A method and system for optimizing state adjustment of a photovoltaic module

Abstract

The present invention discloses a state adjustment optimization method and system for a photoelectric module, and belongs to the field of data processing technology. The present invention has many advantages in the state adjustment of the photoelectric module: when acquiring knowledge and building relationships, the operation scenario, adjustment decision, and performance feedback semantics are integrated to build a relationship spectrum, which can fully explore the complex relationship between various factors and avoid the limitations of traditional methods; based on the temporal commonality metric value, the decision knowledge is optimized to adapt to the operation rhythm, avoid adjustment lag or over-adjustment, and improve stability and performance; based on the optimization of entity semantic transmission characteristics, the state adjustment decision knowledge and device performance feedback semantics can be accurately optimized to improve accuracy and reliability; the synchronously optimized state adjustment decision knowledge is used to determine the current state adjustment strategy of the target photoelectric module, and the synchronously optimized device performance feedback semantics is used to verify the state adjustment, which can improve maintainability and stability, extend the service life of the photoelectric module, and enhance applicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to a state adjustment optimization method and system for a photoelectric module. Background Art

[0002] In the application field of optoelectronic modules, traditional state adjustment methods face many technical problems. With the widespread application of optoelectronic technology in many fields such as modern communications and data centers, the performance and stability requirements of optoelectronic modules are increasing. However, traditional state adjustment methods are often based on experience or local optimization methods, lacking comprehensive consideration of the overall operating environment and internal relationships of optoelectronic modules.

[0003] On the one hand, traditional methods rarely consider the impact of multiple operating scenario factors when dealing with the state adjustment of photovoltaic modules. On the other hand, traditional state adjustment decisions lack timing matching with actual operation. In addition, traditional methods do not adequately explore the intrinsic relationship between state adjustment decisions and device performance feedback. It can be seen that traditional photovoltaic module state adjustment technology is difficult to meet the increasingly complex application requirements. Summary of the invention

[0004] The present invention provides a method and system for optimizing the state adjustment of a photovoltaic module, which can solve or partially solve the technical problems involved in the above-mentioned background technology.

[0005] The present invention provides a state regulation optimization method for a photoelectric module, which is applied to a state regulation optimization system. The method comprises: obtaining operation scenario description knowledge of a reference photoelectric module, state regulation decision knowledge of a plurality of attention dimensions, and X device performance feedback semantics; wherein the state regulation decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics are interconnected to generate a knowledge feature relationship spectrum, each of the plurality of attention dimensions corresponds to one or more state regulation decision knowledge, the state regulation decision knowledge is represented by a terminal relationship entity of the knowledge feature relationship spectrum, the device performance feedback semantics is represented by a linkage relationship entity of the knowledge feature relationship spectrum, and X is a positive integer; based on the reference photoelectric module The temporal commonality measurement value of the operation scenario description knowledge and the state adjustment decision knowledge of the several attention dimensions is used to synchronously optimize the state adjustment decision knowledge of the several attention dimensions; based on the entity semantic transmission characteristics between the terminal relationship entity and the linkage relationship entity, the state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics are synchronously optimized; wherein the state adjustment decision knowledge of the several attention dimensions after synchronous optimization is used to determine the current state adjustment strategy of the target optoelectronic module, the current state adjustment strategy includes at least one of a temperature adjustment strategy, an optical power adjustment strategy, a bias current adjustment strategy, and a signal modulation optimization strategy, and the X device performance feedback semantics after synchronous optimization are used for state adjustment verification.

[0006] The present invention provides a state adjustment optimization system, comprising at least one processor and a memory; the memory stores computer execution instructions; the at least one processor executes the computer execution instructions stored in the memory, so that the at least one processor executes the above method.

[0007] The present invention provides a readable storage medium, on which a program or instruction is stored, and the program or instruction implements the steps of the above method when executed by a processor.

[0008] The present invention has many advantages in the state regulation of photovoltaic modules: when acquiring knowledge and building relationships, the operation scenarios, regulation decisions, and performance feedback semantics are integrated to build a relationship spectrum, which can fully explore the complex relationships between various factors and avoid the limitations of traditional methods; based on the optimization of time series commonality metrics, decision knowledge can be adapted to the operation rhythm, avoid regulation lag or over-regulation, and improve stability and performance; based on the optimization of entity semantic transmission characteristics, the state regulation decision knowledge and device performance feedback semantics can be accurately optimized to improve accuracy and reliability; the synchronously optimized state regulation decision knowledge is used to determine the current state regulation strategy of the target photovoltaic module, and the synchronously optimized device performance feedback semantics is used to perform state regulation verification, which can improve maintainability and stability, extend the service life of the photovoltaic module and enhance applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 The present invention provides a flow chart of a method for optimizing the state adjustment of a photovoltaic module.

[0010] Figure 2 This is a structural schematic diagram of a state adjustment and optimization system provided by the present invention. DETAILED DESCRIPTION

[0011] Figure 1 A state regulation optimization method for a photovoltaic module is shown, which is applied to a state regulation optimization system. The method includes the following steps 202 to 206 .

[0012] Step 202: The state adjustment optimization system obtains the operating scenario description knowledge of the reference photovoltaic module, the state adjustment decision knowledge of several attention dimensions, and X device performance feedback semantics.

[0013] The state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics are interconnected to generate a knowledge feature relationship spectrum, each of the several attention dimensions corresponds to one or more state adjustment decision knowledge, the state adjustment decision knowledge is represented by a terminal relationship entity of the knowledge feature relationship spectrum, the device performance feedback semantics is represented by a linkage relationship entity of the knowledge feature relationship spectrum, and X is a positive integer.

[0014] It can be understood that step 202 involves the generation of a knowledge acquisition and knowledge feature relationship spectrum, and this step can be divided into the following four branches for explanation.

[0015] 1. Obtaining the operating scenario description knowledge of the reference photovoltaic module

[0016] The reference photoelectric module operation scenario description knowledge covers various environments and working conditions related to the operation of the photoelectric module. For example, the operation scenario may include the type of equipment where the photoelectric module is located (such as in optical fiber communication base station equipment or in server interconnection equipment in a data center), the ambient temperature range (such as -20°C to 60°C), the working optical fiber link length (such as short distance within 100 meters, medium distance 100-1000 meters or long distance more than 1000 meters), the transmission data rate requirement (such as 10Gbps, 100Gbps or higher), etc. These operation scenario description knowledge provides basic context information for subsequent state adjustment decisions.

[0017] 2. Acquiring state regulation decision knowledge in several attention dimensions

[0018] 2.1. Concept of Attention Dimension

[0019] The attention dimension is an abstract division of different aspects of attention in the state regulation of the optoelectronic module. Each attention dimension corresponds to one or more state regulation decision knowledge. For example, an attention dimension may be "optical power stability", and the corresponding state regulation decision knowledge may include automatic power control (APC) related knowledge, such as the feedback loop parameter settings of APC (e.g., the proportional coefficient is 0.5, the integral time constant is 0.1 seconds, etc.), the frequency of optical power monitoring (e.g., monitoring once every 10 microseconds), etc.

[0020] Another attention dimension can be "temperature sensitivity", and the corresponding state adjustment decision knowledge includes the control strategy of the thermoelectric cooler (TEC), such as the maximum cooling power of TEC (for example, 5 watts), the accuracy requirement of temperature control (for example, ±0.1℃), etc.

[0021] 2.2 Representation of State Adjustment Decision Knowledge - Terminal Relationship Entity

[0022] The state regulation decision knowledge is characterized by the terminal relation entity of the knowledge feature relation spectrum. Taking the attention dimension of "optical power stability" as an example, if an APC system based on a specific algorithm is adopted, the terminal relation entity may include the key parameters in the mathematical model describing the algorithm. For example, for an exemplary first-order linear feedback control algorithm, its terminal relation entity can be the proportional coefficient K_p (set K_p=0.8) in the control equation P_{out}(t)=K_p×(P_{ref}-P_{mon}(t))+P_{out}(t-δt), where P_{out}(t) is the output optical power at time t, P_{ref} is the reference optical power, P_{mon}(t) is the optical power monitored at time t, and δt is the time interval.

[0023] 3. Obtaining performance feedback semantics of X devices

[0024] 3.1. Meaning of device performance feedback semantics

[0025] The device performance feedback semantics reflects the actual working performance of each device in the optoelectronic module (such as lasers, detectors, amplifiers, etc.). For example, for lasers, the performance feedback semantics may include the actual output power of the laser (for example, 3dBm), the stability of the output wavelength (for example, at a wavelength of 1550nm, the wavelength drift range is ±0.05nm), the change of the threshold current (for example, the initial threshold current is 10mA, and the current threshold current is 10.2mA), etc. For detectors, it may be the detector's responsivity (for example, 0.8A / W), dark current (such as 1nA), etc.

[0026] 3.2 Representation-Linkage Relationship Entities

[0027] The device performance feedback semantics is represented by the linkage relationship entity of the knowledge feature relationship spectrum. Taking the laser as an example, its output power performance feedback semantics is associated with the state adjustment decision knowledge under the attention dimension of optical power stability in the knowledge feature relationship spectrum. If the laser output power is lower than expected (the expectation is set to 5dBm), then the linkage relationship entity can trigger the need to adjust the state adjustment decision knowledge under the attention dimension of optical power stability (such as the relevant parameters of APC).

[0028] 4. Generating knowledge feature relationship spectrum

[0029] The state regulation decision knowledge of several attention dimensions and the X device performance feedback semantics are interconnected to generate a knowledge feature relationship spectrum. This relationship spectrum constructs a complex network structure, in which each node is either a terminal relationship entity (representing the state regulation decision knowledge) or a linkage relationship entity (representing the device performance feedback semantics), and the edge represents the association relationship between them. For example, under the attention dimension of temperature regulation, there is an association between the control strategy of TEC (terminal relationship entity) and the threshold current change of the laser (linkage relationship entity). Because temperature changes affect the threshold current of the laser, and the control strategy of TEC aims to stabilize the temperature and indirectly affect the threshold current of the laser. Through this relationship spectrum, the relationship between different factors in the optoelectronic module can be fully understood, providing a basis for subsequent optimization.

[0030] Step 204: The state adjustment optimization system synchronously optimizes the state adjustment decision knowledge of the several attention dimensions based on the temporal commonality measurement value of the operation scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge of the several attention dimensions.

[0031] Furthermore, the core idea of ​​step 204 is to optimize the state adjustment decision knowledge based on the timing commonality metric synchronously, and this process can be implemented based on the following two aspects.

[0032] 1. Calculation of time series commonality metrics

[0033] 1.1 Data Preparation

[0034] First, the operation scenario description knowledge of the reference photovoltaic module and the state adjustment decision knowledge of several attention dimensions need to be formatted. The time-related information in the operation scenario description knowledge (such as working hours, ambient temperature change patterns in different periods, etc.) and the time-related parameters in the state adjustment decision knowledge (such as the time step and monitoring cycle in the control algorithm, etc.) need to be extracted and sorted.

[0035] For example, the operating scenario description knowledge shows that the ambient temperature of the photovoltaic module changes periodically within a day, rising from 20°C to 25°C from 8:00 to 10:00 in the morning, and dropping from 25°C to 20°C from 4:00 to 6:00 in the afternoon. The optical power monitoring cycle in the state adjustment decision knowledge is once every 15 minutes, and the temperature control update cycle of TEC is once every 5 minutes.

