Method for retrofitting existing buildings with renewable energy

Abstract

This invention relates to the field of photovoltaic system control technology and discloses a method for retrofitting existing buildings using renewable energy. The method includes establishing communication between a central control unit and distributed matrix nodes and a lidar device using an industrial communication bus; using lidar to establish a digital spatial model, measuring the impedance characteristics of the series power supply network and establishing a hierarchical impedance topology matrix; combining a theoretical model to invert the transmittance attenuation coefficient to correct the predicted irradiance; and based on the effectively predicted irradiance and the hierarchical impedance topology matrix, solving for the optimal global current value that maximizes the system's net output power and driving the system to switch to the corresponding circuit topology state. This invention, through active mapping and parameter inversion, helps solve the problems of missing building drawings and unclear line impedance, achieving optimized system net power and safe operation under complex shading environments.

Description

Technical Field

[0001] This invention relates to the field of photovoltaic system control technology, specifically a method for retrofitting existing buildings with renewable energy. Background Technology

[0002] With the acceleration of urban renewal and the advancement of "dual carbon" goals, utilizing photovoltaic technology to carry out green renovations of existing buildings has become an important way to reduce building energy consumption. Unlike the photovoltaic integrated design of new buildings, the renovation of existing buildings faces more complex engineering environments and technical constraints.

[0003] At the physical level of existing buildings, due to their age, precise as-built drawings or digital models are often lacking, making it difficult to accurately obtain data on the actual dimensions, flatness, and orientation of the building facade. When arranging photovoltaic modules, manual on-site measurements are typically relied upon, which is inefficient and difficult to eliminate accumulated errors. Furthermore, electrical wiring in existing buildings is often constrained by existing shaft space and building structure, resulting in tortuous and significantly increased DC cable lengths. During long-distance DC transmission, the voltage drop loss caused by cable impedance is not negligible, and due to varying degrees of aging, the physical impedance characteristics exhibit a non-uniform distribution. However, most existing photovoltaic array control strategies assume ideal line impedance or only consider maximizing the output power of the modules themselves, ignoring power losses on the transmission network. This leads to a suboptimal net benefit from grid connection.

[0004] At the operational environment level, existing buildings are typically situated within complex urban landscapes, where surrounding buildings, trees, and their own attached structures create time-varying and irregular shadows. Simultaneously, cleaning and maintaining building facades is challenging, and photovoltaic modules are prone to uneven accumulation of dust and dirt. Existing maximum power point tracking (MPPT) technologies often struggle to distinguish between the effects of geometric shading and transmittance attenuation (dirt), leading to distorted power prediction models. This prediction bias makes it easy for the control system to get trapped in local extrema when searching for the optimal operating point among multi-peak power curves, or causes the system to frequently switch between different circuit topologies, resulting in power oscillations.

[0005] Furthermore, during control processes involving circuit topology reconfiguration, switching high-voltage DC systems under load can easily generate electric arcs, causing irreversible damage to electrical contacts and even posing a fire risk. Existing retrofit solutions often lack a coordinated control mechanism between the photovoltaic array side and the inverter side, failing to achieve efficient global power optimization while ensuring equipment safety. Therefore, there is an urgent need for a retrofit and control method that can proactively sense the existing building's spatial form and electrical characteristics, and comprehensively consider line losses and environmental factors. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for retrofitting existing buildings with renewable energy, which solves the problems faced by existing buildings in photovoltaic retrofitting, such as missing drawings, unclear line impedance, complex shading leading to low grid connection efficiency and large power prediction deviations.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A method for retrofitting existing buildings with renewable energy includes: arranging photovoltaic array assemblies and distributed matrix nodes on the existing building body; constructing a series power supply network connected to a grid-connected inverter using cables; establishing a communication connection between a central control unit and the distributed matrix nodes and a lidar device; using the lidar device to establish a digital spatial model, measuring the impedance characteristics of the series power supply network, and establishing a hierarchical impedance topology matrix; calculating the predicted irradiance based on the digital spatial model, acquiring actual electrical data, and inverting the transmittance attenuation coefficient using a theoretical model; using the transmittance attenuation coefficient to correct the predicted irradiance to obtain the effective predicted irradiance; generating candidate global current values ​​based on the effective predicted irradiance; calculating the voltage response of each distributed matrix node using the hierarchical impedance topology matrix; and solving for the optimal global current value that maximizes the net output power of the system.

[0009] Preferably, the method of retrofitting existing buildings with renewable energy sends control commands to distributed matrix nodes through a central control unit, driving the system to switch to a circuit topology state that matches the optimal global current value.

[0010] Preferably, the process of establishing a digital spatial model using the lidar device specifically includes: controlling the lidar device to perform multi-viewpoint 3D point cloud data acquisition, performing statistical outlier removal and voxelized mesh downsampling on the acquired point cloud data; subsequently, performing region growing segmentation based on the spatial smoothness and normal vector consistency of the preprocessed point cloud, identifying independent point cloud subsets corresponding to each photovoltaic module from the data; using the covariance matrix decomposition method to perform plane fitting on the point cloud subsets, extracting the geometric centroid of each photovoltaic module, and obtaining the macroscopic orientation vector of the existing building body, thereby correcting the normal vector of the fitted plane and constructing a digital spatial model.

[0011] Preferably, the process of establishing the graded impedance topology matrix includes: calling the geometric coordinates in the digital spatial model, estimating the physical length of the local DC cable using the weighted Manhattan distance model, and calculating the base resistance value; simultaneously, controlling all distributed matrix nodes to enter the series conduction state, controlling the grid-connected inverter to absorb a constant test current, and using the differential voltage measurement method to determine the measured impedance of the high-voltage DC trunk line between adjacent nodes; finally, acquiring the ambient temperature data, performing temperature drift correction on the base resistance value and the measured impedance, and generating the graded impedance topology matrix.

[0012] Preferably, the process of retrieving the transmittance attenuation coefficient includes: calculating the solar vector based on the current time and geographical coordinates, performing ray tracing using a digital spatial model, and calculating the geometric illuminance factor of each photovoltaic module; constructing a theoretical power generation benchmark model under clean conditions by combining the obtained ambient temperature and real-time total radiation intensity; controlling the distributed matrix nodes to enter bypass test mode to obtain the actual maximum power of the photovoltaic modules under their jurisdiction, calculating the ratio of the actual maximum power to the theoretical power generation benchmark model, and eliminating the influence of the geometric illuminance factor and temperature to obtain the transmittance attenuation coefficient characterizing the cleanliness of the module surface.

[0013] Furthermore, the process of retrieving the transmittance attenuation coefficient also includes performing spatial consistency verification and confidence update: based on the digital spatial model, the set of physically adjacent modules of the target photovoltaic module is identified, and the median value of the above ratio of all modules in the set is calculated; if the deviation between the ratio of the target photovoltaic module and the median value exceeds a preset anomaly threshold, the data is determined to be outlier noise and the update is paused; if the deviation does not exceed the threshold, a confidence weight factor positively correlated with the theoretical power level is introduced, and the transmittance attenuation coefficient is smoothly updated using a recursive filter.

[0014] Preferably, the photovoltaic array assembly includes several photovoltaic modules arranged on the building facade of the existing building, and the distributed matrix nodes are deployed in the electrical shaft;

[0015] The series power supply network includes local DC cables and high-voltage DC trunk lines;

[0016] The local DC cable connects the photovoltaic module to the corresponding distributed matrix node, and the high-voltage DC trunk line connects all the distributed matrix nodes in series.

[0017] Preferably, the process of finding the optimal global current value that maximizes the net output power of the system includes: generating a theoretical peak current set for all nodes in the network based on the effective predicted irradiance and the component temperature calculated from the ambient temperature, and using the current values ​​in this set as candidate global current values; for each candidate global current value, determining whether each distributed matrix node is bypassed due to the current exceeding the short-circuit current limit if forced to operate in series under this current value, and reconstructing the voltage response function of each node based on the determined bypass status; then summarizing the reconstructed voltage responses of all nodes, calculating and deducting the impedance voltage drop loss of the series power supply network based on the hierarchical impedance topology matrix, synthesizing a prediction curve of global power changing with current, and determining the current corresponding to the peak value of the prediction curve as the optimal global current value.

[0018] Preferably, the process of the central control unit sending control commands to the distributed matrix nodes specifically includes: calculating network transmission delay and processing time to determine a unified command effective time; calculating the difference between the current operating current and the optimal global current value; if the rate of change corresponding to the difference exceeds the preset physical ramping capability, truncating and smoothing the optimal global current value according to the maximum allowable ramping rate to generate a smoothed current command; if it does not exceed the limit, directly using the optimal global current value as the smoothed current command; encapsulating a control frame containing the command effective time and the smoothed current command and broadcasting it.

[0019] Preferably, the method of the present invention further includes performing a cold switching operation with the inverter: before performing the reconfiguration, calculating the estimated DC bus voltage difference before and after the reconfiguration; if the difference exceeds a preset cold switching trigger threshold, sending an active derating command to the grid-connected inverter to limit the input current to below the safe switching current threshold; after detecting that the bus current has decreased to the safe switching current threshold, triggering each distributed matrix node to perform circuit topology switching; after confirming that the physical switching is completed, controlling the grid-connected inverter to release the power limit and perform soft-start linear recovery.

