BIPV Shading and Power Generation Efficiency Analysis and Monitoring Methods

By analyzing the operational data of BIPV systems, a state correlation model between photovoltaic strings was established, solving the problems of local shading identification and power generation efficiency assessment in BIPV systems, and achieving accurate analysis and cost reduction in complex environments.

CN123122445APending Publication Date: 2026-11-17ZHUHAI SINGYES GREEN BUILDING SCI & TECH CO LTD +4
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611189916.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-06
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies struggle to accurately identify localized shading in BIPV systems and assess its impact on power generation without adding extra detection equipment, especially in complex building environments where it is difficult to distinguish between normal environmental changes and power reduction caused by shading.

Method used

By analyzing the operational data of the BIPV system, a state correlation model between photovoltaic strings is established. Multi-model estimation and verification relationship set are used to identify shading and evaluate power generation efficiency, including multi-model estimation, verification relationship update and error correction recovery, and reconstruct the shading-free power to determine shading loss.

Benefits of technology

It enables accurate identification of local shading and assessment of power generation loss without adding equipment, improving the accuracy and reliability of analysis in complex environments and reducing system modification costs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122445A_ABST
    Figure CN123122445A_ABST
Patent Text Reader

Abstract

This invention belongs to the field of building-integrated photovoltaics (BIPV) technology and provides a method for analyzing and monitoring shading and power generation efficiency in BIPV systems. The method includes processing the BIPV system and establishing operational status correlations between BIPV strings; detecting shading of target BIPV strings based on real-time operational status to identify strings affected by shading; establishing a shading-free power state model based on normal operating status data; reconstructing the theoretical shading-free power of the target strings under current environmental conditions through reference state verification, multi-model prediction fusion, and state correction; and calculating the power generation loss caused by shading and the actual power generation efficiency based on the difference between the actual output power and the theoretical shading-free power of the target strings. This invention utilizes existing operational data from the BIPV system to achieve shading identification and power generation capacity assessment, improving the accuracy and reliability of BIPV system operational status analysis in complex building environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of building photovoltaic technology, and in particular to a method for analyzing and monitoring BIPV shading and power generation efficiency. Background Technology

[0002] With the development of Building Integrated Photovoltaic (BIPV) technology, photovoltaic modules are gradually shifting from traditional centralized installation to integrated applications on building roofs, curtain walls, and building envelopes. While BIPV systems achieve energy conservation and renewable energy utilization in buildings, their operating environment is more complex than that of traditional ground-mounted photovoltaic systems because the installation location of photovoltaic modules is affected by factors such as building structure, surrounding environment, and spatial layout.

[0003] In actual operation, BIPV photovoltaic modules may be partially shaded by surrounding buildings, trees, equipment, and the building's own structure, reducing the solar radiation received by some modules and consequently decreasing their output power. Since BIPV systems typically consist of multiple photovoltaic modules and strings, local shading not only affects the output capacity of the corresponding modules but may also reduce the overall power generation performance of the photovoltaic system.

[0004] Currently, methods for analyzing the impact of shading on photovoltaic (PV) systems mainly include manual inspection, image recognition, and methods based on operational parameters. Manual inspection requires regular on-site checks by staff, which suffers from long inspection cycles, insufficient real-time performance, and difficulty adapting to large-scale building-integrated photovoltaic (BIPV) systems. Image recognition methods typically require additional camera equipment and are affected by lighting conditions, shooting angles, and environmental changes, resulting in high equipment investment and maintenance costs. Methods based on operational parameters primarily collect electrical parameters during BIPV system operation to identify abnormal operating states. However, due to the complex environment of BIPV applications and the significant uncertainty in the irradiance changes experienced by PV modules, existing methods struggle to accurately distinguish between power fluctuations caused by normal environmental changes and power reductions due to localized shading.

[0005] In addition, existing photovoltaic system operation monitoring methods usually focus on fault detection or overall power generation analysis, and lack effective means to assess the degree of power generation reduction caused by local shading in the building environment, making it difficult to further obtain the difference between the actual power generation capacity and the theoretical power generation capacity of photovoltaic modules under the influence of shading.

[0006] Therefore, how to accurately identify the impact of local shading on BIPV systems and effectively analyze the power generation loss caused by shading without adding a large number of additional detection devices has become an urgent problem to be solved in the operation and management of building photovoltaic integrated systems. Summary of the Invention

[0007] To address the aforementioned technical issues, this invention provides a method for analyzing and monitoring BIPV shading and power generation efficiency. This invention utilizes existing operational data from the BIPV system to achieve shading identification and power generation capacity assessment, thereby improving the accuracy and reliability of photovoltaic system operational status analysis in complex building environments.