[0036] 1.2. Calculation method example - based on time series similarity algorithm

[0037] The dynamic time warping (DTW) algorithm can be used to calculate the time series commonality metric. The DTW algorithm can nonlinearly stretch or compress two sequences on the time axis to find the best matching path between them. For the temperature time series (T={t_1, t_2, ..., t_n}) in the operation scenario description knowledge (where t_i is the temperature value at different times, and n represents the number of data points in the temperature time series) and the TEC control cycle time series (C={c_1, c_2, ..., c_m}) in the state regulation decision knowledge (where (c_j) is the different control cycle moments, and m represents the number of data points in the TEC control cycle time series), the DTW algorithm calculates a distance d, which is normalized (for example, mapping the distance d to the interval of 0-1) to obtain the time series commonality metric Sim. If Sim is close to 1, it means that the time series commonality is high; if Sim is close to 0, it means that the time series commonality is low.

[0038] For example, after calculation, the temporal commonality measurement value (Sim=0.6) between the operating scene description knowledge and the state adjustment decision knowledge under a certain attention dimension (such as temperature regulation related) is obtained.

[0039] 2. Synchronous optimization state adjustment decision knowledge

[0040] 2.1 Optimization principles

[0041] Adjust the state adjustment decision knowledge according to the temporal commonality metric. If the temporal commonality metric is low, it means that the current state adjustment decision knowledge does not match the operating scenario in time and needs to be optimized.

[0042] 2.2 Optimization Example

[0043] Taking optical power regulation as an example, if the timing commonality metric value is low, it may mean that the frequency of optical power monitoring does not match the changing pattern of ambient temperature or data transmission rate. For example, the original optical power monitoring frequency is once every 30 minutes, and according to the operating scenario description knowledge, the optical power of the optoelectronic module fluctuates greatly during the peak transmission period (the first 15 minutes of each hour), so the optical power monitoring frequency can be increased to once every 10 minutes. For other state adjustment decision knowledge related to optical power regulation, such as the feedback loop parameters of APC, it can also be adjusted according to the timing commonality metric value. For example, if the timing commonality metric value is low, appropriately increase the proportional coefficient (set from 0.6 to 0.8) to improve the response speed of optical power regulation.

[0044] Step 206: The state adjustment optimization system synchronously optimizes the state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics based on the entity semantic transmission characteristics between the terminal relationship entity and the linkage relationship entity.

[0045] The state adjustment decision knowledge of the several attention dimensions after synchronous optimization is used to determine the current state adjustment strategy of the target optoelectronic module, wherein the current state adjustment strategy includes at least one of a temperature adjustment strategy, an optical power adjustment strategy, a bias current adjustment strategy, and a signal modulation optimization strategy, and the X device performance feedback semantics after synchronous optimization are used for state adjustment verification.

[0046] Finally, step 206 is to synchronously optimize the state adjustment decision knowledge and device performance feedback semantics based on the entity semantic transfer feature. The specific technical ideas are as follows.

[0047] 1.1 Analysis of entity semantic transfer features

[0048] 1.1 Entity Semantic Transfer Path

[0049] In the knowledge feature relationship spectrum, there is a semantic transfer path between the terminal relationship entity (representing the state adjustment decision knowledge) and the linkage relationship entity (representing the device performance feedback semantics). For example, in the scenario of optical power adjustment, the APC control algorithm (terminal relationship entity) affects the output optical power of the laser (linkage relationship entity) by adjusting the driving current of the laser. In this process, there is a semantic transfer from the control algorithm to the output optical power of the laser, which involves semantic information such as the mathematical relationship between the driving current and the optical power (for example, the optical power is proportional to the driving current, and the proportionality coefficient is a specific value, such as 0.5mW / mA).

[0050] 1.2 Quantitative Analysis of Semantic Transfer - Based on Causal Model

[0051] A causal relationship model can be constructed to quantify entity semantic transfer features. For example, for the effect of bias current adjustment on the performance of photodetectors, an exemplary linear model (R=a×I+b) can be established, where R is the responsivity of the photodetector (here only as an introduction to the exemplary linear model, and does not conflict with f_{21} in the subsequent specific feature vector example. In other words, the subsequent f_{21} can also be understood as R), I is the bias current, and a and b are model parameters (set a=0.1, b=0.05). This model describes the semantic transfer features from bias current (terminal relationship entity related) to photodetector responsivity (linkage relationship entity related). By analyzing the model parameters, we can understand the degree of influence of different bias current values ​​on the detector responsivity. For example, when the bias current increases by 1mA, the responsivity increases by 0.1.

[0052] 2. Synchronous optimization of state adjustment decision knowledge and device performance feedback semantics

[0053] 2.1. Optimizing state adjustment decision knowledge

[0054] According to the entity semantic transfer characteristics, if it is found that a certain state adjustment decision knowledge does not significantly improve the performance of the device or has a negative impact, it needs to be optimized. For example, in the above bias current adjustment example, if it is found that the current bias current setting (for example, 5mA) causes the photodetector's responsiveness to be lower than expected (the expected responsiveness is 0.6A / W according to the model calculation, and the actual responsiveness is 0.5A / W), the decision knowledge of bias current adjustment can be adjusted. You can try to increase the bias current to 6mA, and at the same time, according to the model prediction, the responsiveness will increase to 0.6A / W.

[0055] 2.2. Optimizing device performance feedback semantics

[0056] The device performance feedback semantics can also be optimized based on the entity semantic transfer characteristics. Since the device performance feedback semantics may have measurement errors or incompleteness, it can be corrected or supplemented by analyzing the semantic transfer relationship between it and the state regulation decision knowledge. For example, for the output power measurement of the laser, if it is found that the measured value deviates from the value predicted based on the state regulation decision knowledge (such as the APC setting) (the measured value is set to 3dBm and the predicted value is 3.5dBm), the accuracy of the measuring equipment can be checked or the measurement method can be adjusted. At the same time, the expected range of the output power can be more accurately defined based on the state regulation decision knowledge such as the APC control algorithm, and the original wider range (such as 3-4dBm) can be adjusted to a narrower range (such as 3.3-3.7dBm).

[0057] 3. Determine the state adjustment strategy and verification of the target photovoltaic module

[0058] 3.1. Determine the state adjustment strategy

[0059] The state adjustment decision knowledge of several attention dimensions after synchronous optimization is used to determine the current state adjustment strategy of the target optoelectronic module. For example, if the state adjustment decision knowledge under the temperature adjustment dimension is optimized, the cooling power of the TEC is determined to be set to 3 watts (based on the ambient temperature and device performance feedback in the operating scenario), the APC proportional coefficient is determined to be 0.7 and the optical power monitoring frequency is once every 8 minutes under the optical power adjustment dimension, the bias current of the photodetector is determined to be 6mA under the bias current adjustment dimension, and the 16QPSK modulation format is determined to be used with a modulation depth of 50% under the signal modulation optimization dimension, these constitute the current state adjustment strategy of the target optoelectronic module.

[0060] 3.2. Status Adjustment Verification

[0061] The X device performance feedback semantics after synchronous optimization are used for state adjustment verification. After the target optoelectronic module operates according to the above state adjustment strategy, the effect of state adjustment is verified by monitoring the device performance feedback semantics. For example, monitor whether the output power of the laser is within the expected range (such as 3.3-3.7dBm) and whether the responsivity of the photodetector meets the expectations (such as 0.6A / W). If the device performance feedback semantics meet expectations, it means that the state adjustment strategy is effective; if not, further analysis and adjustment of the state adjustment strategy is required.

[0062] Based on the above content, operating scenario description knowledge, state adjustment decision knowledge, device performance feedback semantics, knowledge feature relationship spectrum, terminal relationship entity, linkage relationship entity, and entity semantic transfer features can all be represented in the form of feature vectors. Based on this, a detailed introduction is given below through illustrative examples.

[0063] 1. Examples of technical terms based on feature vectors

[0064] 1. Feature vector representation of operational scenario description knowledge

[0065] The knowledge of setting the operating scenario description can be represented by a feature vector vec{R}, then vec{R}=[r_1, r_2, r_3, r_4]. Among them, r_1 represents the type of equipment where the optoelectronic module is located. If it is an optical fiber communication base station equipment, r_1=1, and if it is a data center server interconnection equipment, r_1=2; r_2 represents the ambient temperature range in degrees Celsius, for example, r_2=[-20, 60]; r_3 represents the length of the optical fiber link in meters, r_3=500 represents medium distance; r_4 represents the transmission data rate requirement in Gbps, r_4=100.

[0066] 2. Feature vector representation of state adjustment decision knowledge

[0067] Taking optical power regulation as an example, for the attention dimension of optical power regulation, the characteristic vector vec{S}_{1} of the state regulation decision knowledge can be expressed as vec{S}_{1}=[s_{11}, s_{12}, s_{13}, s_{14}]. Among them, s_{11} represents the feedback loop proportional coefficient of automatic power control (APC), and s_{11}=0.5 is set; s_{12} represents the optical power monitoring frequency in seconds, and s_{12}=10^{-5} means monitoring every 10 microseconds; s_{13} represents the maximum allowable deviation of optical power regulation in dBm, and s_{13}=0.5; s_{14} represents the type of optical power regulation algorithm, and if it is a first-order linear feedback algorithm, s_{14}=1.

[0068] Taking temperature regulation as an example, for the temperature regulation attention dimension, the feature vector of state regulation decision knowledge can be: vec{S}_{2}=[s_{21}, s_{22}, s_{23}, s_{24}]. Among them, s_{21} represents the maximum cooling power of the thermoelectric cooler TEC, in watts, s_{21}=5; s_{22} represents the temperature control accuracy, in degrees Celsius, s_{22}=±0.1; s_{23} represents the temperature monitoring frequency, in minutes, s_{23}=5; s_{24} represents a key parameter in the TEC control algorithm, such as the initial value of the duty cycle, s_{24}=0.3.

[0069] 3. Feature vector representation of device performance feedback semantics

[0070] For lasers, the device performance feedback semantic feature vector vec{F}_{1} of the laser can be expressed as vec{F}_{1}=[f_{11}, f_{12}, f_{13}, f_{14}]. Among them, f_{11} represents the actual output power of the laser, in dBm, f_{11}=3; f_{12} represents the stability of the output wavelength, in nm, f_{12}=±0.05; f_{13} represents the threshold current, in mA, f_{13}=10.2; f_{14} represents the efficiency of the laser, expressed as a decimal, f_{14}=0.3.

[0071] For photodetectors, the device performance feedback semantic feature vector vec{F}_{2} of the photodetector is vec{F}_{2}=[f_{21}, f_{22}, f_{23}, f_{24}]. Where f_{21} represents the responsivity of the detector, in A / W, f_{21}=0.8; f_{22} represents the dark current, in nA, f_{22}=1; f_{23} represents the bandwidth of the detector, in GHz, f_{23}=10; f_{24} represents the noise equivalent power of the detector, in W, f_{24}=10^{-12}.

[0072] 4. Construction and Representation of Knowledge Feature Relation Spectrum

[0073] In terms of construction, the knowledge feature relationship spectrum can be represented by a matrix, assuming that there are two attention dimensions (optical power regulation and temperature regulation) and two devices (lasers and photodetectors). The rows of the matrix represent different relationship entities (terminal relationship entities and linkage relationship entities), and the columns represent different knowledge sources (operation scenario description knowledge, state regulation decision knowledge, device performance feedback semantics). For example, the first row of the matrix may represent the relationship between the optical power regulation decision knowledge and the laser output power, and the second row represents the relationship between the temperature regulation decision knowledge and the photodetector responsiveness, etc.