[0020] Furthermore, as a preferred embodiment, the method further includes performing a sensor zero-point drift correction step: acquiring the total horizontal radiation intensity of the external environment; when the radiation intensity is lower than a preset sleep threshold and the duration exceeds a stable window, triggering a zero-point calibration command; in response to the command, controlling each distributed matrix node to perform multiple samplings of the port voltage and loop current and calculating the arithmetic mean to obtain the voltage zero-point drift value and the current zero-point drift value; during subsequent power generation operation of the system, subtracting the voltage zero-point drift value and the current zero-point drift value from the real-time collected actual electrical data.

[0021] Through the above technical solution, this invention can build a model based on active sensing by lidar, and combine the measured impedance matrix with the inverted transmittance coefficient to achieve refined modeling and control of existing building photovoltaic systems, thereby improving power generation revenue and system operation safety under complex shading environments.

[0022] This invention provides a method for retrofitting existing buildings using renewable energy. It offers the following advantages:

[0023] 1. This invention utilizes lidar to actively construct a digital spatial model and combines weighted Manhattan distance and differential voltage measurements to establish a hierarchical impedance topology matrix. This helps solve the problems commonly faced by existing buildings, such as missing as-built drawings and unclear routing of concealed wiring. It can quantify the impedance voltage drop caused by aging wiring or complex cabling, incorporate cable losses into global power optimization calculations, and ensure maximum net output power of the system under long-distance DC transmission conditions.

[0024] 2. This invention acquires actual electrical data and combines it with a theoretical model to invert the transmittance attenuation coefficient, using this coefficient to correct the predicted irradiance. This method can, without adding external distributed environmental sensors, perceive the degree of dust accumulation on the surface of photovoltaic modules and the complex shading conditions of building facades in real time, correcting theoretical prediction deviations caused by environmental factors, and improving the control accuracy and response speed of the system under non-uniform illumination and module aging conditions.

[0025] 3. This invention employs a global current optimization strategy based on full-network voltage response reconstruction, and coordinates with the inverter to perform cold switching and current ramp-up control. This mechanism not only avoids the power generation loss caused by traditional algorithms getting stuck in local optima, but also eliminates the arcing impact that may occur during load switching of the high-voltage DC system by actively reducing the bus current before topology reconstruction, extending the service life of power switching devices and ensuring the operational safety of existing building electrical systems. Attached Figure Description

[0026] Figure 1 This is the overall architecture of the existing building renewable energy retrofit system of the present invention;

[0027] Figure 2 This is a flowchart of the full-cycle operation control method of the present invention;

[0028] Figure 3 This is a diagram showing the global power variation characteristic with current of the present invention;

[0029] Figure 4 This is an electrical waveform diagram of the cold switching process of the present invention;

[0030] Figure 5 This is a comparison chart of daily power generation and line heat loss according to the present invention.

[0031] Among them, 100 is the photovoltaic array assembly; 101 is the photovoltaic module; 200 is the distributed matrix node; 301 is the local DC cable; 302 is the high-voltage DC trunk line; 400 is the central control unit; 500 is the grid-connected inverter; 600 is the lidar device; 700 is the industrial communication bus; 800 is the existing building body; 801 is the building facade; and 802 is the electrical shaft. Detailed Implementation

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] See attached document Figure 1This invention provides a method for retrofitting existing buildings using renewable energy. The method first involves constructing a hardware infrastructure based on a distributed cascade architecture within the physical environment of the existing building body 800. The method includes the following specific layout and connection steps.

[0034] A photovoltaic array assembly 100 is arranged on the building facade 801 of the existing building body 800. The photovoltaic array assembly 100 consists of several independent photovoltaic modules 101. The photovoltaic modules 101 adopt a passive structure design, and their back panels or junction boxes do not contain power optimizers, micro-inverters, or power shut-off devices. The photovoltaic modules 101 are anchored to the surface of the building facade 801 or embedded in the curtain wall structure by physical fasteners.

[0035] The electrical shafts 802 within the existing building 800 are used as the node deployment space. Distributed matrix nodes 200 are installed in the space of each floor or each predetermined interval of the electrical shaft 802. Each distributed matrix node 200 is an electrical control cabinet containing power switching devices. The number of distributed matrix nodes 200 is determined by the height of the existing building 800 and the number of partitions for the photovoltaic modules 101.

[0036] A local connection network is established. Photovoltaic modules 101 located on the same floor or within the same physical area are connected to the input of the corresponding distributed matrix node 200 in that area via local DC cables 301. The length of the local DC cable 301 is limited by the height range of the floor, enabling local access to the photovoltaic DC output.

[0037] A cascaded transmission trunk line is constructed. The outputs of all distributed matrix nodes 200 are connected in series end to end via a high-voltage DC trunk line 302. The high-voltage DC trunk line 302 is laid vertically along the electrical shaft 802, forming a single series high-voltage circuit that runs through the entire existing building body 800. The beginning and end of the high-voltage DC trunk line 302 extend to the equipment room of the existing building body 800.

[0038] A grid-connected inverter 500 and a central control unit 400 are installed in the equipment room. The two ends of the high-voltage DC trunk line 302 are connected to the DC input side of the grid-connected inverter 500. The AC output side of the grid-connected inverter 500 is connected to the building's power distribution network. The central control unit 400 establishes a data communication connection with each distributed matrix node 200 via an industrial communication bus 700. This connection is used by the central control unit 400 to send control commands to the distributed matrix nodes 200 and receive status data.

[0039] A lidar device 600 is installed on the top of the existing building 800 or in an area with a clear view. The signal output of the lidar device 600 is connected to the central control unit 400. The location of the lidar device 600 must meet the requirement of optically scanning the building facade 801 and environmental objects within a predetermined radius.

[0040] Through the above layout, the photovoltaic modules 101, distributed matrix nodes 200, high-voltage DC trunk line 302, and grid-connected inverter 500 together constitute a distributed cascaded power generation network with dynamic reconfiguration capabilities. The central control unit 400, as the logic processing center, works with the sensing data from the lidar device 600 to schedule the circuit topology within the distributed matrix node 200.

[0041] See attached document Figure 2 The full-cycle operation control method provided by this invention is implemented based on the aforementioned hardware architecture and includes the following steps:

[0042] Step S100: Perform initial modeling of spatial and electrical parameters. Using a lidar device 600, acquire 3D point cloud data of the existing building 800 and its surrounding environment to establish a spatial model containing geometric location information. Simultaneously, the central control unit 400 controls each distributed matrix node 200 to perform loop tests, measuring the impedance values ​​of the local DC cables 301 from each photovoltaic module 101 to the corresponding distributed matrix node 200, as well as the impedance values ​​of the high-voltage DC trunk lines 302 between adjacent distributed matrix nodes 200, establishing a hierarchical impedance database.

[0043] Step S200: Perform irradiance calculation based on spatiotemporal prediction. The central control unit 400 calculates the solar incidence angle based on the current time and geographical coordinates, combined with a solar position algorithm. It then performs light projection calculations based on a spatial model to determine the geometric shading coefficient of each photovoltaic module 101 surface.

[0044] Step S300: Perform electrical fingerprint inversion calibration periodically or triggered by a specific event. The central control unit 400 selects a specific distributed matrix node 200 to enter bypass test mode. In bypass test mode, the distributed matrix node 200 is temporarily disconnected from the series circuit of the high-voltage DC trunk line 302, and performs open-circuit voltage and short-circuit current scanning on the photovoltaic modules 101 under its jurisdiction. The actual maximum power capacity is calculated by combining the preset fill factor model. The central control unit 400 compares the theoretically calculated power with the actual measured power, inverts the transmittance attenuation coefficient based on the residual between the two, and uses this coefficient to correct the irradiance calculation result in step S200, thereby obtaining the effective predicted irradiance of each photovoltaic module 101.

[0045] Step S400: Based on the corrected effective predicted irradiance, perform an equal-current reverse reconfiguration decision. The central control unit 400 sets a series of candidate global trunk current values. For each candidate current value, each distributed matrix node 200 calculates the maximum port voltage it can provide while outputting that current. The central control unit 400 aggregates the voltage capability data of all distributed matrix nodes 200, combines it with the impedance loss of the high-voltage DC trunk 302, solves for the optimal global current value that maximizes the net output power of the system, and determines the optimal switching state combination within each distributed matrix node 200 corresponding to this optimal current value.

[0046] Step S500: Perform preemptive arc-free reconfiguration. Within a preset time window before the predicted change in illumination arrives, the central control unit 400 sends a current regulation command to the grid-connected inverter 500 and simultaneously sends switching operation commands to each distributed matrix node 200. The grid-connected inverter 500 and the distributed matrix nodes 200 work together to complete the physical switching of the circuit topology at the current zero-crossing point or low current state, enabling the system to enter a new optimal operating state. The system then returns to step S200 to perform the calculation and control for the next cycle.