[0008] This invention provides a method for analyzing and monitoring shading and power generation efficiency in BIPV systems, applicable to BIPV systems comprising multiple photovoltaic strings, and includes the following steps: Based on the shading events of the BIPV system, the target string affected by shading and the reference string associated with the target string are determined. The reference state of the reference string is obtained, and the reference state is associated with the reference state based on the pre-built verification relationship set to obtain the initial confidence level. Based on the initial confidence level, multi-model estimation is performed on the reference state to obtain the model probability and fusion state; The reference state is corrected based on the fusion state, and the verification relationship set is updated based on the model probability; the corrected reference state is then corrected and restored according to the updated verification relationship set to obtain the corrected reference state and the updated confidence level. Based on the correction reference state and the update confidence, the multi-model estimation, the model probability-based verification relationship update, and the error correction recovery are iteratively executed until the preset convergence condition is met, and the collaborative convergence result is obtained. Based on the collaborative convergence results, the shadowless power of the target string is reconstructed, and the shading loss and power generation efficiency are determined according to the shadowless power and the actual power of the target string.

[0009] Preferably, the construction of the verification relation set includes the following steps: Based on the historical operating data of the BIPV system, multiple shadowless operating states are defined; For each of the aforementioned shadowless operating states, a corresponding shadowless power state model is established; Based on the state association between multiple photovoltaic strings under each of the aforementioned shadowless operating states, a sub-verification relationship corresponding to each of the aforementioned shadowless power state models is established, and the aforementioned sub-verification relationships are combined into the verification relationship set.

[0010] Preferably, the fusion verification includes the following steps: The current verification relationship is obtained by fusing the sub-verification relationships in the verification relationship set based on the model probability of the previous monitoring time. The reference state is verified based on the current verification relationship to obtain the verification deviation; The initial confidence level of each reference string is determined based on the degree of influence of each reference string on the verification deviation.

[0011] Preferably, the multi-model estimation includes the following steps: Based on the initial confidence level of each reference string, the observation weight corresponding to each reference string is determined; The reference state is weighted based on the observation weights, and the weighted reference state is input into each of the shadowless power state models to obtain the prediction results of each of the shadowless power state models. Based on the state prediction deviation between each prediction result and the corresponding reference state, the model probability of each shadowless power state model is determined. The fusion state is obtained by fusing the prediction results based on the probabilistic fusion of the model.

[0012] Preferably, the step of correcting the reference state and updating the verification relationship set based on the fusion state includes: Based on the residual between the state variables corresponding to the fused state and the reference state, the reference state is corrected to obtain the corrected reference state; Based on the model probability, the weights and verification parameters of each sub-verification relationship in the verification relationship set are adjusted to obtain the updated verification relationship set; The corrected reference state is constrained and modified based on the updated set of verification relationships to obtain the modified reference state.

[0013] Preferably, the error correction and recovery includes: Based on the updated verification relationship set, the corrected reference state is associated with a consistency verification to determine the abnormal reference group string; The states corresponding to the abnormal reference strings are marked, and the states are restored based on the reference states corresponding to the non-abnormal reference strings and the updated verification relationship set. The reliability of the reference string is updated based on the verification deviation between the restored reference state and the updated verification relationship set.

[0014] Preferably, the iterative execution includes: The corrected reference state and updated confidence are fed back to the multi-model estimation process to obtain the updated model probability and fusion state. Based on the updated model probabilities and fusion states, the reference state correction, verification relationship update and error correction recovery processes are repeated. When the changes in model probability and update confidence in adjacent iterations meet the preset convergence threshold, the collaborative convergence result is determined.

[0015] Preferably, the reconstructing of the shadowless power of the target string includes: Based on the collaborative convergence results, obtain the converged model probability, confidence level, and correction reference state; Based on the updated confidence after convergence, the corrected reference states corresponding to multiple reference groups are weighted and fused, and then input into each of the shadowless power state models to obtain the shadowless power prediction results. The shadowless power prediction results are fused based on the model probability, and the fused results are corrected based on the historical shadowless operation status to obtain the shadowless power of the target string.