[0074] In terms of example elements, the element m_{ij} in the matrix represents the strength or degree of association between the i-th relational entity and the j-th knowledge source. For example, m_{11} represents the degree of association between the optical power regulation decision knowledge and the optical power requirements of the data rate in the operation scenario description knowledge. If the higher the data rate, the higher the requirement for optical power stability, then m_{11} may be a larger value, set to 0.8; m_{13} represents the degree of association between the optical power regulation decision knowledge and the actual output power device performance feedback semantics of the laser. Since the optical power regulation decision knowledge directly affects the laser output power, m_{13} may be 1.

[0075] 5. Feature vector representation of terminal relationship entities and linkage relationship entities

[0076] As for the terminal relationship entity, taking the APC feedback loop proportional coefficient in the optical power regulation decision knowledge (as a terminal relationship entity) as an example, its characteristic vector vec{Z}_{1} can be expressed as vec{Z}_{1}=[z_{11}], where z_{11}=0.5 (that is, the value of the proportional coefficient).

[0077] As for the linkage relationship entity, taking the actual output power of the laser (as a linkage relationship entity) as an example, its eigenvector is: (vec{L}_{1}=[l_{11}]), where l_{11}=3 (that is, the value of the actual output power).

[0078] 6. Feature vector representation of entity semantic transfer features

[0079] Taking the effect of bias current adjustment on the responsivity of photodetectors as an example, the feature vector of entity semantic transfer features can be expressed as vec{E}=[e_1, e_2], where e_1 represents the proportional relationship between the change in bias current and the change in responsivity (such as a=0.1 mentioned above), and e_2 represents the initial responsivity when the bias current is 0 (such as b=0.05 mentioned above).

[0080] Furthermore, the state adjustment decision knowledge of the several attention dimensions after synchronous optimization is used to determine the current state adjustment strategy of the target optoelectronic module, and the X device performance feedback semantics after synchronous optimization are used to perform state adjustment verification as the core invention points of the present invention, which are described in detail as follows.

[0081] 1. The state adjustment decision knowledge after synchronous optimization determines the current state adjustment strategy of the target photovoltaic module

[0082] In terms of temperature regulation, the elements in the state regulation decision knowledge feature vector vec{S}_{2} after synchronous optimization have changed. For example, according to the ambient temperature range r_2 in the operation scenario description knowledge and the laser threshold current f_{13} in the device performance feedback semantics (because temperature affects the threshold current), the optimized s_{21} (maximum cooling power of TEC) may be adjusted to 4 watts, s_{22} (temperature control accuracy) remains unchanged, and s_{23} (temperature monitoring frequency) is adjusted to 3 minutes. Based on these optimized state regulation decision knowledge, the temperature regulation strategy of the target photovoltaic module is determined as follows: use TEC for temperature control, the maximum cooling power is 4 watts, the control accuracy is ±0.1℃, and the temperature monitoring frequency is once every 3 minutes.

[0083] For optical power regulation, the optimized state regulation decision knowledge feature vector vec{S}_{1} determines the optical power regulation strategy. For example, according to the transmission data rate r_4 in the operation scenario and the laser output power f_{11} in the device performance feedback semantics, s_{11} (the feedback loop proportionality coefficient of APC) is adjusted to 0.6, s_{12} (optical power monitoring frequency) is adjusted to once every 8 microseconds, and s_{13} (maximum allowable deviation) is adjusted to 0.3dBm. Then the optical power regulation strategy is: use APC for optical power regulation, the feedback loop proportionality coefficient is 0.6, the monitoring frequency is once every 8 microseconds, and the maximum allowable deviation is 0.3dBm.

[0084] Similarly, in the bias current adjustment dimension, the bias current adjustment strategy is determined by setting the optimized state adjustment decision knowledge feature vector vec{S}_{3} (for bias current adjustment). If s_{31} (bias current value) is determined to be 6mA based on the photodetector responsivity f_{21} in the device performance feedback semantics and other relevant factors in the operating scenario (such as signal noise level, etc.), then the bias current adjustment strategy is to set the bias current of the photodetector to 6mA.

[0085] In the signal modulation optimization dimension, the optimized state adjustment decision knowledge feature vector vec{S}_{4} determines the modulation format and modulation depth. For example, according to the transmission data rate r_4 and the fiber link length r_3 in the operation scenario, the 16QPSK modulation format is determined, and s_{41} (modulation depth) is 50%, which constitutes the signal modulation optimization strategy.

[0086] 2. Synchronous optimization of device performance feedback semantics for state adjustment verification

[0087] After the target optoelectronic module is temperature-regulated according to the above temperature regulation strategy, the optimized device performance feedback semantic feature vectors vec{F}_{1} (for lasers) and vec{F}_{2} (for photodetectors) are used for verification. For example, observe whether the threshold current f_{13} of the laser is stable within a reasonable range. If the threshold current fluctuation range is expected to be within ±0.1mA according to the optimized temperature regulation strategy, and the threshold current fluctuation actually monitored is within this range, it means that the temperature regulation strategy is effective. For the photodetector, observe whether its responsivity f_{21} is optimized due to temperature regulation. If the stability of the responsivity is improved after the temperature is stabilized (for example, the fluctuation is reduced from the original ±0.02A / W to ±0.01A / W), it also means that the temperature regulation strategy has played a positive role.

[0088] After running the optimized optical power regulation strategy, check whether the actual output power f_{11} of the laser is within the expected range. For example, if the maximum allowable deviation set by the optical power regulation strategy is 0.3dBm, the actual output power should be within the set target power ±0.3dBm. If the actual output power monitored is within this range, it means that the optical power regulation strategy is effective. At the same time, you can also observe whether the optical power received by the photodetector is stable within a suitable range to further verify the effectiveness of the optical power regulation strategy in the entire optoelectronic module system.

[0089] For the bias current regulation strategy, it is verified by observing the responsivity f_{21} of the photodetector. If the bias current regulation strategy sets the bias current to 6mA, according to the entity semantic transfer feature (such as the linear model R=a×I+b mentioned above), the expected responsivity is a certain value. If the actual monitored responsivity is close to the expected value (within the allowable error range), it means that the bias current regulation strategy is effective.

[0090] After the signal modulation optimization strategy is implemented, it can be verified by observing the bit error rate and other device performance feedback semantics related to signal transmission quality. For example, if the 16QPSK modulation format and the modulation depth are 50%, the expected bit error rate should be lower than a certain threshold (set to 10^{-9}). By monitoring the actual bit error rate, if the actual bit error rate is lower than this threshold, it means that the signal modulation optimization strategy is effective.

[0091] In summary, the present invention has many beneficial effects in optimizing the state regulation of photovoltaic modules.

[0092] In terms of knowledge acquisition and relationship construction, by acquiring the operating scenario description knowledge of the reference optoelectronic module, the state adjustment decision knowledge of several attention dimensions, and the device performance feedback semantics, and constructing the knowledge feature relationship spectrum, it is possible to comprehensively and systematically integrate various key information in the operation process of the optoelectronic module. This helps to deeply understand the complex relationship between various factors of the optoelectronic module in different operating scenarios. For example, the operating scenario description knowledge covers many factors such as the device type, ambient temperature range, optical fiber link length, and transmission data rate requirements of the optoelectronic module. These factors are interrelated with the state adjustment decision knowledge under different attention dimensions (such as the thermoelectric cooler control strategy in temperature regulation, automatic power control knowledge in optical power regulation, etc.) and device performance feedback semantics (such as the output power, threshold current of the laser, and the responsiveness of the photodetector, etc.). This association is clearly presented through the terminal relationship entity (representing the state adjustment decision knowledge) and the linkage relationship entity (representing the device performance feedback semantics) in the knowledge feature relationship spectrum. This allows the mutual influence of many factors to be comprehensively considered when performing state adjustment, avoiding the problem of poor adjustment effect caused by focusing only on a single factor or a few factors in traditional adjustment methods.

[0093] The optimization of state adjustment decision knowledge based on the timing commonality metric value has significant advantages. By calculating the timing commonality metric value of the operation scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge to synchronously optimize the state adjustment decision knowledge, the adjustment strategy can be better adapted to the actual operation rhythm of the optoelectronic module. For example, in practical applications, the operation scenario of the optoelectronic module may have various time-related changes, such as periodic changes in ambient temperature, fluctuations in transmission data rate in different time periods, etc. If the state adjustment decision knowledge cannot match these time changes, it may cause problems such as adjustment lag or over-adjustment. This optimization method of the present invention can timely adjust the relevant parameters in the state adjustment decision knowledge according to the timing commonality metric value, such as the value calculated by the dynamic time warping (DTW) algorithm. Taking optical power regulation as an example, if it is found that the optical power monitoring frequency does not match the change law of ambient temperature or data transmission rate in time (reflected by the timing commonality metric value), the optical power monitoring frequency can be adjusted, and other parameters related to optical power regulation, such as the feedback loop parameters of automatic power control (APC), can be optimized accordingly. This helps improve the stability and performance of optoelectronic modules under different time conditions and reduces the risk of signal transmission errors or device performance degradation due to timing mismatch.

[0094] Synchronous optimization based on entity semantic transfer features is also of great significance. This optimization method takes into account the entity semantic transfer features between terminal relationship entities and linkage relationship entities, and can more accurately synchronize and optimize state regulation decision knowledge and device performance feedback semantics. In optoelectronic modules, there are inherent causal relationships and semantic connections between different state regulation decision knowledge and device performance feedback semantics. For example, bias current adjustment has a direct impact on the performance of photodetectors (such as responsivity). By constructing a causal relationship model (such as a linear model) to quantify this entity semantic transfer feature, the decision knowledge of bias current adjustment can be accurately adjusted when it is found that the detector responsivity is lower than expected due to unreasonable bias current settings. At the same time, this optimization method can also correct and supplement the device performance feedback semantics. Since device performance feedback semantics may have measurement errors or incompleteness, it can be optimized by analyzing the semantic transfer relationship between it and state regulation decision knowledge. For example, for the measurement of laser output power, if there is a deviation from the value predicted based on state regulation decision knowledge (such as APC setting), this semantic transfer relationship can be used to analyze the cause, adjust the measurement method, or define the expected range of output power more accurately. This helps to improve the accuracy and reliability of state regulation, allowing each device of the photovoltaic module to operate in a better working state, thereby improving the performance of the entire photovoltaic module.

[0095] In terms of determining the state adjustment strategy and verification of the target optoelectronic module, the optimized state adjustment decision knowledge of the present invention can determine a comprehensive and effective state adjustment strategy, including a temperature adjustment strategy, an optical power adjustment strategy, a bias current adjustment strategy, a signal modulation optimization strategy, etc. This enables the optoelectronic module to be optimized and adjusted in multiple dimensions according to different operating requirements and actual conditions. For example, in a long-distance, high-speed transmission scenario, a suitable temperature control strategy can be determined based on the optimized decision knowledge to ensure stable device performance, an accurate optical power adjustment strategy to ensure appropriate signal strength, an appropriate bias current adjustment strategy to optimize detector performance, and an efficient signal modulation optimization strategy to improve spectrum utilization and transmission quality. At the same time, the optimized device performance feedback semantics is used for state adjustment verification, and the effectiveness of the state adjustment strategy can be accurately evaluated. By monitoring the various indicators in the optimized device performance feedback semantics (such as the output power of the laser, the responsiveness of the photoelectric detector, the bit error rate, etc.), problems in the state adjustment process can be discovered in a timely manner, and then the adjustment strategy can be adjusted. This helps to improve the maintainability and stability of the optoelectronic module, reduce failures and performance degradation caused by improper state adjustment, extend the service life of the optoelectronic module, and improve its applicability in various application scenarios.