[0047] In the initialization modeling step S100 for spatial and electrical parameters, the process of reverse modeling the existing environmental geometry specifically includes the following implementation steps:

[0048] Step S110: Perform multi-viewpoint 3D point cloud data acquisition. Given the complexity of the existing building facade structure and the unavoidable blind spots from a single observation perspective, this embodiment employs a multi-station joint scanning strategy to obtain complete environmental data. The lidar device 600 is controlled to optically scan the physical environment within a preset radius (typically set to 100 to 300 meters based on the shadow projection limit) of the existing building facade 800. The scanning range is determined primarily based on a shadow length calculation model under low solar altitude angles in winter, aiming to ensure complete coverage of potential obstructions such as adjacent buildings and tall vegetation. The scanned data covers the external outline of the existing building facade 800 and the photovoltaic module 101 array attached to the facade. In the multi-station data stitching stage, to eliminate systematic errors between different stations, rigid body transformation registration is performed using feature points in the overlapping areas to unify the point cloud in the local coordinate system to the world coordinate system.

[0049] Step S120: Perform preprocessing and filtering of the point cloud data. Because airborne particles (such as dust and fog) or insects flying around in the environment can cause random reflections along the laser transmission path, discrete noise is introduced into the original data. If this noise is not removed, it will reduce the accuracy of subsequent plane fitting. Therefore, the central control unit 400 performs statistical outlier removal on the original point cloud dataset. This operation is based on statistical principles and calculates outlier values ​​at any point in the point cloud. To its The average distance between the nearest neighbors And compare it with the global average distance distribution. Set the judgment threshold as follows. ,in Standard deviation This is the standard deviation factor. In this embodiment, the factor... Typically, a value of 1.0 to 2.0 is used. This range has been experimentally verified to effectively remove discrete noise points while preserving building edge details. After denoising, the point cloud data is downsampled into a voxel grid. This significantly reduces the amount of data while preserving the planar features of the photovoltaic modules, thereby improving the computational efficiency of subsequent algorithms.

[0050] Step S130: Perform semantic segmentation and photovoltaic module entity recognition. Utilizing the spatial smoothness and normal vector consistency characteristics typically found on man-made building surfaces, the preprocessed environmental point cloud set is segmented by region growing. The algorithm uses the point with the minimum curvature as the seed point and compares the deviation of the normal vectors of neighboring points to see if it is less than a preset angle threshold (e.g., set to 5 degrees to tolerate normal vector jitter caused by LiDAR measurement system errors and point cloud discrete noise). Euclidean clustering is used to separate spatially discontinuous point cloud clusters, thereby identifying the independent point cloud subset corresponding to each photovoltaic module 101. To achieve accurate mapping between physical components and control logic, the system assigns a unique logical index number to each identified photovoltaic module based on the spatial coordinates of the geometric center of the point cloud cluster, according to a preset spatial sorting rule (e.g., a top-down, left-to-right scanning order). .

[0051] Step S140: Extract the geometric pose parameters of the photovoltaic module. To transform discrete point cloud data into a continuous mathematical model usable for ray tracing calculations, this embodiment utilizes covariance matrix decomposition to perform planar fitting on a subset of the point cloud. Considering that conventional least squares methods may encounter singularity issues when dealing with planes perpendicular to the coordinate axes, the covariance matrix decomposition method based on principal component analysis (PCA) exhibits good numerical stability when handling spatial planes with arbitrary orientations.

[0052] The specific calculation process is as follows:

[0053] The calculation number is photovoltaic module point cloud subset geometric centroid :

[0054] ;

[0055] in, This represents the total number of point clouds within this subset. For the first The system calculates the three-dimensional coordinate vectors of each point. Before fitting, the system verifies the number of points. Is it greater than the minimum valid point threshold? (e.g., 100 points) to prevent fitting errors caused by missing scan data.

[0056] The physical meaning of constructing the covariance matrix C is to characterize the degree of dispersion (i.e., variance distribution) of point cloud data in various orthogonal directions in space:

[0057] ;

[0058] The superscript T indicates the transpose operation of a vector or matrix.

[0059] For matrix Eigenvalue decomposition yields three non-negative eigenvalues. and its corresponding unit eigenvector , , Since photovoltaic modules have a planar structure, their point cloud has the smallest dispersion in the normal direction, therefore the smallest eigenvalue... Corresponding feature vector This represents the normal vector of the plane of the component.

[0060] To ensure that the fitted result is indeed a plane rather than a cluttered point cloud, the system calculates a flatness index. .like If the flatness exceeds a preset flatness threshold (e.g., 0.05), the area is determined to have severe occlusion or interference from non-planar objects and is marked as an invalid area.

[0061] Furthermore, due to the mathematical eigenvectors There is ambiguity regarding positive and negative directions, and physically, the photosensitive surface of the component must face outwards from the building. Therefore, the system incorporates the macroscopic orientation vector of the existing building body. (This vector can be obtained from a building BIM model or digital compass) is corrected. The final unit normal vector is determined by calculating the dot product sign. :

[0062] ;

[0063] in, It is a symbolic function; For feature vectors; This is the macroscopic orientation vector; This is the corrected normal vector.

[0064] Finally, the system will correct the normal vector. Geometric centroid And the identified environmental obstacle data is encapsulated into a digital spatial model. This provides an accurate geometric reference for subsequent shadow occlusion calculations.

[0065] After completing the spatial geometric modeling, the initialization modeling step S100 for spatial and electrical parameters further includes the calibration process of the graded impedance topology diagram. This process aims to establish the mapping relationship between physical space and electrical properties, transforming discrete geometric coordinates into an electrical network model that can reflect the power transmission loss characteristics of the system. Given that purely theoretical calculations are insufficient to cover the contact resistance of connectors and line aging factors in engineering settings, this embodiment adopts a hybrid calibration strategy using geometric estimation as the benchmark and measured data as correction. Specifically, it includes the following implementation steps:

[0066] Step S150: Perform geometric impedance estimation for local DC cables. Due to the large number of low-voltage DC cables connecting photovoltaic modules and distributed nodes, and the difficulty in performing individual circuit break measurements, the system prioritizes a geometric estimation method based on wiring rules. The central control unit 400 calls the aforementioned digital spatial model. The geometric centroid of each photovoltaic module 101 stored in the middle and digital spatial models Spatial anchor point coordinates of distributed matrix node 200 extracted from Considering that building electrical wiring must strictly adhere to the horizontal and vertical wiring specifications for cable trays (horizontal routing) and vertical shafts (vertical routing), the actual cable length is often greater than the Euclidean distance between two points in space. Therefore, this embodiment preferably uses a weighted Manhattan distance model for length estimation.

[0067] For the number photovoltaic modules up to the first The estimated physical length of the local DC cable 301 of each distributed matrix node 200. The calculation is as follows: ;

[0068] in, To estimate the physical length; This is the wiring detour coefficient, used to compensate for the non-linear path loss caused by avoiding building beams, columns, and ventilation ducts; The geometric centroid of the photovoltaic module The three-dimensional coordinate components, The coordinates of the spatial anchor points of the distributed matrix nodes The three-dimensional coordinate components; This is a weighting factor for vertical cabling. Considering that cables in vertical shafts need to be fixed for stress relief at certain intervals (usually laid in an S-shape), this factor is usually taken as 1.05 to 1.10. To allow for a margin of 0.5 meters for the crimping of junction boxes and terminals, in this embodiment, the coefficient is set between 1.1 and 1.3 depending on the building type (such as the complexity of the internal structure of a high-rise office building).

[0069] Based on this estimated physical length The system further calls the corresponding cable cross-sectional area from the preset cable specification database. Compared with standard resistivity Calculate the basic resistance value of the local circuit based on Joule's law. :

[0070] ;

[0071] in, The standard resistivity of copper conductors (approximately) ), The cross-sectional area of ​​the cable is given, and the coefficient 2 represents the total length of the bidirectional loop between the positive and negative terminals. To estimate the physical length, the system will perform a verification before calculation. Check if it is a positive real number to prevent division by zero errors caused by missing parameters.

[0072] Step S160: Perform differential injection measurement on the high-voltage DC trunk line. This step requires the system to be under effective illumination (e.g., irradiance greater than 200 W / m²). 2 The test is performed under the power generation state. Since the high-voltage DC trunk line 302 runs through the building shaft, there are multiple circuit breaker contacts and connectors along its long transmission path. The contact resistance of these nodes accounts for a large proportion of the total impedance and cannot be determined through geometric calculations. Therefore, this embodiment uses an active differential current injection method based on Kirchhoff's laws for actual measurement and calibration. The central control unit 400 first sends a command to release any possible bypass test mode (i.e., exit the S300 state), then forces all distributed matrix nodes 200 to close their cascade switches and enter the series conduction state, and sends a constant current load command to the grid-connected inverter 500 to control it to absorb a stable test current on the DC side. To ensure the signal-to-noise ratio of the measurement, the system is set with a test current. It needs to be more than 10 times the measurement accuracy of the current sensor (e.g., set to 5A to 10A).