[0016] Compared with the prior art, the present invention has at least the following beneficial effects: (1) By analyzing the operating data of the BIPV system, the present invention realizes the automatic identification of local shadow occlusion in the building photovoltaic integration scenario. It does not require additional image acquisition equipment or occlusion detection sensors, which reduces the system transformation cost and improves the convenience of practical application.

[0017] (2) This invention addresses the problem of local shading caused by changes in building structure and surrounding environment in BIPV systems. By establishing a photovoltaic string operation status analysis mechanism, it improves the accuracy of shading judgment under complex environmental conditions and reduces the possibility of misjudgment caused by normal environmental changes.

[0018] (3) The present invention can analyze the theoretical power generation capacity under shading conditions based on the actual operating status of the target photovoltaic string, and further calculate the power generation loss and actual operating efficiency caused by shading, thus realizing a comprehensive analysis from shading detection to power generation impact assessment. Attached Figure Description

[0019] Figure 1 This is a diagram showing the overall architecture of the BIPV shading and power generation efficiency analysis and monitoring system of the present invention. Figure 2 This is a schematic diagram of shadow occlusion in a BIPV building environment according to the present invention; Figure 3 This is a schematic diagram of the spatial distribution and correlation of photovoltaic strings in this invention; Figure 4 This is a flowchart of the shadow occlusion detection process of the present invention; Figure 5 This is a structural diagram of the shadowless power state model of the present invention. Detailed Implementation

[0020] This invention provides a method for analyzing and monitoring shading and power generation efficiency in Building Integrated Photovoltaics (BIPV) systems. In BIPV scenarios, it identifies localized shading caused by building structure, surrounding environment, and dynamic shading factors. Based on the operational correlation between unshaded PV strings, it reconstructs the shading-free state of target PV strings affected by shading, thereby obtaining the theoretical power generation capacity of the target strings under shading conditions. Furthermore, it calculates the power generation loss and actual operating efficiency caused by shading.

[0021] The BIPV system in this embodiment includes multiple photovoltaic strings, a string monitoring unit, an environmental parameter acquisition unit, a data processing unit, and an analysis and calculation unit. Each photovoltaic module forms multiple photovoltaic strings according to a preset connection method, and each photovoltaic string includes multiple photovoltaic modules connected in series. Each photovoltaic string is equipped with an independent data acquisition module for real-time acquisition of the corresponding photovoltaic string's output voltage, current, and power data. The environmental parameter acquisition unit is used to collect parameters such as solar irradiance, module temperature, ambient temperature, humidity, and wind speed in the current photovoltaic system operating environment. The data processing unit is used to perform time synchronization, outlier removal, and data normalization on the collected data. The analysis and calculation unit is used to perform shading detection, state estimation, model fusion, anomaly recovery, and power generation efficiency analysis.

[0022] like Figure 1 The diagram shown illustrates the overall architecture of the BIPV shading and power generation efficiency analysis and monitoring system of this invention. The system includes a photovoltaic string layer, a data acquisition layer, a data processing layer, and an analysis and calculation layer. The photovoltaic string layer consists of multiple photovoltaic strings, each configured with an independent string monitoring unit to collect voltage, current, and power information for the corresponding string. The data acquisition layer obtains operating environment parameters such as solar irradiance, module temperature, and ambient temperature through an environmental parameter acquisition unit. The data processing unit performs synchronous processing and anomaly correction on the collected data. The analysis and calculation unit further performs shading detection, state reconstruction, and power generation efficiency analysis.

[0023] In this embodiment, the i-th photovoltaic string is taken as the analysis object, and its operating state at monitoring time t is represented as a state vector:

[0024] in, Indicates the first The actual output power of a photovoltaic string at time t, in kW; This indicates the output voltage, in volts (V). This indicates the output current, expressed in amperes (A). This indicates the current solar irradiance, expressed in W / m². 2; This indicates the temperature of the photovoltaic module, in °C.

[0025] In actual operation, although different photovoltaic strings in a BIPV system are installed in different locations, they are usually under the same environmental conditions in the same area. Therefore, the trend of solar irradiance variation among each photovoltaic string is highly consistent. Under unobstructed conditions, the output power changes among different strings are significantly correlated. Therefore, this embodiment first establishes a correlation model of normal operating status among multiple photovoltaic strings based on historical operating data, and uses this correlation model to identify abnormal power drop situations.