[0096] In some optional embodiments, the synchronous optimization of the state adjustment decision knowledge of the several attention dimensions based on the timing commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions described in step 204 includes: determining the target state adjustment decision knowledge based on the timing commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions, the target state adjustment decision knowledge being the state adjustment decision knowledge with the largest timing commonality measurement value with the operation scenario description knowledge of the reference photoelectric module among the state adjustment decision knowledge of the several attention dimensions; synchronously optimizing the target state adjustment decision knowledge by weighted processing of the timing commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the target state adjustment decision knowledge.

[0097] Further, based on the time series commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions, the target state adjustment decision knowledge is determined, and the target state adjustment decision knowledge is the state adjustment decision knowledge with the largest time series commonality measurement value with the operation scenario description knowledge of the reference photoelectric module among the state adjustment decision knowledge of the several attention dimensions; on the basis of synchronously optimizing the target state adjustment decision knowledge by weighted processing of the time series commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the target state adjustment decision knowledge, the method also includes: moving towards the convergence of the first time series The direction of the distance between the sequence commonality metric value and the second sequence commonality metric value is used to synchronously optimize the state adjustment decision knowledge of the several attention dimensions, wherein the first sequence commonality metric value is the sequence commonality metric value of the operation scenario description knowledge of the reference photoelectric module and any one of the state adjustment decision knowledge of the several attention dimensions, and the second sequence commonality metric value is the sequence commonality metric value of the operation scenario description knowledge of the corrected photoelectric module and any one of the state adjustment decision knowledge of the several attention dimensions, and the corrected photoelectric module is determined after troubleshooting the reference photoelectric module.

[0098] In the state regulation of the photovoltaic module, the reference photovoltaic module's operating scenario description knowledge contains many aspects of information. For example, the ambient temperature change in the operating scenario description knowledge may show a certain periodic law, setting the temperature to rise from 10°C in the morning to 25°C at noon, and then drop to 15°C at night within a day. At the same time, the state regulation decision knowledge of different attention dimensions also has its own time-related characteristics. Taking optical power regulation as an example, the optical power monitoring frequency in its state regulation decision knowledge may be once every 10 minutes, and the feedback loop parameter adjustment cycle of the automatic power control (APC) may be once every 30 minutes.

[0099] In order to calculate the commonality metric of the time series, it is necessary to quantify the time-related information. The time series analysis method can be used to convert the time series data such as temperature changes in the operation scenario description knowledge into discrete numerical sequences. For the state adjustment decision knowledge of optical power regulation, the monitoring frequency and adjustment cycle are also converted into corresponding numerical representations.

[0100] Taking the dynamic time warping (DTW) algorithm as an example, the temporal commonality measurement value between the operating scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge of each attention dimension is calculated. It is assumed that there are three attention dimensions of state adjustment decision knowledge, namely optical power adjustment knowledge A, temperature adjustment knowledge B, and bias current adjustment knowledge C. For optical power adjustment knowledge A, the temporal commonality measurement value Sim_A with the operating scenario description knowledge is calculated by the DTW algorithm. It is assumed that during the calculation process, after the distance calculation and normalization processing of the algorithm, Sim_A=0.6. Similarly, the temporal commonality measurement value Sim_B=0.8 of the temperature adjustment knowledge B and the temporal commonality measurement value Sim_C=0.5 of the bias current adjustment knowledge C are calculated.

[0101] According to the magnitude of the time series commonality metric, the target state adjustment decision knowledge is determined. In the above example, the time series commonality metric of temperature adjustment knowledge B is the largest, so temperature adjustment knowledge B is determined as the target state adjustment decision knowledge.

[0102] Once the target state adjustment decision knowledge (such as temperature adjustment knowledge B) is determined, it needs to be weighted for optimization. The purpose of weighted processing is to adjust the relevant parameters in the target state adjustment decision knowledge according to the temporal commonality measurement between the operation scenario description knowledge and the target state adjustment decision knowledge. For example, the relevant parameters of the thermoelectric cooler (TEC) control strategy in the temperature adjustment knowledge B may include the cooling power adjustment step s (set to an initial value of 0.5 watts), the proportional coefficient k of the temperature control (set to an initial value of 0.3), etc.

[0103] The weighting coefficient w is set based on the timing commonality metric value Sim_B=0.8. An exemplary method for determining the weighting coefficient may be w=Sim_B. Then for the optimization of the cooling power adjustment step s, the new s' can be calculated by the formula s'=s×w, that is, s'=0.5×0.8=0.4 watts. For the proportional coefficient k of temperature control, the new k' can be calculated in a similar manner and set according to k'=k×w+0.1 (0.1 is added in this embodiment to introduce a certain offset adjustment, and the specific offset value can be determined according to the actual situation), then k'=0.3×0.8+0.1=0.34.

[0104] Furthermore, during the operation of the optoelectronic module, faults may occur. For example, it is assumed that the laser in the optoelectronic module has an unstable output power fault. Through troubleshooting operations, it is found that the laser temperature is too high due to poor heat dissipation, which in turn affects its output power. During the troubleshooting process, the operation scenario description knowledge of the optoelectronic module will be corrected. For example, the heat dissipation in the original operation scenario description knowledge may be evaluated as normal, but after troubleshooting, it is found that the heat dissipation efficiency is lower than expected, so the heat dissipation-related operation scenario description knowledge is corrected, such as the heat dissipation coefficient from the original h=0.8 (set as an indicator to represent the heat dissipation efficiency) to h=0.5. In this way, the corrected operation scenario description knowledge of the optoelectronic module is obtained, which contains more accurate information about the actual operation of the optoelectronic module, especially information related to the fault.

[0105] For any state adjustment decision knowledge (taking optical power adjustment knowledge A as an example), calculate the first time series commonality metric value. According to the DTW algorithm mentioned above, calculate the time series commonality metric value Sim_{A1} of the reference photoelectric module's operating scenario description knowledge and the optical power adjustment knowledge A (set Sim_{A1}=0.6). Then calculate the second time series commonality metric value, that is, the time series commonality metric value Sim_{A2} of the corrected photoelectric module's operating scenario description knowledge and the optical power adjustment knowledge A. Set the calculated Sim_{A2}=0.4.

[0106] In order to optimize the optical power regulation knowledge A in the direction of converging the distance between the first timing commonality metric value and the second timing commonality metric value, an adjustment strategy needs to be determined. Set the parameters in the optical power regulation knowledge A according to the distance d=|Sim_{A1}-Sim_{A2}| (in this embodiment d=|0.6-0.4|=0.2). For example, the optical power monitoring frequency f in the optical power regulation knowledge A (the initial value is set to once every 10 minutes), if d is larger (such as greater than 0.1), the optical power monitoring frequency can be increased. An exemplary adjustment formula can be f'=frac{f}{1+d}, then f'=frac{10}{1+0.2}≈once every 8.33 minutes. In this way, by adjusting the optical power monitoring frequency, it is expected that the timing commonality metric value of the optical power regulation knowledge A in the operating scenarios of the reference optoelectronic module and the corrected optoelectronic module will be closer, thereby improving the adaptability and effectiveness of optical power regulation.

[0107] The state adjustment decision knowledge of other attention dimensions (such as temperature adjustment knowledge B and bias current adjustment knowledge C) are also calculated and optimized in a similar way. In this way, each state adjustment decision knowledge can be more adapted to the actual situation in the operation scenario after troubleshooting, thereby improving the overall performance and stability of the photovoltaic module.

[0108] It can be seen that in terms of determining the target state adjustment decision knowledge and weighted optimization, by calculating the timing commonality metric value to find the decision knowledge that best matches the operation scenario and weighted optimization, the parameters in the decision knowledge can be accurately adjusted. For example, in temperature adjustment, through specific numerical calculations and weighting, the parameters such as the cooling power adjustment step and the temperature control proportional coefficient of the thermoelectric cooler are reasonably adjusted to improve the adaptability of temperature adjustment to the operation scenario. In terms of optimizing in the direction of the convergence timing commonality metric value distance, considering the operation scenario after troubleshooting, by calculating different timing commonality metric values ​​and adjusting the parameters in the decision knowledge, such as adjusting the optical power monitoring frequency, each state adjustment decision knowledge can better adapt to different operation scenarios, whether it is normal or after fault correction. This helps to improve the stability, reliability and overall performance of the optoelectronic module, reduce the risk of failure caused by improper state adjustment, extend the service life of the optoelectronic module and optimize its operation effect in various complex scenarios.

[0109] In addition, based on the temporal commonality measurement value of the operating scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions, the target state adjustment decision knowledge is determined, and the target state adjustment decision knowledge is the state adjustment decision knowledge with the largest temporal commonality measurement value with the operating scenario description knowledge of the reference photoelectric module among the state adjustment decision knowledge of the several attention dimensions; on the basis of synchronously optimizing the target state adjustment decision knowledge by weighted processing of the temporal commonality measurement value of the operating scenario description knowledge of the reference photoelectric module and the target state adjustment decision knowledge, the method also includes: synchronously optimizing the operating scenario description knowledge of the reference photoelectric module and the operating scenario description knowledge of the corrected photoelectric module in the direction of increasing the temporal commonality measurement value of the operating scenario description knowledge of the corrected photoelectric module and the operating scenario description knowledge of the reference photoelectric module.

[0110] In the state regulation system of the photovoltaic module, the reference photovoltaic module operation scenario description knowledge covers a lot of time series information related to the operation. For example, if the ambient temperature changes over time, the 24 hours of a day are discretized into one data point per hour, and the temperature sequence T={t_1, t_2, ..., t_{24}} is obtained, where t_i represents the temperature value of the i-th hour. At the same time, the state regulation decision knowledge of different attention dimensions also has time-related characteristics. Taking the optical power regulation dimension as an example, its state regulation decision knowledge may involve the time interval sequence M={m_1, m_2, ...} of optical power monitoring, where m_i represents the time interval of the i-th monitoring.

[0111] Calculate the temporal commonality measurement value between the operation scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge of each attention dimension. Use a dynamic time warping (DTW) algorithm for calculation. For the optical power adjustment knowledge, set the temporal commonality measurement value between it and the operation scenario description knowledge calculated by the DTW algorithm to be Sim_{1}. For example, in the specific calculation process, the DTW algorithm constructs a distance matrix, calculates the cumulative distance from the starting point of T to the starting point of M, and obtains a total distance value d_total after a series of path searches and distance accumulation. Then, d_total is obtained by a specific normalization function f (d_total) to obtain Sim_{1}, and Sim_{1}=0.6 is set. Similarly, for the temperature adjustment knowledge, the temporal commonality measurement value between it and the operation scenario description knowledge is calculated to be Sim_{2}, and Sim_{2}=0.8; for the bias current adjustment knowledge, the temporal commonality measurement value calculated is Sim_{3}, and Sim_{3}=0.5. According to the calculated time series commonality metric value, the target state adjustment decision knowledge is determined. In the above example, the time series commonality metric value Sim_{2}=0.8 of the temperature adjustment knowledge is the largest, so the temperature adjustment knowledge is determined as the target state adjustment decision knowledge.