[0073] Simultaneously, the central control unit 400 sends synchronization sampling commands to all distributed matrix nodes 200 via an industrial communication bus 700 supporting IEEE 1588 PTP (Precision Time Protocol). Each distributed matrix node 200 utilizes its internal high-precision voltage sensor to acquire the port voltage at its output terminal within the same global clock cycle (within a microsecond error range). The synchronization mechanism here is crucial, designed to eliminate spurious voltage differentials introduced by dynamic fluctuations in the system bus voltage.

[0074] For the connection of the first The node and the first The trunk section between nodes, its trunk impedance The calculation formula is:

[0075] ;

[0076] in, For trunk line impedance, Port voltage, For testing current.

[0077] In its computational logic, the system incorporates a numerical stability protection mechanism: real-time monitoring of the measured current value flowing through the main line. If this current value is lower than a preset minimum effective threshold (e.g., 0.5A, indicating the line is unloaded or open-circuited), the system will forcibly terminate the division operation and mark the measurement as invalid to avoid numerical divergence caused by the denominator approaching zero. Furthermore, to eliminate zero-point drift error from the sensor, the system will adjust the current under different current conditions (e.g., ...). and Repeat the above process and fit the voltage and current curves using the least squares method, taking the slope as the final trunk impedance calibration value.

[0078] Step S170: Perform temperature drift correction based on environmental awareness. According to the physical properties of metallic conductors, their resistivity increases linearly with increasing temperature. Given the temperature gradients existing between different floors and facades of the existing building (e.g., the top floor receives more direct sunlight and has a higher temperature, while the basement is cooler), the static impedance model will introduce significant errors. The system incorporates temperature correction logic, with temperature sensors integrated within each distributed matrix node 200 uploading the ambient temperature in real time. .

[0079] The following linear temperature drift model is used to correct the impedance of local DC cables and trunk lines:

[0080] ;

[0081] in, To correct the impedance value, The nominal impedance or measured impedance obtained from the preceding steps. The temperature coefficient of resistance of a conductor material (copper is approximately...) ), Reference temperature (usually) The corrected impedance value can more accurately reflect the electrical loss characteristics of the system under the current actual operating conditions.

[0082] Step S180: Construct a hierarchical impedance topology matrix. To support subsequent algorithms in solving the power flow and planning the optimal path for the entire system, the central control unit 400 structurally encapsulates the temperature-corrected local cable impedance set and the trunk impedance set to generate an electrical topology adjacency matrix. The matrix Defined as a The weighted square matrix ( (total number of electrical nodes in the system), its elements Indicates the first The node and the first The direct connection impedance between nodes. The specific assignment rules are as follows:

[0083] If there is a physical connection between the two nodes, then Equal to the corresponding corrected impedance value If there is no physical connection, then Initialized as infinity in computer floating-point representation, representing electrical isolation. Through the above steps, the system completes the precise mapping from physical space geometric information to electrical network parameters, providing a complete data foundation for subsequent isostatic reconfiguration decisions.

[0084] Based on the aforementioned initialized spatial and electrical coupling model, this embodiment further executes the transmittance correction method S200 based on electrical fingerprint inversion. The core technical concept of this step is to establish a digital twin benchmark, calculate the theoretical power that the photovoltaic module should have under ideal clean conditions through physical modeling, and compare it with the measured power. Since geometric shading is a deterministic variable based on the building model, while dust shading is a random variable, the system accurately isolates the effects of geometric shading and temperature drift from the total power loss, attributing the remaining power residual solely to transmittance loss, thereby achieving a quantitative inversion of the cleanliness of the module surface. Specifically, the implementation steps are as follows:

[0085] Step S210: Perform dynamic calculation of time-varying solar geometry and shading factor. To ensure strict alignment between physical calculations and electrical sampling in the time dimension, the central control unit 400 extracts the global timestamp when the distributed matrix nodes upload electrical data. Combining the existing building's latitude and longitude coordinates, the system uses a built-in high-precision solar position algorithm (such as the SPA algorithm released by NREL) to calculate the current solar altitude angle. and solar azimuth Construct a unit direction vector pointing towards the center of the sun. .

[0086] Based on this, the system utilizes a digital spatial model For numbered The photovoltaic modules perform high-resolution ray tracing. To accurately capture the shading effects of minute obstacles such as lightning rods and cables, this embodiment abandons the coarse center-point determination method and instead employs a grid discretization strategy. The system discretizes the light-receiving surface of the module into... A uniformly distributed sampling point (e.g.) (Grid), checking one by one whether the rays emitted from each sampling point in the opposite direction of the solar vector intersect with the surrounding obstacles.

[0087] Meanwhile, considering that the reflectivity of light on the glass surface increases nonlinearly and rapidly with increasing incident angle (Fresnel effect), using only the cosine law would introduce errors. This embodiment calculates the geometric illumination efficiency factor. At that time, an incident angle correction factor (IAM) based on the ASHRAE model was introduced, and its calculation formula is as follows:

[0088] ;

[0089] Among them, the first item Characterizing the geometric occlusion ratio, where This is a binary occlusion function (1 if not occluded, 0 otherwise); the second term Characterizing geometric cosine loss, For component normal vector, The solar vector; the third term To be based on the angle of incidence The Fresnel reflection correction coefficients are obtained by looking up a table or fitting a model. Through this step, the system mathematically quantifies and isolates the hard shadow loss caused by the external environment.

[0090] Step S220: Construct a theoretical power generation benchmark model under all operating conditions. To invert transmittance, a theoretical expected power output value must be constructed under ideal conditions of complete cleanliness and no aging. The system acquires the total radiation intensity of the horizontal plane measured by the total radiation meter on site in real time. and ambient temperature .

[0091] Since total radiation meters are typically installed horizontally, while photovoltaic modules are tilted along with the building roof, this embodiment utilizes an anisotropic sky scattering model (such as the Hav-Davies model) to convert horizontal radiation into effective irradiance on the inclined surface of the module. Furthermore, considering the physical fact that the operating temperature of the solar cells is higher than the ambient temperature, the system uses a steady-state thermal model to extrapolate the actual junction temperature of the solar cells based on the rated operating temperature parameters of the module. Based on the above parameters, a theoretical DC power prediction model is constructed. :

[0092] ;

[0093] in, This refers to the nominal peak power of the component under standard test conditions. This is the irradiance normalization factor; It is the power temperature coefficient (usually a negative value, characterizing the band contraction effect caused by high temperature). Standard reference temperature; This is the calculated junction temperature of the solar cell. This is the geometric illumination efficiency factor.

[0094] The model defines the physical upper limit of the power that the component should output under the current geometric location and weather conditions.

[0095] Step S230: Perform transmittance inversion calculation based on residual analysis. After eliminating the effects of geometric obstruction and temperature, the remaining deviation between the actual power and the theoretical power characterizes the optical health of the component surface. In this embodiment, this deviation ratio is defined as the transmittance attenuation coefficient.

[0096] The central control unit 400 utilizes the actual power of the component at its maximum power point obtained in the bypass test mode of step S300. Based on the principle of energy conservation, the transmittance attenuation coefficient is inverted. :

[0097] ;

[0098] in, This is the inherent annual degradation factor of the component (used to eliminate the interference of irreversible material aging on transmittance calculation). To efficiently calculate the power threshold (e.g.) The technical purpose of setting this threshold is to avoid numerical singularities: during periods of low irradiance in the early morning and late evening, the denominator approaches zero, which can lead to calculation divergence. In such cases, the system automatically retains the estimated value from the previous moment. Finally, the system applies a sliding time window filter to the instantaneous values ​​and outputs the corrected transmittance. This parameter directly quantifies the degree of dust accumulation. When it is lower than the preset cleaning threshold (such as 0.85), intelligent operation and maintenance decisions can be triggered.

[0099] To ensure the long-term stability and data confidence of the aforementioned transmittance inversion results, the system needs to address issues such as sensor zero-point drift, sampling asynchrony caused by communication bus congestion, and interference from outlier noise. This embodiment constructs a closed-loop calibration mechanism by executing online polling calibration steps S240 to S270, combining static zero-point correction, dynamic synchronous snapshots, and spatial consistency verification. This mechanism aims to provide a high-fidelity, time-aligned, and denoised input data stream for electrical fingerprint inversion. Specific implementation details are as follows:

[0100] To address the zero-point drift problem caused by long-term sensor operation, this embodiment performs a sensor zero-point drift correction step S240. After prolonged operation, the Hall current sensor and voltage sampling circuit are affected by changes in ambient temperature and component aging, resulting in zero-point shift. If not eliminated, this shift will introduce a significant relative error during low-power operation. Here, utilizing the natural characteristic that the photovoltaic system does not generate electricity at night, the central control unit 400 continuously monitors the total horizontal radiation intensity. When the system is detected to be at night or in a state of extremely low irradiance (i.e. Less than the preset hibernation threshold If the duration exceeds the stabilization window (e.g., 10 minutes), a zero-point calibration command is triggered. In this embodiment, The preferred setting is 5W / m 2 The value is selected because it is slightly higher than the nighttime zero-point noise limit of the total radiation meter, which can effectively prevent false triggering.