[0026] like Figure 2 The diagram illustrates shading in a BIPV building environment, as presented in this invention. During the application of building-integrated photovoltaics (BIPV), photovoltaic modules at different locations may experience varying degrees of shading due to the building structure, surrounding buildings, trees, and other dynamic environmental factors. The photovoltaic strings affected by shading receive reduced solar radiation, resulting in actual output power lower than the theoretical output power under normal operating conditions. Therefore, this invention identifies the impact of localized shading by analyzing the operational relationships between photovoltaic strings.

[0027] Specifically, the historical operating data of the BIPV system is first acquired. This historical operating data includes at least 30 consecutive days of data; in one embodiment, the acquisition period is set to 1 minute, resulting in 43,200 sets of historical sampling data. For each sampling moment, the operating status information of all photovoltaic strings is recorded.

[0028] To reduce the impact of equipment measurement errors, the collected data is preprocessed.

[0029] First, anomaly detection is performed on the acquired power data. When anomalies are detected... or If so, it is considered that the data point may be affected by abnormal collection. A preset change threshold is used. In this embodiment, Set to 30% of the rated power.

[0030] For outlier data points, interpolation using adjacent time windows is employed for correction.

[0031] After preprocessing, a standardized historical operation database is formed.

[0032] Furthermore, based on data from historical unshaded operation phases, a state correlation relationship is established between each photovoltaic string.

[0033] like Figure 3The diagram illustrates the spatial distribution and correlation of photovoltaic strings according to the present invention. Since multiple photovoltaic strings in the same BIPV system are typically affected by the same solar motion patterns and environmental changes, there are certain operational correlations between different strings. This embodiment constructs a photovoltaic string correlation network and correlation matrix by calculating the correlation between strings under historical unshaded operating conditions, and selects reference strings with a high degree of matching to the target string state based on the correlation strength.

[0034] Because all photovoltaic strings within the same BIPV system are affected by the same solar motion, under normal operating conditions, the first... The string and the first The following conditions must be met between the strings:

[0035] in, Indicates the first The string and the first The state mapping relationship between each string group.

[0036] In one embodiment, the correlation coefficient method is used to determine the degree of association between strings.

[0037] The specific calculation formula is as follows:

[0038] in: Indicates the first The string and the first Power correlation coefficient between each string group; Indicates the number of historical samples; and These represent the historical average power values ​​of the two strings, respectively.

[0039] Construct a string association matrix based on the correlation coefficients between all strings:

[0040] in This indicates the total number of photovoltaic strings in a BIPV system.

[0041] During real-time monitoring, when a photovoltaic string is shaded, its actual output power will be lower than its theoretical output power under the current environmental conditions. Therefore, this embodiment predicts the state of each photovoltaic string based on the string association relationship.

[0042] like Figure 4The diagram shown is a flowchart of the shadowing effect analysis based on reference state verification and multi-model fusion according to the present invention. Specifically, it includes steps such as determining the target string and reference string, evaluating the reliability of the reference state, multi-model estimation, updating the verification relationship and error correction recovery, and reconstructing the shadowless power. S1: Based on the shading events of the BIPV system, determine the target string affected by shading and the reference string that has a state association with the target string, obtain the reference state of the reference string, and perform association verification on the reference state based on the pre-built verification relationship set to obtain the initial confidence level. S2: Based on the initial confidence level, perform multi-model estimation on the reference state to obtain the model probability and fusion state; S3: Correct the reference state based on the fusion state, and update the verification relationship set based on the model probability; perform error correction and recovery on the corrected reference state according to the updated verification relationship set to obtain the corrected reference state and updated confidence. S4: Based on the correction reference state and the update confidence, iteratively execute the multi-model estimation, the verification relationship update based on model probability, and the error correction recovery until the preset convergence condition is met, and obtain the collaborative convergence result; S5: Reconstruct the shadowless power of the target string based on the cooperative convergence result, and determine the shading loss and power generation efficiency according to the shadowless power and the actual power of the target string.

[0043] For the For each string, calculate the predicted power based on the states of other strings that are related to it:

[0044] in Indicates the first String pairs The prediction contribution weights of the strings satisfy the following conditions: and Further calculations were performed to determine the deviation between the actual power and the predicted power:

[0045] when When, then determine the first The string exists in an abnormal state, among which This is the threshold for shading detection.

[0046] In one embodiment, based on actual BIPV system operating experience, The value is set to 0.15, meaning that if the actual power is more than 15% lower than the predicted power, the string is considered to be affected by shading.