[0112] It can be understood that after determining the target state adjustment decision knowledge (such as temperature adjustment knowledge), weighted processing is performed to optimize the knowledge. The set temperature adjustment knowledge includes multiple parameters, such as the control parameters of the thermoelectric cooler (TEC), including the cooling coefficient k (the initial value is set to 0.5) and the temperature adjustment step s (the initial value is set to 0.3 degrees Celsius).

[0113] The weighting coefficient w is determined according to the temporal commonality measurement value Sim_{2}=0.8 between the operation scenario description knowledge of the reference photovoltaic module and the target state adjustment decision knowledge (temperature adjustment knowledge). For example, an exemplary linear relationship w=Sim_{2} is adopted.

[0114] For the optimization of the cooling coefficient k, the new k' is calculated by the formula k'=k×w, that is, k'=0.5×0.8=0.4. For the temperature adjustment step s, the new s' is calculated by the formula s'=s×w+0.05 (the addition of 0.05 in this embodiment is to introduce a small offset adjustment, and the specific value can be determined based on actual experience), and s'=0.3×0.8+0.05=0.29 degrees Celsius is obtained.

[0115] In addition, during the operation of the optoelectronic module, the operation scenario description knowledge may need to be revised due to various reasons. For example, during the operation of a certain optoelectronic module, the initial operation scenario description knowledge believes that the loss of the optical fiber link is stable, and the estimated loss value is L=2dB (in this embodiment, L represents the optical fiber link loss). However, after a period of operation and testing, it was found that due to factors such as optical fiber aging or external interference, the actual optical fiber link loss has increased. Through a series of detection and analysis methods, it was determined that the actual optical fiber link loss has become L'=3dB, and the loss change pattern is also different from the previous estimate.

[0116] Based on these new findings, the corrected photovoltaic module operation scenario description knowledge is updated. In addition to the numerical changes in fiber link loss, other corrections may also be involved. For example, it was originally believed that the effect of ambient temperature on photovoltaic modules was linear, but after troubleshooting or more in-depth analysis, it was found that within a specific temperature range, the effect of ambient temperature on photovoltaic modules is nonlinear. This relationship will also be updated to the corrected photovoltaic module operation scenario description knowledge.

[0117] It is worth mentioning that in order to optimize in the direction of increasing the temporal commonality metric value between the operating scenario description knowledge of the corrected photovoltaic module and the operating scenario description knowledge of the reference photovoltaic module, it is first necessary to recalculate the temporal commonality metric value between the two.

[0118] The ambient temperature change sequence in the operation scenario description knowledge of the modified photovoltaic module is set to T'={t'_1, t'_2, ..., t'_{24}}, and the temperature sequence in the operation scenario description knowledge of the reference photovoltaic module is set to T={t_1, t_2, ..., t_{24}}. The DTW algorithm is used again to calculate the temporal commonality metric Sim_{4} between the two. For example, after calculation, Sim_{4}=0.4 is obtained.

[0119] The goal of optimization is to increase the timing commonality metric value Sim_{4}, which can be achieved by adjusting the relevant parameters in the operating scenario description knowledge of the reference photovoltaic module and the operating scenario description knowledge of the modified photovoltaic module.

[0120] Taking the ambient temperature sequence as an example, in the operating scenario description knowledge of the reference photovoltaic module, if the temperature sequence T has a smoothing factor p (used to smooth the temperature data, with an initial value set to 0.2), in the operating scenario description knowledge of the corrected photovoltaic module, the corresponding smoothing factor is p' (with an initial value set to 0.3).

[0121] In order to increase the timing commonality metric value Sim_{4}, the smoothing factor can be adjusted according to the value of Sim_{4}. For example, an exemplary adjustment strategy is adopted. For the smoothing factor p in the operating scenario description knowledge of the reference photovoltaic module, the new p'' is calculated by the formula p''=p+(1-Sim_{4})×0.1, and p''=0.2+(1-0.4)×0.1=0.26 is obtained. For the smoothing factor p' in the operating scenario description knowledge of the corrected photovoltaic module, the new p'''=p'-(1-Sim_{4})×0.1 is obtained, and p'''=0.3-(1-0.4)×0.1=0.24 is obtained.

[0122] In this way, the parameters in the operation scenario description knowledge are adjusted so that the operation scenario description knowledge of the modified photovoltaic module is closer to the operation scenario description knowledge of the reference photovoltaic module in terms of timing, thereby increasing the timing commonality metric between the two. This optimization helps to better utilize existing knowledge and experience in the state adjustment decision-making process and improve the accuracy and effectiveness of state adjustment, especially when the operation scenario of the photovoltaic module changes or there is uncertainty.

[0123] With this design, in terms of determining the target state adjustment decision knowledge and weighted optimization, the target decision knowledge is determined by accurately calculating the time series commonality metric value, and then the relevant parameters are optimized by weighted processing. For example, the optimization of the cooling coefficient and the temperature adjustment step in temperature regulation can make the adjustment decision knowledge better match the operation scenario. In terms of optimizing the operation scenario description knowledge in the direction of increasing the time series commonality metric value, considering the case of the corrected photovoltaic module, by recalculating the metric value and adjusting the relevant parameters (such as the smoothing factor of the ambient temperature series), the operation scenario description knowledge can be more matched in time. This helps to improve the accuracy, adaptability and stability of the photovoltaic module state regulation, better cope with changes and uncertainties in the operation scenario, reduce adjustment errors caused by scenario changes, and improve the overall performance and reliability of the photovoltaic module.

[0124] In a preferred embodiment, the synchronous optimization of the state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics based on the entity semantic transmission characteristics between the terminal relationship entity and the linkage relationship entity recorded in step 206 specifically includes: determining a first linkage relationship entity from the linkage relationship entity, the first linkage relationship entity representing a terminal relationship entity of the first state adjustment decision knowledge and a core upstream relationship entity of the terminal relationship entity representing the second state adjustment decision knowledge, the first state adjustment decision knowledge being the state conduction influence knowledge of the second state adjustment decision knowledge, the second state adjustment decision knowledge being the state conduction influence knowledge of the first state adjustment decision knowledge, and the first state adjustment decision knowledge and the second state adjustment decision knowledge both belonging to the state adjustment decision knowledge; and synchronously optimizing the state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics in the direction of increasing the temporal commonality metric value between the first state adjustment decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity.

[0125] In detail, determining the first linkage relationship entity from the linkage relationship entities includes: determining the discrimination possibility of each linkage relationship entity, the discrimination possibility being a quantitative score of the linkage relationship entity as a core upstream relationship entity of a terminal relationship entity representing first state adjustment decision knowledge and a terminal relationship entity representing second state adjustment decision knowledge; determining the confidence weight of each linkage relationship entity, the confidence weight being a weighted result of the discrimination possibility and the dynamic quantitative score; and determining the first linkage relationship entity from the linkage relationship entities based on the confidence weight of the linkage relationship entity.

[0126] In the state regulation system of optoelectronic modules, terminal relationship entities are used to represent state regulation decision knowledge, while linkage relationship entities are used to represent device performance feedback semantics. For example, in the state regulation decision knowledge of optical power regulation, key parameters in the automatic power control (APC) algorithm (such as proportional coefficient, integral time constant, etc.) can constitute terminal relationship entities. Device performance feedback semantics, such as the actual output power of the laser and the responsiveness of the photodetector, are represented by linkage relationship entities. These entities are interrelated in the knowledge feature relationship spectrum to form a complex semantic transfer relationship.

[0127] The first state regulation decision knowledge is the state conduction influence knowledge of the second state regulation decision knowledge, and vice versa. Taking temperature regulation and optical power regulation as examples, the temperature regulation decision knowledge (first state regulation decision knowledge) will affect the performance of the devices in the optoelectronic module, thereby indirectly affecting the optical power regulation decision knowledge (second state regulation decision knowledge). Specifically, changes in temperature may cause changes in the threshold current of the laser, thereby affecting its output power, which requires the optical power regulation decision knowledge to make corresponding adjustments. Similarly, changes in the optical power regulation decision knowledge, such as an increase in optical power, may cause the device temperature to rise, thereby affecting the temperature regulation decision knowledge. This mutual state conduction influence relationship is very critical in the state regulation of the entire optoelectronic module.

[0128] For each linkage relationship entity, its discrimination possibility needs to be determined, which is a quantitative score of its core upstream relationship entity for the terminal relationship entity representing the first state regulation decision knowledge and the terminal relationship entity representing the second state regulation decision knowledge. For example, consider an optoelectronic module system containing multiple linkage relationship entities, one of which is the threshold current change of the laser (set as linkage relationship entity entity_A). Its discrimination possibility for temperature regulation decision knowledge (first state regulation decision knowledge) and optical power regulation decision knowledge (second state regulation decision knowledge) needs to be determined.

[0129] First, the relationship between the threshold current change and the key parameters in the temperature regulation decision knowledge (such as the control parameters of the thermoelectric cooler (TEC)) and the key parameters in the optical power regulation decision knowledge (such as the proportional coefficient in the APC algorithm) is analyzed. According to historical data and theoretical models, the threshold current of the laser changes by 0.1mA for every 1^{circ}C change in temperature, and a 0.1mA change in threshold current will result in a 0.05dBm change in optical power.

[0130] Based on these relationships, a function can be constructed to calculate the discriminant likelihood. Assume that an exemplary linear function is used:

[0131] P(entity_A)=k_1×|δI_{th} / δT|+k_2×|δP_{out} / δI_{th}|

[0132] Where δI_{th} is the threshold current change, δT is the temperature change, δP_{out} is the optical power change, and k_1 and k_2 are weight coefficients set to k_1=0.3, k_2=0.7). If the current temperature change is δT=2^{circ}C, the threshold current change is δI_{th}=0.2mA, and the optical power change is δP_{out}=0.1dBm, then the discrimination probability P(entity_A)=0.3×|0.2 / 2|+0.7×|0.1 / 0.2|=0.4.

[0133] When determining the confidence weight, in addition to determining the possibility, it is also necessary to introduce a dynamic quantitative score. The dynamic quantitative score is a value dynamically determined based on the current operating status of the photovoltaic module. For example, if the current operating temperature of the photovoltaic module is in a critical range (close to the critical operating temperature of some devices), the dynamic quantitative score of the temperature-related linkage relationship entity will be increased accordingly. For the linkage relationship entity entity_A, the dynamic quantitative score D(entity_A) determined based on the current temperature condition is set to 0.6.

[0134] The confidence weight is the weighted result of the discrimination possibility and the dynamic quantitative score. For the linkage relationship entity_A, the confidence weight W(entity_A)=(P(entity_A)+D(entity_A)) / 2=(0.4+0.6) / 2=0.5 is set to be calculated by using the exemplary weighted average. The confidence weight of each linkage relationship entity is calculated in this way.

[0135] After calculating the confidence weights of all linkage relationship entities, the first linkage relationship entity is determined from the linkage relationship entities based on these confidence weights. For example, assume that there are three linkage relationship entities entity_A, entity_B, and entity_C, and their confidence weights are W(entity_A)=0.5, W(entity_B)=0.3, and W(entity_C)=0.4 respectively. Since the confidence weight of the linkage relationship entity entity_A is the highest, the linkage relationship entity entity_A is determined to be the first linkage relationship entity.