[0101] Upon receiving the instruction, each distributed matrix node 200 samples the port voltage and loop current of each photovoltaic module 101 connected to it multiple times (e.g., continuously sampling for 100 cycles to eliminate random thermal noise) and calculates the arithmetic mean to obtain the voltage zero-point drift value. With current zero-point drift value (where subscript) , Indicates the first The node connected to the first (The node consists of several components, hereinafter the same), and stores these two correction parameters in the node's non-volatile memory. During subsequent normal power generation operation, the node directly performs linear compensation at the hardware level, that is, subtracts the corresponding zero-point drift value stored above from the real-time collected electrical quantities, thereby eliminating systematic measurement errors at the source end.

[0102] To address the data asynchrony issue caused by bus transmission delays, this embodiment implements a time-division multiplexing polling step S250 based on a synchronous snapshot mechanism. Given the bandwidth limitations of industrial fieldbuses (such as RS485), traditional sequential polling results in sampling time differences of up to seconds between the first and last nodes, making it impossible to precisely align electrical data with the instantaneously changing irradiance data. Therefore, at the beginning of each control cycle, the central control unit 400 sends a high-priority synchronous snapshot broadcast frame to the entire network.

[0103] Upon receiving the frame, all distributed matrix nodes 200 immediately lock the current electrical parameters and temperature data using a hardware interrupt, and add a unified global timestamp. Subsequently, each node, according to the preset logical address order, sequentially sends data with a global timestamp within its allocated time slot window. The data packet upload mechanism achieves virtual simultaneity at the physical layer, ensuring strict alignment of the electrical fingerprint data of all photovoltaic modules in the time dimension, fully supporting the technical features of synchronous acquisition in the claims.

[0104] Based on this, in order to remove abnormal data caused by sensor malfunctions or localized occlusion (such as bird droppings), the system executes an outlier removal step S260 based on spatial neighborhood correlation. According to the first law of geography, spatially adjacent photovoltaic modules should have similar power generation performance under the condition of no anomalies.

[0105] Based on this principle, the system, for any photovoltaic module 101 to be verified (denoted as the target module), uses a digital spatial model... Identify its physical space A set of nearest neighbor components (e.g., adjacent components within a radius of 10 meters, and) To ensure statistical accuracy.

[0106] To avoid interference caused by differences in power levels among different components, this embodiment uses a dimensionless instantaneous performance ratio as a verification metric. The system calculates the global timestamp of all components within this neighborhood set. Instantaneous performance ratio And count the number of items in the set. median value with standard deviation The median, rather than the mean, is chosen to avoid the influence of existing outliers in the neighborhood on the baseline value (a robust statistical property). Based on this, a spatial consistency criterion for the target component is constructed. :

[0107] ;

[0108] in, The instantaneous performance ratio of the target component. To prevent extremely small positive numbers with a denominator of zero (e.g., 10) −6 When the calculation is obtained When the value exceeds the preset abnormal threshold, the system determines that the data of this component is outlier noise.

[0109] In this embodiment, the preset anomaly threshold is preferably set to 3.0. According to the normal distribution theory, this threshold covers 99.7% of the normal data fluctuation range, and data exceeding this range has a very high anomaly confidence level. For components identified as outliers, the system does not update their transmittance but directly uses the average transmittance of the neighborhood as a replacement value, and marks the photovoltaic module 101 as a suspected fault in the operation and maintenance log.

[0110] Finally, to obtain a stable and reliable cleanliness index, the system performs a weighted confidence update step S270 for transmittance. Since the signal-to-noise ratio of a photovoltaic system deteriorates with decreasing irradiance (signal is weak while noise remains constant), transmittance calculated during early morning, late evening, or cloudy hours is often unreliable. This embodiment introduces a confidence weighting factor that is positively correlated with the theoretical power level. And update the current smooth transmittance using a recursive filter. :

[0111] ;

[0112] in, This is the inversion transmittance attenuation coefficient (original value with noise) obtained from the inversion. This is the smoothed value from the previous time step. Weighting factor. Defined as , The squared term is introduced to non-linearly and rapidly suppress noise weights in the low-power range, ensuring that transmittance is significantly updated only when there is sufficient light. The base update rate (preferably ranging from 0.01 to 0.05) determines the system's sensitivity to changes in cleanliness.

[0113] Through this weighted filtering, the system can effectively filter out transient fluctuations and output smooth and reliable component health indicators.

[0114] In a series-connected distributed cascaded system, to achieve equal-current reverse reconfiguration control, the system must first establish an accurate physical constraint model, that is, quantify how much voltage support each distributed matrix node 200 can provide at the current moment. This evaluation process is not a simple numerical reading, but a dynamic reconfiguration process involving environmental perception, photoelectric conversion physical modeling, and hardware safety boundary constraints. In this embodiment, by executing step S300, discrete environmental parameters are transformed into voltage boundary commands that the control system can execute.

[0115] In step S310, the central control unit or each distributed matrix node 200 performs synchronous reading of the environment and device status. The system reads the global timestamp. Corresponding effective irradiance of the inclined plane Smooth transmittance of photovoltaic module 101 and component backsheet temperature At the same time, the system retrieves the component's Standard Test Condition (STC) parameters, including the standard open-circuit voltage, pre-stored in non-volatile memory. Temperature coefficient of open circuit voltage In this embodiment, the consistency of the data source within the system is ensured by directly reusing the geometrically corrected irradiance data calculated in the preceding steps. In step S320, the system performs a net effective irradiance calculation based on transmittance correction. To quantify the actual attenuation effect of the dust accumulation layer on the light energy received by the photovoltaic module, this embodiment defines net effective irradiance. This represents the actual light intensity that penetrates the glass layer and reaches the surface of the photovoltaic cell's PN junction.

[0116] In step S320, the system performs a net effective irradiance calculation based on transmittance correction. To quantify the actual attenuation effect of the dust accumulation layer on the light energy received by the photovoltaic module, this embodiment defines net effective irradiance. The formula for calculating the actual light intensity that penetrates the glass layer and reaches the surface of the photovoltaic cell's PN junction is as follows:

[0117] ;

[0118] in, For the first Net effective irradiance corresponding to each node The effective irradiance of the inclined plane, This represents the normalized transmittance. This step establishes the energy transfer mapping from surface light to light within the junction, providing real input variables for solving subsequent electrical equations.

[0119] In step S330, the system reconstructs the theoretical maximum open-circuit voltage based on an improved engineering single-diode model. As a typical nonlinear DC source, the output voltage characteristics of a photovoltaic cell are affected by the dual coupling of temperature and light intensity: the bandgap of crystalline silicon narrows with increasing temperature, leading to a linear decrease in voltage, while reduced light intensity decreases the photocurrent concentration, causing the voltage to decay logarithmically. To cover all operating conditions, this embodiment uses an analytical model that includes linear temperature correction and logarithmic net irradiance correction to calculate the theoretical maximum open-circuit voltage at the current moment.

[0120] ;

[0121] in, The maximum open-circuit voltage, Standard open-circuit voltage, For the component backsheet temperature; Standard irradiance; δ is the open-circuit voltage temperature coefficient; δ is a minimal positive compensation term (e.g., 10⁻⁶). Specifically, the coefficient... It characterizes the overall thermal-voltage characteristics of the component (related to the Boltzmann constant and the number of battery series stages).

[0122] In this embodiment, the parameter (typically 0.05V-0.15V) is preferably obtained by offline fitting of the measured V-I curve of the component. This formula, by introducing a logarithmic term, accurately characterizes the physical characteristics of the rapid drop in the open-circuit voltage of the component under low light conditions such as shade or dirt blockage.

[0123] In step S340, the system determines the final maximum voltage capability of the node by combining hardware limits and control safety margins. .Although This characterizes the potential of the physical potential difference, but the actual usable voltage range must be limited by the safe operating area (SOA) and control stability requirements of the power device. Therefore, this embodiment introduces a multi-dimensional decision logic that includes low-light cutoff of net irradiance and safety margin cutoff:

[0124] ;

[0125] in, For maximum voltage capability, The maximum allowable hardware voltage for distributed matrix nodes 200; Minimum net irradiance threshold (e.g., 20 W / m²) to maintain normal operation of the converter drive circuit 2 , The modulation coefficient is set to a safe value (preferably 0.90-0.95).

[0126] This decision logic ensures that the system only includes the node in the voltage regulation resource pool when the actual light energy entering the solar cell is sufficient to maintain stable control, thereby avoiding invalid oscillations under low-energy-efficiency conditions at the algorithm level.

[0127] After defining the physical boundaries of node-level voltage capabilities, the system's control focus shifts from local constraints to maximizing global efficiency. Given that the distributed cascaded system follows series circuit characteristics in its electrical topology, according to Kirchhoff's current law, all series-connected distributed matrix nodes 200 must operate under the same current value (i.e., the global common current). However, due to environmental factors such as component aging differences, uneven dirt distribution, and shading, the optimal operating current point of each photovoltaic module 101 exhibits discrete characteristics in numerical terms. Under this condition, simply using the average current or minimum current as the control benchmark will inevitably force some high-efficiency nodes to deviate from their peak power regions, resulting in a potential loss in system-level power generation. Therefore, this embodiment executes the global equal-current optimization decision step S400, using a reverse reconstruction algorithm based on discrete characteristic spectra to accurately locate the unique optimal current solution that maximizes the total system power on the multi-peak power curve.