[0047] Meanwhile, to avoid misjudgment caused by short-term irradiance fluctuations, this embodiment further sets a duration judgment condition that the continuous holding time exceeds [a certain value]. If so, a shading event is confirmed to have occurred, in which Set to 10 minutes.

[0048] For example, in a certain BIPV system, a total of 48 photovoltaic strings are set up, each with a rated power of 5kW.

[0049] At a certain moment, solar irradiance: Component temperature: 42℃.

[0050] The theoretical output power of the 12th string is calculated based on the associated string groups: .

[0051] The actual measured output power: Since 31.25% is greater than the threshold of 15%, it is determined that the 12th string is affected by shadow occlusion, and the 12th string is identified as the target string.

[0052] After determining the target string, it is necessary to further select a reference string that has a state association with the target string and is not affected by occlusion.

[0053] Specifically, based on the historical association matrix It performs association evaluation on strings other than the target string.

[0054] For candidate reference strings Its evaluation indicators are expressed as follows: .

[0055] in This represents the comprehensive evaluation value of the candidate string. This indicates the current level of trustworthiness.

[0056] according to Sort by largest to smallest, then select the first... Each string is used as a reference string.

[0057] In this embodiment, That is, select the three normal strings with the highest degree of correlation as reference strings.

[0058] For example, for the 12th string that is shaded:

[0059] Then it is determined that: As a reference set of strings.

[0060] After determining the set of reference strings that have state associations with the target string, in order to enable the reference strings to accurately represent the shadowless operation state of the target string under different lighting conditions, temperature conditions, and building shadow changes, this embodiment further establishes a verification relationship set based on the historical operation data of the BIPV system, and uses the verification relationship set to perform credibility evaluation, multi-model estimation, state correction, and error correction and recovery on the reference string state.

[0061] Specifically, the historical operating data of the BIPV system is first processed to obtain unobstructed operating samples for constructing the verification relation set.

[0062] Because BIPV systems may experience issues such as localized shading, component performance degradation, data acquisition anomalies, and inverter operation fluctuations during actual operation, historical operating data cannot be directly used for model construction. This embodiment first filters historical operating data based on the operational consistency between photovoltaic strings, changes in environmental parameters, and equipment operating status.

[0063] The historical operating data includes at least: the output power, voltage, and current of each photovoltaic string at multiple historical moments; the solar irradiance at the corresponding moment; the module temperature; the ambient temperature; and time information.

[0064] In one embodiment, by acquiring 180 consecutive days of historical operating data and setting the sampling period to 1 minute, each photovoltaic string can obtain approximately 259,200 sets of operating status data.

[0065] For any sampling moment in the historical operating data, determine whether the output power change of each photovoltaic string conforms to the characteristics of normal unshaded operation. When the power change trends among multiple photovoltaic strings are consistent and there is no significant power drop in a single string, the data corresponding to that moment is taken as the unshaded operation sample.

[0066] Meanwhile, to avoid equipment malfunctions affecting the verification relationship set, this embodiment further filters health status based on the average power generation capacity of each photovoltaic string during historical operating cycles. If the long-term output capacity of a photovoltaic string is significantly lower than the average level of strings of the same type, then that string will not be used as a valid reference object in the verification relationship establishment process.

[0067] After the above screening, a sample set of unobstructed operating states is obtained, and a shadowless power state model corresponding to multiple typical operating states is established based on the unobstructed operating state samples.

[0068] like Figure 5The diagram shown illustrates the structure of the shadowless power state model of this invention. In this embodiment, based on the environmental parameter variation characteristics in historical unobstructed operation data, the operating state is divided into multiple typical states, and corresponding shadowless power state models are established for each. The model takes the current environmental parameters and the operating state of a reference string as input, and obtains the mapping relationship between the environmental state and the theoretical output power of the target string through training with historical data, thereby obtaining the theoretical shadowless power of the target string under the current environmental conditions.

[0069] Since the output characteristics of photovoltaic modules in BIPV systems vary under different irradiance, temperature and time stages, this embodiment does not use a single fixed model, but instead divides multiple shadowless operating states based on historical operating data.

[0070] Specifically, based on solar irradiance, module temperature, and historical power variation trends, the unshaded operation samples are divided into multiple operating states. For example, they can be divided into high irradiance operating state, medium irradiance operating state, and low irradiance operating state.

[0071] For each shadowless running state, a corresponding shadowless power state model is established.