[0136] Taking the first linkage relationship entity (such as the threshold current change of the laser mentioned above) as an example, consider the temporal commonality measurement value between the first state adjustment decision knowledge (temperature adjustment decision knowledge) and the device performance feedback semantics (threshold current of the laser) represented by the first linkage relationship entity. Set the temperature monitoring time series in the temperature adjustment decision knowledge to T={t_1, t_2,..., t_n}, where t_i represents the time point of the (i)th monitoring. The time series of the threshold current change of the laser is I={current_1, current_2,..., current_n}, where current_j represents the threshold current value measured for the jth time.

[0137] The dynamic time warping (DTW) algorithm is used to calculate the timing commonality metric Sim between the two. First, a distance matrix D is constructed, where D_{ij} represents the distance metric between t_i and current_j (for example, the Euclidean distance d_Euclidean=sqrt{(t_i-i_j)^2} can be used). Then, the optimal path is searched in the distance matrix through the dynamic programming algorithm, and the cumulative distance d_summation is calculated. Finally, d_summation is normalized by the function f(d_summation) to obtain the timing commonality metric Sim. Set Sim=0.6 after calculation.

[0138] The state adjustment decision knowledge is optimized in the direction of increasing the temporal commonality metric between the first state adjustment decision knowledge (temperature adjustment decision knowledge) and the device performance feedback semantics (threshold current of the laser) represented by the first linkage relationship entity. For example, the TEC control parameters in the temperature adjustment decision knowledge, such as the cooling power adjustment step s (set to an initial value of 0.5 watts) and the proportional coefficient k of the temperature control (set to an initial value of 0.3).

[0139] If you want to increase the timing commonality metric value Sim, you can adjust these parameters according to the value of Sim. An exemplary adjustment strategy is that for the cooling power adjustment step s, the new s' is calculated by the formula s'=s×(1+Sim), and s'=0.5×(1+0.6)=0.8 watts is obtained. For the proportional coefficient k of temperature control, the new k' is calculated by the formula k'=k×(1+Sim)+0.1 (0.1 is added in this embodiment to introduce a small offset adjustment, and the specific value can be determined based on actual experience), and k'=0.3×(1+0.6)+0.1=0.58 is obtained.

[0140] The device performance feedback semantics also need to be optimized. Since the device performance feedback semantics may have measurement errors or incompleteness, in this optimization process, adjustments can be made according to the optimization of the state regulation decision knowledge. For example, for the threshold current measurement of the laser, if the original measurement accuracy is pm0.05mA, the measurement accuracy can be improved according to the optimization of the temperature regulation decision knowledge. If the threshold current change is expected to be more stable according to the new temperature regulation parameters, the measurement accuracy can be improved to pm0.03mA. At the same time, the expected range of threshold current changes can also be adjusted according to the state regulation decision knowledge. If the threshold current is originally expected to vary in the range of 10-10.5mA, the expected range can be adjusted to 10.1-10.4mA according to the new temperature regulation strategy.

[0141] It can be seen that in terms of determining the first linkage relationship entity, by calculating the discrimination possibility and confidence weight, the key linkage relationship entity can be accurately found. For example, when considering the relationship between temperature and optical power regulation, the laser threshold current change is accurately determined as the core upstream relationship entity. In terms of synchronous optimization, the state regulation decision knowledge is optimized based on the timing commonality metric, such as adjusting the TEC control parameters in temperature regulation, which can improve the accuracy and adaptability of regulation. At the same time, optimizing the device performance feedback semantics, such as improving the laser threshold current measurement accuracy and adjusting the expected range, helps to improve the overall performance, stability and reliability of the optoelectronic module state regulation, and reduce the risk of performance degradation due to improper state regulation.

[0142] Further, a first linkage relationship entity is determined from the linkage relationship entities, the first linkage relationship entity represents a terminal relationship entity of a first state regulation decision knowledge and a core upstream relationship entity of a terminal relationship entity representing a second state regulation decision knowledge, the first state regulation decision knowledge is the state conduction influence knowledge of the second state regulation decision knowledge, the second state regulation decision knowledge is the state conduction influence knowledge of the first state regulation decision knowledge, and the first state regulation decision knowledge and the second state regulation decision knowledge both belong to the state regulation decision knowledge; in the direction of increasing the temporal commonality metric value between the first state regulation decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity, on the basis of synchronously optimizing the state regulation decision knowledge of the several attention dimensions and the X device performance feedback semantics, the method also includes one of the following three embodiments.

[0143] In embodiment 1, the method further includes: determining a second linkage relationship entity from the linkage relationship entity, the second linkage relationship entity being a core upstream relationship entity of a terminal relationship entity representing the first state regulation decision knowledge, a terminal relationship entity representing the second state regulation decision knowledge, and a terminal relationship entity representing a third state regulation decision knowledge, the third state regulation decision knowledge not belonging to the state conduction influence knowledge of the first state regulation decision knowledge, but belonging to the state conduction influence knowledge of the second state regulation decision knowledge; synchronously optimizing the state regulation decision knowledge of the several attention dimensions and the X device performance feedback semantics in a direction of weakening at least one of the following temporal commonality metrics: the temporal commonality metric between the first state regulation decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity; the temporal commonality metric between the second state regulation decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity; the temporal commonality metric between the third state regulation decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity.

[0144] In embodiment 2, the method further includes: synchronously optimizing the state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics in a direction of weakening the temporal commonality metric value of the state adjustment decision knowledge represented by the terminal relationship entity and the device performance feedback semantics represented by the top-level relationship entity, wherein the top-level relationship entity is one of the linkage relationship entities.

[0145] In embodiment 3, the method further includes: synchronously optimizing the state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics in a direction of weakening the temporal commonality metric value between the state adjustment decision knowledge represented by any two of the terminal relationship entities.

[0146] In the previous technical solution, a method for finding a first linkage relationship entity from linkage relationship entities has been determined, and this first linkage relationship entity has a specific relationship with the first and second state adjustment decision knowledge, and by increasing the temporal commonality metric between the first state adjustment decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity, the state adjustment decision knowledge of several attention dimensions and X device performance feedback semantics are simultaneously optimized. On this basis, the following is a detailed description of three embodiments.

[0147] Embodiment 1:

[0148] In the state regulation system of the photovoltaic module, the second linkage relationship entity is to be determined from the linkage relationship entity. This second linkage relationship entity has a special relationship definition. It is the core upstream relationship entity of the terminal relationship entity representing the first state regulation decision knowledge, the terminal relationship entity representing the second state regulation decision knowledge, and the terminal relationship entity representing the third state regulation decision knowledge. Among them, the third state regulation decision knowledge does not belong to the state conduction influence knowledge of the first state regulation decision knowledge, but belongs to the state conduction influence knowledge of the second state regulation decision knowledge.

[0149] For example, if the first state adjustment decision knowledge is set as temperature adjustment knowledge, its terminal relationship entity may involve the control parameters of the thermoelectric cooler (TEC); the second state adjustment decision knowledge is optical power adjustment knowledge, and its terminal relationship entity includes the parameters in the automatic power control (APC) algorithm; the third state adjustment decision knowledge is bias current adjustment knowledge, and its terminal relationship entity is related to the set value of the bias current, etc. Then the second linkage relationship entity may be an entity that has a common upstream influence relationship with these state adjustment decision knowledge in the device performance feedback semantics, such as the linkage relationship entity related to the internal quantum efficiency of the laser (because the internal quantum efficiency will be affected by temperature, optical power, bias current, etc.).

[0150] The method for determining the second linkage relationship entity may involve analyzing and evaluating each linkage relationship entity. Similar to determining the first linkage relationship entity, it is necessary to quantitatively analyze the relationship between each linkage relationship entity and the terminal relationship entities of the first, second, and third state adjustment decision knowledge.

[0151] Assume that there are multiple linkage relationship entities (E_1, E_2, E_3, ...), and for each linkage relationship entity E_i, calculate the quantitative index of the relationship between it and the terminal relationship entity of the first, second, and third state adjustment decision knowledge. For example, taking E_1 as an example, calculate the quantitative index of the relationship between it and the temperature adjustment knowledge terminal relationship entity (TEC control parameter related), the optical power adjustment knowledge terminal relationship entity (APC parameter related), and the bias current adjustment knowledge terminal relationship entity (bias current setting value related). If a relationship function R (E_1) is established based on the theoretical model and historical data of the optoelectronic module, it contains a weighted combination of parameters such as the influence coefficient of temperature on internal quantum efficiency, the influence coefficient of optical power on internal quantum efficiency, and the influence coefficient of bias current on internal quantum efficiency (set R (E_1) = 0.3 × a_1 + 0.4 × a_2 + 0.3 × a_3; a_1, a_2, a_3 are coefficients related to temperature, optical power, and bias current, respectively). By calculating such a relationship function value for each linkage relationship entity, the linkage relationship entity with the largest relationship function value is found as the second linkage relationship entity. Assuming that after calculation, the relationship function value of E_2 is the largest, then E_2 is determined to be the second linkage relationship entity.

[0152] For the temporal commonality metric between the first state adjustment decision knowledge (temperature adjustment knowledge) and the device performance feedback semantics (laser internal quantum efficiency) represented by the second linkage relationship entity (such as the linkage relationship entity related to the laser internal quantum efficiency), a suitable algorithm is used for calculation. The dynamic time warping (DTW) algorithm is assumed to be used, taking the temperature monitoring time series T={t_1, t_2,..., t_n} in the temperature adjustment knowledge and the measurement time series Q={q_1, q_2,..., q_n} of the laser internal quantum efficiency as examples.

[0153] First, construct the distance matrix D, where D_{ij} represents the distance metric between t_i and q_j. Then, use the dynamic programming algorithm to search for the optimal path in the distance matrix, calculate the cumulative distance d_summation, and finally use the normalization function f(d_summation) to obtain the time series commonality metric Sim_1. Set Sim_1=0.7 after calculation.

[0154] Similarly, the timing commonality metric value Sim_2 between the second state adjustment decision knowledge (optical power adjustment knowledge) and the device performance feedback semantics represented by the second linkage relationship entity, and the timing commonality metric value Sim_3 between the third state adjustment decision knowledge (bias current adjustment knowledge) and the device performance feedback semantics represented by the first linkage relationship entity are also calculated according to the above method. Sim_2=0.6, Sim_3=0.5 are set.

[0155] The optimization direction is to weaken at least one of these timing commonality metrics. For example, for the timing commonality metric Sim_1 between the first state adjustment decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity, it can be weakened by adjusting the parameters in the temperature adjustment knowledge. The proportional coefficient k of the temperature control in the temperature adjustment knowledge is set (the initial value is 0.3), and the formula k'=k×(1-Sim_1) can be used for adjustment, and k'=0.3×(1-0.7)=0.09 is obtained.

[0156] For the temporal commonality metric value Sim_2 between the second state adjustment decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity, if the optical power monitoring frequency f in the optical power adjustment knowledge (the initial value is once every 10 minutes) can be adjusted according to the value of Sim_2. For example, using the formula f'=f×(1-Sim_2), we get f'=10×(1-0.6)=4 minutes.

[0157] For the timing commonality measurement value Sim_3 between the device performance feedback semantics represented by the third state adjustment decision knowledge and the first linkage relationship entity, the bias current adjustment step s (initial value is 0.1mA) in the bias current adjustment knowledge is set, and adjusted by the formula s'=s×(1-Sim_3), and s'=0.1×(1-0.5)=0.05mA is obtained.