[0128] In step S410, the central control unit constructs the network-wide current characteristic spectrum based on the net effective irradiance. To determine the feasible search domain for the global common current, the system first needs to know the theoretical maximum current potential of each node under the current environment. The system calls the net effective irradiance calculated in step S320. and the real-time backplane temperature of the components Using the photoelectric conversion principle commonly used in this field, the theoretical peak current of each distributed matrix node 200 (after series or parallel connection of its internal photovoltaic modules 101) at the current moment is estimated. Specifically, the system uses the component's standard short-circuit current as a benchmark, linearly converts it according to the ratio of the current net effective irradiance to the standard irradiance, and adds a correction term based on the temperature coefficient. This is achieved by analyzing the entire network... The system iterates through each node to generate a discrete current set containing the peak current characteristics of all nodes. As a preferred processing method, after generating the set, the system will perform deduplication and ascending sorting operations, and remove sets with currents below the system's minimum starting current (e.g., ...). By identifying invalid extrema, a compact and efficient finite search space is constructed, avoiding the waste of computational resources caused by blindly searching in the continuous current domain.

[0129] It should be noted that, in this embodiment, the distributed matrix node 200 can be configured as a multi-MPPT input parallel structure or an optimized structure for a single component. During the reverse reconstruction calculation in step S420, the system treats the node as an equivalent Thevenin power supply model.

[0130] In step S420, the system performs a voltage-constrained power reverse reconstruction calculation. For the set Each candidate current value (in (As a set index), the system needs to predict the terminal voltage that each node can provide when the entire network is forced into series operation under this current. A voltage response reconstruction function is introduced here. Based on the physical characteristics of photovoltaic cells and their bypass protection circuits: when the series circuit current exceeds the short-circuit current under the current illumination of the module... At the maximum photocurrent limit, the high impedance inside the component causes the bypass diode to conduct in the forward direction, shorting the node and causing the output voltage to drop rapidly to near zero. This embodiment uses the following segmented logic to digitally model this physical phenomenon:

[0131] ;

[0132] in, In the candidate current Next The estimated terminal voltage of each node; To prevent extremely small thresholds where the denominator is zero (e.g., 0.01A); This represents the voltage-current coupling droop factor. In this embodiment, The preferred value range for this coefficient is 0.05 to 0.1. This coefficient reflects the slope characteristic of the photovoltaic module's IV characteristic curve to the left of the maximum power point (constant voltage region), and characterizes the change in output current. The increase of the port voltage relative to the open circuit voltage It exhibits the physical characteristic of linear decay.

[0133] In step S430, the system synthesizes a global power prediction curve and determines the optimal common current reference. For each candidate current... The system calculates the expected total system-level power by multiplying the sum of the reconfiguration voltages of all nodes by the candidate current. This calculation process reveals the core technical characteristics of reverse reconstruction: instead of passively responding to current by adjusting voltage, it actively predicts the voltage distribution under different current setpoints and then derives the power. This is achieved by traversing the set... For all candidate points, the system can construct discrete global power versus current prediction curves. The central control unit identifies the candidate current value corresponding to the maximum power value through numerical comparison, solves for the optimal global current value that maximizes the system's net output power, and locks it as the current provisional global optimal common current value. This decision-making logic ensures that the system can automatically identify and discard local suboptimal solutions that, although they have high current, would lead to too many nodes being bypassed, thus reducing the total power.

[0134] In step S440, the system performs instruction smoothing and output based on hysteresis comparison. To avoid frequent jumps in control commands between two adjacent candidate values ​​due to computational noise, sensor errors, or minor environmental fluctuations, this embodiment introduces hysteresis filtering logic in the final output stage. The system only outputs the result when a new optimal target value is calculated. The corresponding power gain, compared to the system power in the previous control cycle, exceeds the preset power sensitivity threshold. (Preferably set to 1.5% to 3.0%), only then will the target value be updated to the final globally optimal common current command. Otherwise, the system maintains the current command from the previous moment. This mechanism, while maintaining maximum power point tracking accuracy, sacrifices a small amount of theoretical power gain to achieve smooth steady-state operation of the system, reducing thermal stress and regulation losses of the power converter switching devices.

[0135] In the preceding step S400, the system-level globally optimal common current command is determined. Subsequently, if the control system directly sends the instruction as a step signal to each distributed matrix node 200, due to the parasitic inductance of the DC trunk and the differences in the dynamic response characteristics of the power converters at each node, the sudden and drastic change in current can easily induce high-frequency, high-voltage oscillations in the circuit, and may even induce DC arcing at the connection point. To avoid this safety risk from a physical perspective, this embodiment executes the instruction generation and issuance step S500 in the predictive time domain, adopting a preemptive control strategy that decouples the instruction timing from the execution action, transforming a single steady-state target value into a dynamic, smooth trajectory with timing constraints.

[0136] In step S510, the central control unit establishes the predictive control time domain and calculates action delay compensation. Given the non-negligible physical time required for data transmission from the central unit to each edge node in a distributed architecture, and the potential slight deviations in the clock phases of each node, the system must introduce the concept of future-time control to ensure that all nodes in the network can synchronously execute actions at a precise moment. The central control unit reads the current system global reference clock. In conjunction with historical statistical data of network communication, the unified command take-off time is calculated. :

[0137] ;

[0138] in, The time when the instruction takes effect. This serves as the statistical upper bound for network transmission latency (e.g., taking the 99th percentile of the round-trip time of the past 100 heartbeat packets). This allows time for the underlying DSP of the node to perform unpacking and computing power scheduling. This is the safety redundancy factor.

[0139] In this embodiment, The preferred value range for this coefficient is 1.1 to 1.3. This coefficient is introduced to cover random factors such as network jitter, ensuring that commands can be executed even under the worst communication conditions. All nodes arrive before the command is executed, thus achieving a synchronization mechanism where the clock precedes the command.

[0140] In step S520, the system performs arc-free trajectory smoothing based on ramp rate limits. This is to prevent abrupt changes in the current setpoint from inducing overvoltage in the equivalent inductance of long-distance DC cables (following the laws of physics). The central control unit needs to output the step target in step S440. This is converted into a ramp command that complies with physical safety constraints. The system first calculates the prediction time window length. To avoid computational overflow or control divergence due to an excessively small time window, if Less than the system minimum control period If the update fails to complete within 1ms, the system will either abandon the update or postpone it. .exist Provided it is effective, the system operates based on the preset maximum allowable rate of current change. (Unit: A / ms), Calculation for future time periods Security command value :

[0141] ;

[0142] in, For security command values, For the maximum gradeability, This represents the measured operating current at the current moment. sgn represents the desired current increment. It is a symbolic function.

[0143] The above formula clarifies the cutoff logic: when the desired change in current... Within the allowed physical climbing ability When the target value is within the acceptable range, the system responds directly; conversely, when the expected change is too large, the system forces the maximum ramp rate. Perform output truncation. The specific value depends on the system insulation withstand voltage rating. Total inductance of the circuit Typically, it needs to meet the following requirements. This is to ensure that the induced voltage is always within the insulation safety margin.

[0144] In step S530, the system encapsulates an atomic control frame containing a timing tag and performs a broadcast. To ensure control consistency and atomicity, the central control unit packages the calculated control parameters into a standard data frame. As a preferred data structure, this control frame strictly includes the following fields: a multicast group ID specifying the range of received groups, a type bit identifying the synchronization priority, and a current command smoothed by S520. The corresponding overvoltage protection clamping value and the core instruction activation time. Subsequently, the central control unit performs a distribution operation via a communication medium such as power line carrier (PLC) or wireless LAN, that is, the central control unit 400 sends control commands (which are encapsulated atomic control frames) to the distributed matrix nodes 200. Upon receiving the frame, each node does not immediately refresh its hardware registers, but instead parses and stores it in its local cache queue until the local clock ticks precisely reach the specified value. In that microsecond, the underlying digital signal processors (DSPs) of all nodes simultaneously update the reference values ​​of their pulse width modulation (PWM) modules. This timestamp-based triggering mechanism eliminates the reception time difference caused by the different mounting positions of nodes on the communication bus, achieving millisecond-level synchronous reconstruction across the entire network.

[0145] In the preceding step S500, a slope smoothing control strategy is employed to address small current fluctuations, effectively suppressing overvoltages caused by line inductance. However, when the system encounters extreme conditions such as sudden disappearance of local shadows, rapid cloud movement, or drastic changes in lighting conditions, the system often requires significant reconfiguration of the grid topology (e.g., series-parallel switching) or duty cycle. This is based on the principle of capacitor charging and discharging. The step-like voltage jump in the DC bus accompanying this reconfiguration will generate a huge inrush current on the DC bus capacitor of the grid-connected inverter. This can easily trigger the inverter's overcurrent or overvoltage hardware protection, leading to a shutdown. Furthermore, it may generate a destructive DC arc due to the forced disconnection of the inductive circuit under high current, burning the switch contacts. Therefore, this embodiment executes a cold switching operation step S600 in coordination with the inverter. Through collaborative interaction with the back-end grid-connected inverter, a safe time window is artificially constructed, sequentially including active power derating, topology reconfiguration, and power soft-start recovery. This allows the system state transition to be completed under the physical state with minimal electrical stress.