[0072] In this embodiment, the shadowless power state model is used to describe the theoretical output state of each photovoltaic string when it is not affected by shadows under the current environmental conditions. Its input parameters include the current environmental parameters and the operating state of the reference string, and the output parameter is the theoretical shadowless operating state corresponding to the target string.

[0073] The shadowless power state model can be a state prediction model trained based on historical data, such as a multivariate regression model, a support vector regression model, a neural network model, or other models that can establish a mapping relationship between environmental parameters and photovoltaic output power.

[0074] In one embodiment, for the first The state model of a shadowless running state is as follows:

[0075] in Indicates the first The shadowless power predicted under various operating conditions; Indicates the state of the reference string; Indicates solar irradiance; Indicates component temperature; Indicates the first Power prediction models corresponding to various operating states.

[0076] The model parameters were obtained through training with historical unobstructed operating data. During model training, the reference string operating data under unobstructed operating conditions was used as input, and the actual output power of the target string was used as the training output. The mapping relationship between the input state and the normal power of the target string was obtained by fitting historical data.

[0077] Furthermore, in order to describe the correlation between multiple photovoltaic strings under different operating conditions, this embodiment establishes a corresponding sub-verification relationship for each unshaded operating condition.

[0078] The sub-verification relationship is used to describe: Under a given environmental condition, when the reference string is in normal operation, the target string and other associated strings should satisfy the state constraints.

[0079] Specifically, for the first Establish an unobstructed operating state:

[0080] in Indicates the first The state relationships between each photovoltaic string under different conditions. This indicates the threshold for anomaly detection in this state. This represents the weight corresponding to the relationship between each state.

[0081] The state correlation is determined based on the synchronous change pattern among multiple photovoltaic strings in historical unshaded operation data; the anomaly judgment threshold is determined based on the prediction error range under historical normal conditions; and the weight is determined based on the contribution of different reference strings to the state prediction of the target string.

[0082] All sub-verification relationships corresponding to all running states together form a verification relationship set:

[0083] in, This indicates the number of running states without shadows.

[0084] During real-time monitoring, the current operating status is determined based on the currently collected solar irradiance, component temperature, and time information, and the corresponding sub-verification relationship is selected from the verification relationship set as the current verification basis.

[0085] Since the operating environment of the BIPV system is continuously changing, the current operating state may be affected by multiple historical operating states at the same time. Therefore, this embodiment further integrates multiple sub-verification relationships based on the model probability obtained at the previous monitoring time to form a comprehensive verification relationship at the current time.

[0086] Specifically, let the previous time step be... The model probability corresponding to each unshaded state model is: The current verification relationship is obtained by weighted fusion of multiple sub-verification relationships according to the model probability.

[0087] By using the above methods, we can avoid the problem of inaccurate predictions from a single-state model due to sudden changes in the operating state.

[0088] Furthermore, the current verification relationship is used to perform a consistency check on the reference string state.

[0089] For reference string Get its current running status ; The predicted state is obtained based on the current verification relationship. Calculate the residual between the actual state and the predicted state of the reference string:

[0090] Among them, residual The smaller the value, the more the reference string conforms to the current verification relationship, and the more reliable its operating status.

[0091] The initial confidence level of the reference string is calculated based on the residual:

[0092] in Indicates the first The initial confidence level of each reference string; This is a reliability adjustment parameter, which is determined based on the statistical results of the prediction error of the reference string under historical unobstructed operation conditions.

[0093] When the reference string's operating status is highly consistent with its historical normal status, its residual is small, and its confidence level is close to 1. When the reference string is affected by unknown occlusion or equipment malfunction, its residual increases, and its confidence level decreases.

[0094] Furthermore, a multi-model estimation process is performed based on the obtained initial confidence level.

[0095] Since BIPV systems may be under different irradiance and temperature conditions, different shadowless state models have different degrees of adaptability to the current state. Therefore, this embodiment calls multiple shadowless power state models for prediction at the same time, and dynamically determines the model probability based on the prediction results of each model.

[0096] First, the observation weights of different reference strings are determined based on the initial confidence of the reference strings.

[0097] Specifically, reference strings with higher credibility have higher weights, while reference strings with lower credibility have reduced participation.

[0098] Subsequently, based on the weighted state input of each reference group string, multiple shadowless running state models are input to obtain the predicted state corresponding to each model.

[0099] Let the first The state model outputs the predicted state as follows: The model fitness level is then determined based on the prediction deviation between the predicted state of each model and the current reference state.