[0158] In this process, by adjusting the relevant parameters in the state regulation decision knowledge, the state regulation decision knowledge of several attention dimensions and the X device performance feedback semantics are simultaneously optimized in the direction of weakening the specific temporal commonality metric value.

[0159] Embodiment 2:

[0160] The top-level relationship entity is determined in the linkage relationship entity. This top-level relationship entity has a special status in the entire relationship system of state adjustment decision knowledge and device performance feedback semantics. For example, the top-level relationship entity may be a representation entity of device performance feedback semantics that is most directly related to the overall performance of the photovoltaic module. Assuming that in the photovoltaic module, the total detection efficiency of the photodetector is a very critical performance indicator, then the linkage relationship entity related to the total detection efficiency of the photodetector may be defined as the top-level relationship entity.

[0161] The relationship between the state adjustment decision knowledge represented by the terminal relationship entity and the device performance feedback semantics represented by the top-level relationship entity needs to be quantified. For example, the state adjustment decision knowledge represented by the terminal relationship entity is set as the optical power adjustment knowledge, which has a certain relationship with the top-level relationship entity (the linkage relationship entity related to the total detection efficiency of the photoelectric detector). According to the theoretical model of the optoelectronic module, there is a functional relationship between the optical power and the total detection efficiency of the photoelectric detector, P(E)=b×P_{out}+c, where P(E) represents the total detection efficiency of the photoelectric detector, P_{out} represents the optical output power, and b and c are coefficients determined according to the module characteristics (set b=0.5, c=0.1).

[0162] Calculate the temporal commonality metric of the state adjustment decision knowledge represented by the terminal relation entity and the device performance feedback semantics represented by the top-level relation entity. Set up a DTW algorithm similar to the previous one, taking the optical power adjustment time series P={p_1, p_2, ..., p_n} in the optical power adjustment knowledge and the measurement time series E={e_1, e_2, ..., e_n} of the total detection efficiency of the photodetector as examples.

[0163] Construct the distance matrix D, calculate the cumulative distance d_summation, and obtain the time series commonality metric Sim_4 through the normalization function. Set Sim_4=0.8 after calculation.

[0164] The state adjustment decision knowledge and device performance feedback semantics are optimized in the direction of weakening this timing commonality metric value Sim_4. For the optical power adjustment parameters in the optical power adjustment knowledge, such as the step size s of the optical power adjustment (the initial value is 0.2dBm), it can be adjusted according to the value of Sim_4. Using the formula s'=s×(1-Sim_4), we get s'=0.2×(1-0.8)=0.04dBm.

[0165] For device performance feedback semantics, such as the measurement accuracy of the total detection efficiency of the photodetector, if the original measurement accuracy is pm0.02, it can be adjusted according to the optimization of the state adjustment decision knowledge. Since the timing commonality metric value with the optical power adjustment knowledge needs to be weakened, it may mean reducing the impact of optical power on the detection efficiency measurement, so the measurement accuracy can be improved to pm0.01.

[0166] Embodiment 3:

[0167] In the state regulation of the photovoltaic module, consider the state regulation decision knowledge represented by any two terminal relationship entities. For example, one terminal relationship entity represents the temperature regulation knowledge, and the other terminal relationship entity represents the bias current regulation knowledge. These two state regulation decision knowledge have a certain relationship in the operation process of the photovoltaic module. The change of temperature may affect the performance of the device, thereby affecting the optimal setting of the bias current; conversely, the adjustment of the bias current may also affect the temperature characteristics of the device.

[0168] Calculate the temporal commonality measure between the state adjustment decision knowledge represented by the two terminal relationship entities. Assume that the DTW algorithm is used, taking the temperature monitoring time series T={t_1, t_2, ..., t_n} in the temperature adjustment knowledge and the bias current adjustment time series I={current_1, current_2, ..., current_n} in the bias current adjustment knowledge as examples.

[0169] Construct the distance matrix D, calculate the cumulative distance d_summation, and obtain the time series commonality metric Sim_5 through the normalization function. Set Sim_5=0.6 after calculation.

[0170] The state regulation decision knowledge and device performance feedback semantics are optimized in the direction of weakening this timing commonality metric Sim_5. For the TEC control parameters in the temperature regulation knowledge, such as the cooling power adjustment step s (initial value is 0.5 watts), it can be adjusted according to the value of Sim_5. Using the formula s'=s×(1-Sim_5), we get s'=0.5×(1-0.6)=0.2 watts.

[0171] The bias current adjustment step s_1 (initial value is 0.1 mA) in the bias current adjustment knowledge can also be adjusted according to Sim_5. For example, using the formula s_1'=s_1×(1-Sim_5), s_1'=0.1×(1-0.6)=0.04 mA is obtained.

[0172] For device performance feedback semantics, such as device performance indicators related to temperature and bias current, such as the noise figure of the device, if the original noise figure measurement range is 1-2dB, it can be adjusted according to the optimization of the state adjustment decision knowledge. Since the timing commonality measurement value between the temperature adjustment knowledge and the bias current adjustment knowledge needs to be weakened, it may mean reducing the impact of the mutual influence between the two on the noise figure measurement, so the measurement range can be adjusted to 1.2-1.8dB.

[0173] In summary, Example 1 can accurately adjust multiple parameters in the state adjustment decision knowledge, such as temperature control proportional coefficient, optical power monitoring frequency, bias current adjustment step, etc., by determining the second linkage relationship entity and optimizing in the direction of weakening the commonality metric value of a specific timing, while optimizing the device performance feedback semantics, such as the measurement accuracy of the internal quantum efficiency of the laser, etc., which helps to finely adjust the state of the optoelectronic module and reduce unnecessary correlation effects.

[0174] Example 2 can improve the independence of state adjustment decision knowledge and key device performance feedback semantics by weakening the temporal commonality measurement value of the state adjustment decision knowledge represented by the terminal relationship entity and the device performance feedback semantics represented by the top-level relationship entity, optimize parameters such as optical power adjustment step size and detector detection efficiency measurement accuracy, and improve overall performance.

[0175] Embodiment 3 optimizes the timing commonality measurement value between the state adjustment decision knowledge represented by any two terminal relationship entities, which can reduce the mutual interference between different state adjustment decision knowledge, optimize TEC cooling power, bias current adjustment step and device noise coefficient measurement range, and improve stability and reliability.

[0176] On the basis of the above embodiment, the method also includes: obtaining the prior state adjustment strategy of the reference photoelectric module; analyzing the temporal commonality measurement value between the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions based on the long short-term memory model, and determining the state adjustment training strategy of the reference photoelectric module; debugging the long short-term memory model according to the prior state adjustment strategy of the reference photoelectric module and the state adjustment training strategy of the reference photoelectric module.

[0177] In the state regulation system of photovoltaic modules, the prior state regulation strategy refers to the methods and solutions for photovoltaic module state regulation that already exist in previous operations or experience. These strategies may be based on previous engineering practices, theoretical research or preliminary experimental results. For example, in a photovoltaic module system that has been running for a period of time, there may be an a priori state regulation strategy for temperature regulation, that is, when the ambient temperature is within a certain range, such as 0-40°C, a thermoelectric cooler (TEC) is used to cool or heat at a fixed power (such as 3 watts) to keep the internal temperature of the photovoltaic module in a relatively stable range (such as 25°C±2°C).

[0178] For optical power regulation, the prior state regulation strategy may be to set the initial parameters of automatic power control (APC), such as a proportional coefficient of 0.5, an integral time constant of 0.1 seconds, and an optical power monitoring frequency of once every 10 minutes. These prior state regulation strategies form the basis for subsequent analysis and optimization.

[0179] There are many ways to obtain the prior state adjustment strategy. One way is to read it from the existing system configuration file. In the control system of the photovoltaic module, various previously set adjustment parameters and strategies are usually saved, which can be directly obtained as the prior state adjustment strategy. For example, in the storage system, there is a special configuration file that records the parameter settings of various aspects such as temperature adjustment, optical power adjustment, and bias current adjustment. Another way is to query the historical operation records. If there are detailed log records of the operation process of the photovoltaic module, the state adjustment strategy previously adopted can be analyzed from these logs. For example, it can be found from the historical log that in a certain period of time in the past, for specific environmental conditions, the set value of the bias current and its adjustment frequency and other information all belong to the category of the prior state adjustment strategy.

[0180] Next, we will discuss the technical idea of ​​determining the state adjustment training strategy based on the long short-term memory model (LSTM) to analyze the commonality measurement values ​​of time series.

[0181] 1. Introduction to Long Short-Term Memory Model

[0182] The long short-term memory model is a special recurrent neural network (RNN) that aims to solve the problem of gradient vanishing or gradient exploding when traditional RNN processes long sequence data. LSTM controls the flow of information by introducing memory cells (Cell), input gates (InputGate), forget gates (ForgetGate) and output gates (OutputGate).

[0183] In the state regulation analysis of photovoltaic modules, memory cells can be used to store long-term information about the operating scenario description knowledge and state regulation decision knowledge. For example, for the temperature change sequence of the photovoltaic module (part of the operating scenario description knowledge), the LSTM memory cell can remember the temperature change trend over a longer time range, not just the short-term fluctuations. The input gate determines how new information (such as the temperature measurement value at the current moment or the new state regulation decision knowledge parameter) enters the memory cell, the forget gate controls the degree of retention of old information in the memory cell, and the output gate determines how the information in the memory cell affects the current output (such as the calculation result of the temporal commonality metric).

[0184] 2. Analyzing the commonality of time series

[0185] In order to use LSTM to analyze the temporal commonality measurement between the operating scenario description knowledge of the reference optoelectronic module and the state regulation decision knowledge of several attention dimensions, the data needs to be preprocessed first. The data in the operating scenario description knowledge may include the sequence of ambient temperature over time, the length of the optical fiber link (if it changes over time, such as in a reconfigurable optical network), the change of the transmission data rate requirement over time, etc. These data are normalized, for example, the temperature value is mapped to the interval [0, 1], and the transmission data rate requirement is also normalized similarly.

[0186] For state adjustment decision knowledge, such as TEC control parameters in temperature adjustment and APC parameters in optical power adjustment, they are also converted into a format suitable for LSTM input. For example, the cooling power and the proportional coefficient of temperature control in the TEC control parameters are encoded into vector form according to certain rules. The parameter vector in the temperature adjustment decision knowledge is set to vec{T}=[t_1, t_2, ..., t_n], where t_1 represents the cooling power and t_2 represents the proportional coefficient, and these parameters are normalized so that their numerical range is suitable for the input requirements of LSTM.

[0187] Construct an LSTM model and determine the structure of its input layer, hidden layer, and output layer. For example, the number of neurons in the input layer is determined according to the dimension of the input data. Assuming that the total input dimension of the running scenario description knowledge and the state adjustment decision knowledge after preprocessing is m, the number of neurons in the input layer is m. The number of neurons in the hidden layer can be determined based on experience or experiments, and is set to h neurons. The output layer is used to output the time series commonality measurement value, so the number of neurons in the output layer is 1.

[0188] When training the LSTM model, existing data samples are used. Paired data of the running scenario description knowledge and the state adjustment decision knowledge are used as input samples, and the corresponding time series commonality measurement values ​​(pre-calculated by other methods, such as using the dynamic time warping (DTW) algorithm to calculate the time series commonality measurement values ​​of a part of the samples as labels) are used as output samples. Use a suitable loss function, such as the mean square error (MSE) function L=frac{1}{N}sum_{i=1}^{N}(y_i-hat{y}_i)^2, where y_i is the true time series commonality measurement value, hat{y}_i is the time series commonality measurement value predicted by the LSTM model, and N is the number of samples. The weights of the LSTM model are adjusted through the back-propagation algorithm to minimize the loss function, thereby training an LSTM model that can accurately predict the time series commonality measurement value.