[0146] In step S610, the central control unit performs threshold determination and mode decision for the reconfiguration amplitude. To accurately identify when a high-priority cold switching process needs to be initiated, the system needs to predict in real-time the voltage surge amplitude of the upcoming reconfiguration action on the DC bus based on the current electrical state. This decision-making process does not rely on a single instantaneous voltage value, but rather calculates the estimated difference in the system's steady-state voltage before and after the reconfiguration. The system bases this on the maximum voltage capability of each distributed matrix node 200 under the current irradiance. The target duty cycle set with respect to the reconstruction target The following formula can be used to estimate the expected total voltage change of the entire series branch at the moment of reconstruction. :

[0147] ;

[0148] in, This represents the current DC bus voltage measured by the sensor; the summation term represents the theoretically expected superimposed output voltage of the entire network after reconstruction. The system will calculate the obtained... Compared with the preset cold switching trigger threshold A logical comparison is performed. In this embodiment, The value is mainly determined by the withstand voltage margin of the DC bus capacitor of the downstream inverter and the overvoltage protection trigger threshold, and is preferably set to the rated operating voltage of the inverter. Up to 10%. If The system determines that the voltage fluctuation is within the buffering capacity of the capacitor and continues to use the smooth transition strategy; however, when When the system identifies a potential impact risk, it immediately activates the cold switch mode, suspends the current refactoring command, and enters an active intervention state.

[0149] In step S620, the system issues an active derating command to the inverter and monitors the bus current decay. To eliminate the risk of arcing during load switching at its source, the central control unit uses standard industrial fieldbus protocols such as Modbus-TCP, SunSpec, or Power Line Carrier (PLC) to send a high-priority power dispatch command to the grid-connected inverter. This command forcibly freezes the inverter's Maximum Power Point Tracking (MPPT) algorithm and lowers the DC-side input absorption current limit to the safe switching current threshold. As a preferred method, Set as to The microcurrent level is significantly lower than the minimum physical current required to sustain a DC arc in dry air (typically 1). (Approximately), thereby disrupting the physical conditions for arc generation. At this point, the central control unit enters a closed-loop monitoring cycle, reading the actual value of the bus current fed back by the inverter in real time, and determining whether it has dropped below the safety threshold, including the sensor measurement noise margin. Considering that packet loss or delay may occur in industrial field communication, this step is also configured with a watchdog timeout mechanism: if the bus current fails to complete the derating within a preset time window (e.g., 500ms), the system will forcibly terminate this reconfiguration task and report a communication link anomaly alarm, ensuring that no physical action is performed under uncontrolled conditions.

[0150] In step S630, the system performs a physical switching reconfiguration under zero-current conditions while in a micro-current state. Once it is confirmed that the bus current has decreased below the safety threshold, according to the principles of physics, the magnetic field energy stored in the loop inductor has decayed to the microjoule level. The central control unit then triggers all distributed matrix nodes 200 to perform a predetermined topology reconfiguration or a significant duty cycle adjustment via a broadcast command. Since the current flowing through the power switching devices is extremely small at this time, this switching process is electrically similar to a no-load operation, which not only completely avoids the risk of contact arcing but also reduces the switching losses and thermal stress impacts of insulated-gate bipolar transistors or metal-oxide-semiconductor field-effect transistors during high-current load switching. After the operation is completed, each node feeds back a status ready signal to the central control unit, indicating that the physical reconfiguration at the hardware level has been fully completed.

[0151] In step S640, the system removes the power limit and executes the soft-start recovery logic. After confirming that all nodes are operating stably at the new topology or voltage setpoint, the central control unit sends a power recovery command to the inverter, disabling the MPPT function. To avoid a sudden current surge causing bus voltage oscillation or overshoot again, this embodiment does not immediately restore full power, but instead instructs the inverter to perform a linear soft start. Specifically, the system controls the inverter's current absorption limit value from... Start by restoring the climbing rate at the preset power. The current is linearly increased (e.g., 5 A / s) until the new maximum power point current after reconfiguration is reached. Through this logic, the system can smoothly increase the operating current from the microamp level to the new optimal operating point, allowing the DC bus voltage to smoothly transition to the new equilibrium value under controlled conditions. This completes the cold switching closed-loop control, which includes system derating, physical reconfiguration, and power recovery. Although this process sacrifices hundreds of milliseconds of instantaneous power generation, it ensures the electrical safety and hardware lifespan of the system under extreme conditions, achieving a dynamic balance between efficiency and safety.

[0152] Specific application examples:

[0153] Project Background:

[0154] A high-rise office building (built in 2005) located in a business district of a certain city was selected as the renovation target.

[0155] The existing building body is 800 meters high (20 floors) with a south-facing facade.

[0156] Environmental challenges:

[0157] There is a Shin Kong Tower on the southeast side, which casts a moving shadow on the east facade of the office building between 9:00 and 11:00 in the morning.

[0158] The building is located near an elevated highway, and the photovoltaic modules on the lower floors (floors 1-8) accumulate dust much faster than those on the upper floors.

[0159] The electrical shaft 802 has a narrow space, forcing the DC trunk line to adopt a roundabout wiring method, and the resistance of some of the original old cable joints is unknown.

[0160] System Configuration:

[0161] Photovoltaic array assembly 100: A total of 400 monocrystalline silicon photovoltaic modules 101 are arranged, each with a rated power of 450W, and a total installed capacity of 180kW.

[0162] Distributed matrix node 200: One node is set up for each layer, and each node manages 20 components in that layer, with a total of 20 nodes connected in series.

[0163] Communication and Sensing: RS-485 industrial bus 700 connection is used; a solid-state LiDAR 600 is deployed on the roof.

[0164] Detailed explanation of the operation process:

[0165] Scene time: 10:30 am on the winter solstice, temperature 5°C, southeast shadow covers the eastern part of the components of the 1st to 5th floors.

[0166] Step S100 (Sensing and Impedance Calibration): LiDAR scanning detected geometric occlusion on the right side of layers 1-5.

[0167] Key finding: During impedance self-testing, the system detected an abnormally high trunk impedance (0.8Ω, normally 0.05Ω) between the nodes on the 12th and 13th floors, presumably due to oxidation of circuit breaker contacts somewhere in the shaft. This data was incorporated into the hierarchical impedance topology matrix.

[0168] Steps S200-S300 (Transmittance Inversion): A comparison with the theoretical model revealed that although layers 6-8 had no geometric obstruction, the actual power was only 88% of the theoretical value. Inversion calculations showed that the transmittance attenuation coefficient for these layers was 0.88 (heavily dusty), while the transmittance for layers 15 and above was 0.96 (clean). The system corrected the effective predicted irradiance: the effective light intensity of the lower-layer components was significantly reduced.

[0169] Step S400 (Global Optimization Decision):

[0170] Traditional solution logic: Typically, the MPPT tracks the upper-layer module with the strongest light exposure and sets the current to 11A (peak). This results in all lower-layer modules with accumulated dust and those that are shaded being bypassed, and when the current flows through the high-impedance section of layers 12-13, it generates... Heat loss.

[0171] The logic of this invention is as follows: The system generates candidate currents, and the results show that the overall system efficiency is highest when the global current drops to 7.5A. The specific reasons are as follows:

[0172] The lower-level contaminated components can operate in a non-bypass state and contribute voltage.

[0173] Although the upper-layer components were not fully loaded, the total voltage increased significantly.

[0174] The heat loss in the high impedance section is reduced to This reduced line heat generation by more than half.

[0175] Decision: The system locks in the optimal global current value of 7.5A, at which point the net output power of the entire network is 12% higher than that of the 11A solution.

[0176] See attached document Figure 3 , Figure 3 The horizontal axis represents the global series current, and the vertical axis represents the system's net output power. The dashed line in the figure represents the theoretical power curve without considering line impedance, with the traditional MPPT misjudgment point located at 11A. The solid line in the figure represents the net power curve after impedance sensing correction according to this invention. It can be seen that, due to the consideration of the 0.8Ω trunk impedance and the local obstruction bypass effect, the peak value of the solid line shifts significantly to the left to 7.5A (marked as a square point), and the net power at this point is significantly higher than the actual output under the traditional MPPT misjudgment point of 11A, directly verifying the correctness of the system's current reduction and power increase decision.

[0177] Steps S500-S600 (Safety Reconfiguration): At 10:45, the clouds suddenly dispersed, and sunlight surged. The calculated predicted voltage difference exceeded the threshold. The central control unit 400 triggered a cold switch: first, it instructed the inverter to reduce the current to 0.5A, and then completed the duty cycle reconfiguration of all nodes within 30ms, followed by a soft start recovery. No electric arc was generated throughout the entire process.

[0178] See attached document Figure 4 , Figure 4 The top figure records the change of DC bus current over time, showing that before the topology switching moment (marked by the vertical dashed line), the system actively enters the active current reduction region, reducing the current to the safe threshold of 0.5A. The bottom figure compares the transient response of the bus voltage: the dashed line represents the dangerous overvoltage spike of several hundred volts (arc risk) generated when the control group directly switches under load; the solid line represents the cold switching of the present invention under low current, with a smooth voltage waveform rise and no overshoot spikes, verifying the electrical safety of the control strategy.

