using a voltage sensor, and Figure 11 is a graph showing an enlarged portion of Figure 10.

[0182] As shown in Fig. 10, starting at 450[V], V MPP It can be seen that MPPT is being performed normally up to 400[V]. If we look at the enlarged view in Fig. 11... It can be seen that MPPT is being performed normally at 400[V].

[0183] Figure 12 is a graph showing the MPPT simulation results when MPPT continuous control is performed due to a voltage sensor failure, and Figure 13 is a graph showing an enlarged portion of Figure 12.

[0184] As shown in Fig. 12, starting at 450[V] It can be seen that MPPT is being performed normally up to 400[V]. If we look at the enlarged view in Fig. 11... It can be seen that MPPT is being performed normally at 400[V].

[0185] In other words, it was verified that a fault persistence control method capable of detecting voltage sensor failure and performing MPPT even in the absence of a voltage sensor operates normally. When this invention is applied to a photovoltaic power generation system, it can eliminate economic losses incurred during inverter maintenance or replacement.

[0186] A conventional inverter cannot perform MPPT if a single sensor fails, so the inverter must be shut down. Consequently, power cannot be produced until the sensor or the inverter itself is repaired or replaced, resulting in economic losses for the seller of the generated electricity.

[0187] Accordingly, the present invention provides a method for detecting sensor failure and a failure persistence control method capable of performing MPPT even when one sensor is missing.

[0188] Mathematical formulas 7 through 21 represent formulas that allow sensor data to be calculated using known variables even if the sensor fails. By appropriately utilizing these formulas, a fault persistence control method capable of performing MPPT even if the sensor fails is possible.

[0189] FIG. 14 is a flowchart illustrating a continuous control method using a sensor failure detection device of a photovoltaic power generation system according to the present invention.

[0190] As illustrated in FIG. 14, a continuous control method using a sensor fault detection device of a photovoltaic power generation system according to the present invention sets an arbitrary reference value that becomes the fault boundary of the sensor using a fault boundary setting unit (120) (S110).

[0191] Next, voltage and current are measured using a voltage and current measuring unit (110) installed at the input terminal of the PV array, and power is calculated using the measured voltage and current (S120).

[0192] Next, the power calculated by the MPPT execution unit (130) is received, a voltage reference suitable for MPPT is determined, and voltage PI control is performed in the control unit (140) (S140). At this time, when voltage PI control is performed through the control unit (140), the resulting control amount can be known, and the duty ratio, which is the On / Off ratio of the switching element, can be known.

[0193] Next, voltage and current are calculated using various variables (S150), and fault conditions are compared to determine the fault (S160).

[0194] If the above sensor is not faulty, MPPT is performed (S170), and if the above voltage sensor is faulty, the reference of the PI control is converted to current and the data of the voltage being measured is changed to the calculated voltage (S180).

[0195] Then, using the above-mentioned calculated voltage and measured current, power is calculated and MPPT is continuously performed (S190).

[0196] If the flowchart of the fault-continuous MPPT control method described above is applied to a solar power inverter, continuous control is possible even if a single sensor required for a single MPPT fails, thereby eliminating economic losses that occur while the inverter is being maintained or replaced.

[0197] Meanwhile, the embodiments of the present invention described above are merely illustrative, and those skilled in the art to which the present invention pertains will readily understand that various modifications and equivalent alternative embodiments are possible therefrom.

[0198] Therefore, it can be well understood that the present invention is not limited only to the forms mentioned in the detailed description above.

[0199] Accordingly, the true scope of technical protection of the present invention should be determined by the technical concept of the appended claims, and the present invention should be understood to include all variations, equivalents, and substitutions within the scope of the technical concept defined by the appended claims.

[0200] In addition, the method for controlling the boost converter of a solar inverter according to the present invention can be implemented as computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes all types of recording devices in which data that can be read by a computer system is stored. Examples of recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, hard disk, flash drive, etc., and also include implementation in the form of a carrier wave (e.g., transmission via the Internet). Furthermore, the computer-readable recording medium may be distributed across networked computer systems, and computer-readable code may be stored and executed in a distributed manner. Explanation of the symbols

[0201] 100: Arc and spark detection unit 110: Image capturing unit 120: Sensor section 130: Frame buffer 140: Region of Interest Generation Unit 150: Enhanced Detection Unit 160: Feature Calculation Unit 170: Confirmation Decision Unit 180: Control signal generation unit 200: Sensor failure and control unit