[0100] The smaller the prediction bias, the more the corresponding state model matches the current operating environment, and the higher its model probability.

[0101] The model probability update process is as follows:

[0102] in This represents the prior probability obtained based on the model probability and state transition relationship from the previous time step; Indicates the first The degree of prediction matching corresponding to the model.

[0103] Through the above calculations, the dynamic probabilities corresponding to multiple shadowless running state models are obtained.

[0104] Furthermore, the outputs of multiple models are fused based on the model probabilities:

[0105] in This indicates a fusion state.

[0106] The fusion state can integrate the prediction results of different operating state models, which can reduce the prediction error caused by environmental changes compared with the single model prediction method.

[0107] After obtaining the fusion state, this embodiment further uses the fusion state to correct the state of the reference string.

[0108] Specifically, the state deviation between the fusion state and the current reference state is calculated, and the reference state is corrected according to the preset correction coefficient, so that the reference string state gradually approaches the actual operating state under the current environmental conditions.

[0109] After completing the reference state correction, the influence weights of each sub-verification relation in the verification relation set are adjusted according to the updated model probability.

[0110] Specifically, when a certain shadowless state model has a high model probability during continuous monitoring, the weight of the corresponding sub-verification relationship of that state is increased; when a certain state model has a low long-term matching degree, the influence of its corresponding sub-verification relationship is reduced.

[0111] By dynamically updating the verification relationship set, the verification relationship can adapt to seasonal changes, component performance changes, and environmental changes that occur during the long-term operation of the BIPV system.

[0112] Furthermore, reference string error correction and recovery are performed based on the updated check set.

[0113] First, perform consistency checks on all reference strings based on the updated check relationships.

[0114] When the deviation between the state of a reference string and the verification relationship exceeds the corresponding running state threshold, the reference string is considered to be possibly abnormal.

[0115] For abnormal reference strings, they do not directly participate in the prediction of the target string's shadowless state, but instead utilize the states of other normal reference strings and the updated check relationships for recovery.

[0116] The confidence level is recalculated after the reference state is restored.

[0117] If the restored credibility meets the preset requirements, the restored result will be used as the new reference state; if it is still lower than the preset threshold, other candidate reference strings will be selected and added to the reference string set.

[0118] After one round of state estimation, verification relationship update, and error correction recovery, the obtained corrected reference state and updated confidence are fed back to the next round of multi-model estimation.

[0119] The process is executed cyclically: reference state fusion → model probability update → state correction → verification relationship adjustment → anomaly detection → state recovery, forming a closed-loop iterative process.

[0120] During the iteration process, the change between the results of two adjacent iterations is calculated in real time.

[0121] When both the change in model probability and the change in the confidence level of the reference string meet the preset convergence conditions, the current state estimation result is considered to be stable.

[0122] The specific judgment method is as follows: ,and in Indicates the current iteration number. This represents the probability convergence threshold of the model. This represents the confidence convergence threshold.

[0123] In one embodiment, the maximum number of iterations is set to 10. When the maximum number of iterations is reached or the convergence condition is met, the iteration ends and the final co-convergence result is obtained.

[0124] Based on the final obtained collaborative convergence results, the target string is reconstructed without shadows.

[0125] First, based on the confidence level of the converged reference strings, the correction states corresponding to multiple reference strings are fused:

[0126] Subsequently, the fused reference state is input into multiple shadowless power state models to obtain shadowless power prediction results for different models. .

[0127] Then, the probabilities of the finally converged model are fused:

[0128] The theoretical shadowless power of the target string under the current environmental conditions is obtained.

[0129] Finally, based on the actual output power of the target string Calculate the power generation loss caused by shading. ; and shading loss rate ; Further calculate the actual power generation efficiency of the target string under the current state. .

[0130] For example, in a BIPV system containing 48 photovoltaic strings, the 12th string is detected to be shaded. After reference string reliability evaluation, multi-model probabilistic fusion, and iterative correction, the theoretical unshaded power of the target string is obtained as 4.6kW, while the actual output power is 3.3kW. Therefore: The power generation loss is 4.6 - 3.3 = 1.3 kW; the shading loss rate is 1.3 / 4.6 × 100% = 28.26%; the corresponding actual power generation efficiency is 3.3 / 4.6 × 100% = 71.74%.