[0179] Experimental verification and effect comparison:

[0180] To verify the effectiveness of the present invention, a high-fidelity digital twin simulation platform was built using MATLAB R2025a / Simulink.

[0181] Simulation environment settings:

[0182] Photovoltaic model: Based on a single diode physical model, the building geometry parameters and shadow occlusion files from the embodiments are imported.

[0183] Comparison group settings:

[0184] Control group A (traditional string type): It only has the global scanning MPPT function, does not consider line impedance, and does not have single module communication capability.

[0185] Control group B (conventional optimizer): It has component-level optimization capabilities, but each node operates independently, lacks central unified scheduling, and ignores trunk impedance loss.

[0186] Experimental group (this invention): Enable impedance sensing, transmittance inversion and global isocurrent reverse reconstruction functions.

[0187] Experimental results data:

[0188] The test results are as follows, under the complex dynamic shading and uneven dust accumulation conditions that simulate a whole day (8:00-17:00):

[0189] Performance indicators Control group A (traditional) Control group B (independent optimization) Experimental group (this invention) Improvement rate (comparison A / B) Daily power generation (kWh) 485.2 560.5 592.8 +22.1% / +5.7% Line heat loss (kWh) 18.5 16.2 9.4 Significant reduction in losses Number of voltage oscillations (>50V) 45 times 12 times 2 times Stability greatly improved Risk of forced switching of electric arc high middle None (cold switch) Optimal security

[0190] See attached document Figure 5 , Figure 5The graph visually compares the daily power generation and line heat loss of three different schemes. The light-colored bars on the left axis represent daily power generation (higher values ​​are better), and the dark-colored bars on the right axis represent line heat loss (lower values ​​are better).

[0191] As can be seen from the comparison, the present invention (the rightmost group) not only achieved the highest power generation (592.8kWh), but also reduced the line heat loss to the lowest level (9.4kWh), demonstrating the significant advantages of global equal current reconfiguration in solving the problem of photovoltaic retrofitting of high-impedance old buildings.

Claims

1. A method for retrofitting existing buildings with renewable energy, characterized in that, Includes the following steps: A photovoltaic array assembly (100) and a distributed matrix node (200) are arranged on the existing building body (800), connected to the series power supply network of the grid-connected inverter (500), and a communication connection is established between the central control unit (400) and the distributed matrix node (200) and the lidar device (600). A digital spatial model is established using the lidar device (600), the impedance characteristics of the series power supply network are measured, and a hierarchical impedance topology matrix is ​​established. Based on the digital spatial model, the predicted irradiance is calculated, the actual electrical data is obtained and the transmittance attenuation coefficient is inverted by combining the theoretical model, and the predicted irradiance is corrected by the transmittance attenuation coefficient to obtain the effective predicted irradiance. Based on the effective predicted irradiance, candidate global current values ​​are generated. The voltage response of each distributed matrix node (200) is calculated in combination with the hierarchical impedance topology matrix, and the optimal global current value that maximizes the net output power of the system is solved.

2. The method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, Also includes: The central control unit (400) sends control commands to the distributed matrix nodes (200) to drive the system to switch to a circuit topology state that matches the optimal global current value.

3. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The process of establishing a digital spatial model by the lidar device (600) specifically includes: The lidar device (600) is controlled to perform multi-view three-dimensional point cloud data acquisition, and statistical outlier removal and voxelized mesh downsampling are performed on the acquired point cloud data; Based on the spatial smoothness and normal vector consistency of the preprocessed point cloud, region growing segmentation is performed to identify the independent point cloud subset corresponding to each photovoltaic module (101); The point cloud subset is fitted with a plane using the covariance matrix decomposition method. The geometric centroid of each photovoltaic module (101) is extracted, the macroscopic orientation vector of the existing building body (800) is obtained, and the normal vector of the fitted plane is corrected accordingly to construct the digital spatial model.

4. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The photovoltaic array assembly (100) includes a plurality of photovoltaic modules (101), and the series power supply network includes a local DC cable (301) and a high-voltage DC trunk line (302). The process of establishing the hierarchical impedance topology matrix specifically includes: The geometric coordinates in the digital spatial model are used to estimate the physical length of the local DC cable (301) using the weighted Manhattan distance model, and the base resistance value is calculated. Control all the distributed matrix nodes (200) to enter the series conduction state, control the grid-connected inverter (500) to absorb a constant test current, and use the differential voltage measurement method to determine the measured impedance of the high voltage DC trunk line (302) between adjacent nodes; Ambient temperature data is acquired, and temperature drift correction is applied to the base resistance value and the measured impedance to generate the graded impedance topology matrix.

5. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The method further includes performing a sensor zero-point drift correction step: The total horizontal radiation intensity of the external environment is obtained. When the total horizontal radiation intensity is lower than a preset sleep threshold and the duration exceeds the stable window, a zero-point calibration command is triggered. In response to the zero-point calibration command, each of the distributed matrix nodes (200) is controlled to perform multiple samplings of the port voltage and loop current and calculate the arithmetic mean to obtain the voltage zero-point drift value and the current zero-point drift value; During subsequent power generation operation of the system, the voltage zero-point drift value and the current zero-point drift value are subtracted from the actual electrical data collected in real time.

6. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The process of retrieving the transmittance attenuation coefficient specifically includes: The solar vector is calculated based on the current time and geographical coordinates, and ray tracing is performed using the digital spatial model to calculate the geometric illuminance factor of each photovoltaic module (101). By combining the obtained ambient temperature and real-time total radiation intensity, a theoretical power generation benchmark model under clean conditions is constructed. The distributed matrix node (200) is controlled to enter the bypass test mode to obtain the actual maximum power of the photovoltaic module (101) under its jurisdiction. The ratio of the actual maximum power to the theoretical power generation benchmark model is calculated, and the geometric illuminance factor and temperature influence are eliminated to obtain the transmittance attenuation coefficient characterizing the surface cleanliness of the module.

7. A method for retrofitting existing buildings with renewable energy according to claim 6, characterized in that, The process of retrieving the transmittance attenuation coefficient also includes performing spatial consistency verification and confidence update: Based on the digital spatial model, identify the set of physically adjacent components of the target photovoltaic module (101), and calculate the median value of the ratio of the actual maximum power of all components in the set to the theoretical power generation benchmark model. If the deviation between the ratio of the actual maximum power of the target photovoltaic module (101) to the theoretical power generation benchmark model and the median value exceeds a preset anomaly threshold, the data is determined to be outlier noise and updates are paused. If the deviation does not exceed the preset abnormal threshold, a confidence weight factor positively correlated with the theoretical power level is introduced, and the transmittance attenuation coefficient is smoothly updated using a recursive filter.

8. A method for retrofitting existing buildings with renewable energy according to claim 4, characterized in that, The process of finding the optimal global current value that maximizes the system's net output power specifically includes: Based on the effective predicted irradiance and the component temperature calculated from the ambient temperature, a theoretical peak current set for all nodes in the entire network is generated, and the current values ​​in the set are used as the candidate global current values. For each candidate global current value, determine whether each of the distributed matrix nodes (200) is bypassed due to the current exceeding the short-circuit current limit if forced to operate in series under the candidate global current value, and reconstruct the voltage response function of each node according to the determined bypass state; The reconfigured voltage response of all nodes is summarized, and the impedance voltage drop loss of the series power supply network is calculated and subtracted based on the hierarchical impedance topology matrix. A prediction curve of global power as a function of current is synthesized, and the current corresponding to the peak value of the prediction curve of global power as a function of current is determined as the optimal global current value.

9. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The process of the central control unit (400) sending control commands to the distributed matrix nodes (200) specifically includes: Calculate network transmission latency and processing time to determine a unified command take-off time; Calculate the difference between the current operating current and the optimal global current value. If the rate of change corresponding to the difference exceeds the preset physical ramping capability, the optimal global current value is truncated and smoothed according to the maximum allowable ramping rate to generate a smoothed current command. If it does not exceed the limit, the optimal global current value is directly used as the smoothed current command. A control frame containing the effective time of the instruction and the smoothed current instruction is encapsulated and broadcast to each of the distributed matrix nodes (200) via an industrial communication bus (700) to instruct each node to switch to the circuit topology state corresponding to the smoothed current instruction at the effective time of the instruction.

10. A method for retrofitting existing buildings with renewable energy according to claim 1, characterized in that, The method also includes performing a cold switching operation in conjunction with the inverter: Before performing the reconfiguration, calculate the estimated DC bus voltage difference before and after the reconfiguration; If the difference exceeds the preset cold switching trigger threshold, an active derating command is sent to the grid-connected inverter (500) to limit the input current to below the safe switching current threshold. After detecting that the bus current has dropped to the safe switching current threshold, each of the distributed matrix nodes (200) is triggered to perform circuit topology switching; After confirming that the physical switch is completed, control the grid-connected inverter (500) to release the power limit and perform soft-start linear recovery.

Images