Claims

Claim 1 delete Claim 2 The system comprises a plurality of strings configured by connecting multiple solar modules that generate electricity using solar power in series or parallel, a junction box that collects DC power produced from each string, an inverter that receives DC power transmitted from the junction box, tracks MPPT using a boost converter equipped with voltage and voltage sensors for measuring voltage and current, converts it into AC power, and outputs it, and a control server that receives and monitors fire and power generation information through communication with the junction box and the inverter, an arc and spark detection unit configured inside the junction box that detects arcs or sparks by combining image analysis and current and acoustic sensor data, and a sensor failure and control unit configured inside the inverter that detects a failure of a voltage sensor detecting voltage and controls the system to continuously perform MPPT even if the voltage sensor fails, wherein the arc and spark detection unit comprises an image capturing unit positioned to photograph the interior of the junction box at close range, a sensor unit that detects current and acoustics inside the junction box, a frame buffer that receives image data captured by the image capturing unit and stores it in frame units for a predetermined time interval, and among the images stored in the frame buffer, a short-time light emission or an interest region generation unit that generates an interest region corresponding to an arc / spark candidate based on a sudden change in brightness; an enhanced detection unit that divides the interest region generated from the interest region generation unit into a plurality of sub-regions, performs bounding box-based object detection in each sub-region, and merges the detection results for each sub-region to enhance detection of arcs / sparks as small objects in the interest region; and a feature calculation unit that calculates a current score based on high frequency features from the current signal of the sensor unit and / or an acoustic score based on arc-related features from the acoustic signal.A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized by comprising: a confirmation determination unit that determines whether an arc / spark is confirmed by fusing at least one of an image-based value calculated from the reinforcement detection unit and a current score and / or acoustic score calculated from the feature quantity calculation unit; and a control signal generation unit that generates and outputs a control signal for alarm, output limiting, or cutoff according to the confirmation determination result of the confirmation determination unit. Claim 3 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 2, the region of interest generating unit detects light emission candidates based on inter-frame differences or temporal high-pass components in the entire frame, and the enhancement detection unit operates only for the region of interest where candidates exist. Claim 4 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 2, the sub-regions are divided to overlap each other, and the merging includes at least one of non-maximum suppression (NMS), weighted merging, or stochastic merging. Claim 5 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 2, the reinforced detection unit adaptively determines the sub-area size and / or overlap ratio and / or detection threshold according to at least one of reflectance, flicker index, noise index or blur index. Claim 6 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 2, the frame buffer provides an event interval including frames before and after a candidate detection point, and the reinforcement detection unit selectively performs reinforcement detection on the event interval. Claim 7 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 2, the feature quantity calculation unit calculates at least one of band power, spike frequency, kurtosis, or spectral flatness by frequency band from a current signal. Claim 8 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 2, the feature quantity calculation unit calculates at least one of a Mel spectrogram, MFCC, high-frequency energy, or click / burst event frequency from an acoustic signal. Claim 9 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 2, the confirmation determination unit includes confirmation gate logic that determines arc / spark confirmation when the image-based spark score is greater than or equal to a first threshold and at least one of the current-based score or the acoustic-based score is greater than or equal to a second threshold. Claim 10 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 9, the confirmation gate condition is determined as confirmation only when it persists for a predetermined time or longer, or hysteresis is applied to the confirmation determination. Claim 11 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 2, the control signal generating unit includes a NORMAL / WATCH / WARNING / DANGER state machine, generates an output limit signal in the WARNING state, and generates a cutoff signal or cutoff recommendation signal in the DANGER state. Claim 12 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 2, the control signal generating unit is configured such that a separate microcontroller performs final blocking for safety interlocking and provides an output limiting signal and / or a blocking recommendation signal to the microcontroller. Claim 13 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, wherein, in claim 2, the reinforcement detection unit forms a priority queue according to risk scores for multiple regions of interest and performs reinforcement detection on the selected regions of interest to satisfy a frame delay limit or computation time budget. Claim 14 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 2, when an event occurs, it stores or transmits an evidence package including frames before and after the event, reinforced detection results (bounding boxes), and current / acoustic feature quantities or scores. Claim 15 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 2, the image capturing unit is