[0131] Therefore, this embodiment can accurately analyze the degree of shadow shading impact without the need for additional shading detection equipment, using only the existing operating data of the BIPV system. Furthermore, through dynamic reference string verification, multi-model fusion, and state recovery mechanisms, it improves the accuracy and reliability of photovoltaic power generation efficiency assessment in complex building environments.

Claims

1. A method for analyzing and monitoring shadow blocking and power generation efficiency of a BIPV, applied to a BIPV system including a plurality of photovoltaic strings, characterized in that, Includes the following steps: Based on the shading events of the BIPV system, the target string affected by shading and the reference string associated with the target string are determined. The reference state of the reference string is obtained, and the reference state is associated with the reference state based on the pre-built verification relationship set to obtain the initial confidence level. Based on the initial confidence level, multi-model estimation is performed on the reference state to obtain the model probability and fusion state; The reference state is corrected based on the fusion state, and the verification relationship set is updated based on the model probability; the corrected reference state is then corrected and restored according to the updated verification relationship set to obtain the corrected reference state and the updated confidence level. Based on the correction reference state and the update confidence, the multi-model estimation, the model probability-based verification relationship update, and the error correction recovery are iteratively executed until the preset convergence condition is met, and the collaborative convergence result is obtained. Based on the collaborative convergence results, the shadowless power of the target string is reconstructed, and the shading loss and power generation efficiency are determined according to the shadowless power and the actual power of the target string. 2.The method of claim 1, wherein, The construction of the verification relationship set includes the following steps: Based on the historical operating data of the BIPV system, multiple shadowless operating states are defined; For each of the aforementioned shadowless operating states, a corresponding shadowless power state model is established; Based on the state association between multiple photovoltaic strings under each of the aforementioned shadowless operating states, a sub-verification relationship corresponding to each of the aforementioned shadowless power state models is established, and the aforementioned sub-verification relationships are combined into the verification relationship set.

3. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 2, characterized in that, The fusion verification includes the following steps: The current verification relationship is obtained by fusing the sub-verification relationships in the verification relationship set based on the model probability of the previous monitoring time. The reference state is verified based on the current verification relationship to obtain the verification deviation; The initial confidence level of each reference string is determined based on the degree of influence of each reference string on the verification deviation.

4. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 3, characterized in that, The multi-model estimation includes the following steps: Based on the initial confidence level of each reference string, the observation weight corresponding to each reference string is determined; The reference state is weighted based on the observation weights, and the weighted reference state is input into each of the shadowless power state models to obtain the prediction results of each of the shadowless power state models. Based on the state prediction deviation between each prediction result and the corresponding reference state, the model probability of each shadowless power state model is determined. The fusion state is obtained by probabilistically fusing the prediction results of the model.

5. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 4, characterized in that, The step of correcting the reference state and updating the verification relationship set based on the fusion state includes: Based on the residual between the state variables corresponding to the fused state and the reference state, the reference state is corrected to obtain the corrected reference state; Based on the model probability, the weights and verification parameters of each sub-verification relationship in the verification relationship set are adjusted to obtain the updated verification relationship set; The corrected reference state is constrained and modified based on the updated set of verification relationships to obtain the modified reference state.

6. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 5, characterized in that, The error correction and recovery includes: Based on the updated verification relationship set, the corrected reference state is associated with a consistency verification to determine the abnormal reference group string; The states corresponding to the abnormal reference strings are marked, and the states are restored based on the reference states corresponding to the non-abnormal reference strings and the updated verification relationship set. The reliability of the reference string is updated based on the verification deviation between the restored reference state and the updated verification relationship set.

7. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 6, characterized in that, The iterative execution includes: The corrected reference state and updated confidence are fed back to the multi-model estimation process to obtain the updated model probability and fusion state. Based on the updated model probabilities and fusion states, the reference state correction, verification relationship update and error correction recovery processes are repeated. When the changes in model probability and update confidence in adjacent iterations meet the preset convergence threshold, the collaborative convergence result is determined.

8. The BIPV shading and power generation efficiency analysis and monitoring method according to claim 7, characterized in that, The shadowless power for reconstructing the target string includes: Based on the collaborative convergence results, obtain the converged model probability, confidence level, and correction reference state; Based on the updated confidence after convergence, the corrected reference states corresponding to multiple reference groups are weighted and fused, and then input into each of the shadowless power state models to obtain the shadowless power prediction results. The shadowless power prediction results are fused based on the model probability, and the fused results are corrected based on the historical shadowless operation status to obtain the shadowless power of the target string.