configured to control exposure, shutter, or gain according to internal lighting conditions of the photovoltaic power generation device, or to emphasize a spark in the form of short-time light emission including flicker suppression lighting. Claim 16 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 2, the confirmation judgment unit adjusts a confirmation threshold or fusion weight based on a reliability index of an image input or sensor input, and limits DANGER judgment when the reliability is below a standard. Claim 17 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 2, the reinforcement detection unit additionally performs object detection for smoke or flame candidates, and the confirmation determination unit additionally fuses temperature feature quantities and / or smoke / gas feature quantities to determine fire confirmation. Claim 18 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 2, the region of interest generating unit or enhanced detection unit performs flicker removal or background modeling to suppress false detections caused by reflection, flicker, or dust / smoke. Claim 19 A photovoltaic power generation system capable of detecting junction box fires and inverter sensor failures, characterized in that, in claim 2, it further includes a predictive maintenance analysis unit that accumulates and stores time-series data of the image-based spark score, current score, or acoustic score, and quantifies and calculates a health index or failure probability indicating the deterioration state inside the junction box using an artificial intelligence model. Claim 20 A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized in that, in claim 19, the confirmation judgment unit learns the correlation between the frequency and intensity of detected arcs / sparks and power generation data, predicts a decrease in power generation efficiency at a specific point in time in the future, and generates this as numerical data. Claim 21 It comprises a plurality of strings configured by connecting multiple solar modules that generate electricity using solar power in series or parallel, a junction box that collects DC power produced from each string, an inverter that receives DC power transmitted from the junction box, tracks MPPT using a boost converter equipped with voltage and current sensors for voltage and current measurement, converts it into AC power for output, and a control server that receives and monitors fire and power generation information through communication with the junction box and the inverter, an arc and spark detection unit configured inside the junction box that detects arcs or sparks by utilizing a combination of image analysis and current and acoustic sensor data, and a sensor failure and control unit configured inside the inverter that detects a failure of a voltage sensor that detects voltage and controls the system to continuously perform MPPT even if the voltage sensor fails, wherein the sensor failure and control unit comprises a failure threshold setting unit that sets a threshold value to determine whether the current sensor and voltage sensor are faulty, an MPPT execution unit that performs MPPT according to the current and voltage measured through the current sensor and voltage sensor, and a target voltage and the voltage that receives the MPPT results performed through the MPPT execution unit A control unit that generates a switching control signal to control the boost converter by compensating for the voltage difference measured through a sensor; a voltage calculation unit that receives the control signal through the control unit and calculates the voltage by subtracting the measured voltage from the target voltage to reduce the voltage error by PI control; a calculation value determination unit that receives the voltage value calculated through the voltage calculation unit and simultaneously calculates the voltage and current in the CCM and DCM to determine whether to operate in CCM or DCM depending on the current of the inductor constituting the boost converter; and a sensor fault determination unit that receives the voltage value determined through the calculation value determination unit and determines whether the voltage sensor is faulty by comparing it with a threshold value set in the fault threshold setting unit.A photovoltaic power generation system capable of detecting connection box fires and inverter sensor failures, characterized by comprising a control target conversion unit that, when a fault is determined based on the result of the sensor failure determination unit, changes to current PI control and calculates the current error used by subtracting the measured current from the target current to perform PI control and continuously performs MPPT. Claim 22 In claim 21, the control target converter switches to current PI control in the event of a failure of the voltage sensor, and the current error is used in the following mathematical formula 1. is the target current The measured current at A photovoltaic power generation system capable of detecting junction box fires and inverter sensor failures, characterized by calculating by subtracting and performing PI control.[Mathematical Formula 1] Claim 23 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, wherein, in claim 21, the calculation value determining unit determines the voltage inside the inverter by reflecting Equations 2 and 3, reflecting the continuous change of the CCM and DCM without a voltage sensor.[Equation 2] [Mathematical Formula 3] (Here, L is the capacitance of the inductor, V Link is the voltage of the inverter's DC Link capacitor, Ipv is the measured value of the current sensor, D is the duty cycle which is the rate at which the power component is turned on per cycle, and T is time) Claim 24 A photovoltaic power generation system capable of detecting connection box fire and inverter sensor failure, characterized in that, in claim 21, the sensor failure determination unit uses the absolute values ​​of the measured voltage and the calculated voltage, and determines a voltage sensor failure when the absolute value is greater than or equal to a preset voltage failure threshold value.

Images