[0189] Once the LSTM model is trained, it can be used to determine the state regulation training strategy for the reference optoelectronic module. For example, for different operating scenario description knowledge (such as different ambient temperature change trends, different transmission data rate requirement changes, etc.), the LSTM model can predict the temporal commonality metrics between the state regulation decision knowledge of each attention dimension. Based on these predicted temporal commonality metrics, the parameters in the state regulation decision knowledge can be adjusted to form a state regulation training strategy.

[0190] Taking optical power regulation as an example, it is assumed that the LSTM model predicts that when the ambient temperature fluctuates periodically (such as sinusoidal fluctuations with a period of 24 hours and an amplitude of 10°C) and the transmission data rate is required to be high during the day and low at night (this operating scenario description knowledge), the APC proportional coefficient in the optical power regulation decision knowledge should be adjusted to 0.6, and the optical power monitoring frequency should be increased to once every 8 minutes. This constitutes an optical power regulation training strategy for this specific operating scenario.

[0191] 3. Debugging the LSTM model based on the prior state adjustment strategy and state adjustment training strategy

[0192] Although the LSTM model has certain capabilities in analyzing common metrics of time series and determining state adjustment training strategies, the initially constructed and trained models may have some inaccuracies or be inappropriate for actual situations. The prior state adjustment strategy contains experience from actual projects and existing effective adjustment methods. Combining it with the state adjustment training strategy obtained through the LSTM model can debug the LSTM model to make it more accurate and practical.

[0193] One debugging method is to adjust the weights of the LSTM model. For example, the parameters in the state adjustment training strategy obtained by the LSTM model are compared with the known effective parameter settings in the prior state adjustment strategy. If the temperature adjustment parameters in the prior state adjustment strategy (such as TEC cooling power of 3 watts) are proven to be very effective in actual applications in a specific operating scenario, and the TEC cooling power in the temperature adjustment training strategy predicted by the LSTM model is 2.5 watts, then the weights related to temperature adjustment in the LSTM model can be adjusted through a certain algorithm. A simple weighted adjustment method is set to adjust the weights according to the difference ratio between the two. Let the prior cooling power be P_{prior}, the cooling power predicted by the LSTM model be P_{LSTM}, and the weight adjustment factor α=frac{P_{prior}-P_{LSTM}}{P_{prior}}. If the weight related to cooling power in the LSTM model is β, then the adjusted weight β'=β×(1+α).

[0194] Another debugging method is to adjust the input data preprocessing method of the LSTM model. If there is a large difference between some parameter settings in the prior state adjustment strategy and the results obtained by the LSTM model based on the input data, it may be because some information is lost or converted irrationally during the input data preprocessing process. For example, the prior state adjustment strategy is very sensitive to ambient temperature, and the LSTM model may not fully reflect this sensitivity when processing temperature data due to problems with the normalization method. Then you can adjust the normalization range of the temperature data or use other preprocessing methods, such as performing differential processing on the temperature data before inputting it into the LSTM model, to improve the model's adaptability to key information in the prior state adjustment strategy.

[0195] It can be understood that obtaining a priori state adjustment strategies provides an empirical and practical basis for the entire state adjustment process. Based on LSTM analysis of time series commonality metrics, the state adjustment training strategy is determined. By taking advantage of LSTM's ability to process long sequence data, the adjustment strategy can be adjusted more accurately based on the relationship between the operating scenario and the state adjustment decision knowledge. For example, in optical power adjustment, the APC parameters can be adjusted according to the environment and transmission requirements. The LSTM model is debugged based on a priori and training strategies, and the model weights or input preprocessing methods are adjusted in combination with prior experience to improve the model accuracy and practicality, optimize state adjustment, and enhance the performance and stability of optoelectronic modules.

[0196] Based on step 202 to step 206, the method further includes: obtaining a target photoelectric module; and determining a current state adjustment strategy of the target photoelectric module based on a temporal commonality measurement value of the target photoelectric module and state adjustment decision knowledge of several attention dimensions.

[0197] The target photovoltaic module is the object of state adjustment. In the entire photovoltaic system, it is determined by a specific identification or screening mechanism. This may be based on factors such as the system topology, the functional characteristics of the device, or the operating status.

[0198] For the target photovoltaic module, operating data such as temperature, optical power, and bias current are collected. Taking temperature as an example, assume that the current temperature of the target photovoltaic module is 20°C. At the same time, the state adjustment decision knowledge of several attention dimensions is sorted out. For example, the control parameters of the thermoelectric cooler (TEC) in the temperature adjustment decision knowledge, such as the cooling power adjustment step size of 0.3 watts and the temperature control proportional coefficient of 0.4.

[0199] The dynamic time warping (DTW) algorithm is used for calculation. For example, for the temperature-related timing commonality metric, the temperature time series of the target photovoltaic module and the cooling power adjustment time series in the temperature regulation decision knowledge are calculated. A distance matrix is ​​constructed, the distance between elements is calculated using the Euclidean distance, the optimal path is searched through dynamic programming to calculate the cumulative distance, and then the timing commonality metric is obtained by normalization, which is assumed to be 0.5.

[0200] Temperature adjustment strategy: Determine the temperature adjustment strategy based on the above time series commonality measurement value. For example, the cooling power adjustment step is adjusted according to the formula. Assume that the new cooling power adjustment step is the original step multiplied by (1 + time series commonality measurement value), that is, 0.3 × (1 + 0.5) = 0.45 watts. The temperature control proportional coefficient can also be adjusted similarly.

[0201] Optical power regulation strategy: If the optical power monitoring frequency in the optical power regulation decision knowledge is initially once every 15 minutes, and the timing commonality metric value related to optical power is 0.4, the new optical power monitoring frequency is the original frequency multiplied by (1-timing commonality metric value), that is, 15×(1-0.4)=once every 9 minutes.

[0202] Bias current adjustment strategy: Assuming that the bias current adjustment step size in the bias current adjustment decision knowledge is initially 0.2 mA, and the timing commonality metric value related to the bias current is 0.3, the new bias current adjustment step size is 0.2×(1-0.3)=0.14 mA.

[0203] With this design, the regulation strategy is determined based on the timing commonality metric value. Through numerical calculations such as specific adjustments in temperature, optical power, and bias current regulation, the regulation strategy can be matched to the actual situation of the target optoelectronic module, improve its performance stability, ensure the normal and efficient operation of the optoelectronic module, and reduce problems caused by mismatch.

[0204] In some independent embodiments, after determining the current state adjustment strategy of the target photoelectric module based on the temporal commonality measurement value of the target photoelectric module and the state adjustment decision knowledge of several attention dimensions, the method also includes: storing the current operation data record of the target photoelectric module in association with the current state adjustment strategy.

[0205] Furthermore, the current operation data record of the target optoelectronic module is associated with the current state adjustment strategy and stored, including: obtaining a module operation text data stream, wherein the module operation text data stream includes a plurality of module operation text data with a time sequence relationship; obtaining an adjustment annotation text data stream according to the module operation text data stream and the policy pointing description feature of the current state adjustment strategy, wherein the adjustment annotation text data stream includes a plurality of adjustment annotation text data with a time sequence relationship; based on the module operation text data stream, obtaining a module operation semantic entity topology stream through a first knowledge entity mapping algorithm included in an associated storage processing network, wherein the module operation semantic entity topology stream The body topology flow includes several modules running semantic entity topologies; based on the adjusted annotation text data flow, the adjusted annotation semantic entity topology flow is obtained through the second knowledge entity mapping algorithm included in the associated storage processing network, wherein the adjusted annotation semantic entity topology flow includes several adjusted annotation semantic entity topologies; based on the module running semantic entity topology flow and the adjusted annotation semantic entity topology flow, the storage mode matching score corresponding to the module running text data flow is obtained through the storage mode matching algorithm included in the associated storage processing network; according to the target storage mode corresponding to the storage mode matching score, the module running text data flow and the current state adjustment strategy are associated and stored.

[0206] Under the following innovative design idea, based on the module running semantic entity topology flow and the adjustment and annotation semantic entity topology flow, the storage mode matching score corresponding to the module running text data flow is obtained through the storage mode matching algorithm included in the associated storage processing network, including: based on the module running semantic entity topology flow, a plurality of first one-hot linear codes are obtained through the first one-hot semantic coding layer included in the associated storage processing network, wherein each first one-hot linear code corresponds to a module running semantic entity topology; based on the adjustment and annotation semantic entity topology flow, a plurality of second one-hot linear codes are obtained through the second one-hot semantic coding layer included in the associated storage processing network, wherein each second one-hot linear code corresponds to an adjustment and annotation semantic entity topology; the plurality of first one-hot linear codes and the plurality of second one-hot linear codes are fused to obtain a plurality of target one-hot linear codes, wherein each target one-hot linear code includes a first one-hot linear code and a second one-hot linear code; based on the plurality of target one-hot linear codes, the storage mode matching score corresponding to the module running text data flow is obtained through the storage mode matching algorithm included in the associated storage processing network.

[0207] Furthermore, the method of running a semantic entity topology stream based on the module and obtaining a plurality of first one-hot linear codes through a first one-hot semantic coding layer included in the associated storage processing network includes: running a semantic entity topology for each module in the module-running semantic entity topology stream and obtaining a first local down-sampling semantic entity topology through a local down-sampling layer included in the first one-hot semantic coding layer, wherein the first one-hot semantic coding layer belongs to the associated storage processing network; running a semantic entity topology for each module in the module-running semantic entity topology stream and obtaining a first global down-sampling semantic entity topology through a global down-sampling layer included in the first one-hot semantic coding layer; running a semantic entity topology for each module in the module-running semantic entity topology stream and obtaining a first integrated semantic entity topology through a coding feature integration layer included in the first one-hot semantic coding layer based on the first local down-sampling semantic entity topology and the first global down-sampling semantic entity topology; running a semantic entity topology for each module in the module-running semantic entity topology stream and obtaining a first one-hot linear code through a first global down-sampling layer included in the first one-hot semantic coding layer based on the first integrated semantic entity topology and the module-running semantic entity topology.

[0208] In other embodiments of the present invention, after determining the current state adjustment strategy of the target photovoltaic module, the core content of this embodiment is to associate and store the current operation data record of the target photovoltaic module with the current state adjustment strategy. This process involves multiple steps, through a series of algorithms and processing flows, to ensure the effective storage of data and the subsequent queryability and analyzability.

[0209] Among them, the module operation text data stream is a textual description of the operation process of the target optoelectronic module, which contains several module operation text data with a time sequence relationship. These data may include records of the working state parameters and operating environment parameters of each component of the optoelectronic module (such as lasers, detectors, etc.) at different time points. For example, at a certain moment, the recorded laser output power is 3dBm, the detector responsivity is 0.8A / W, the ambient temperature is 25°C, and other information is arranged in chronological order to form a module operation text data stream.

[0210] Next, the adjustment annotation text data stream is obtained according to the module operation text data stream and the policy pointing description feature of the current state adjustment strategy. The current state adjustment strategy includes policy contents such as temperature adjustment strategy (for example, the cooling power adjustment step of the thermoelectric cooler (TEC), the temperature control proportional coefficient, etc.), optical power adjustment strategy (such as the proportional coefficient of the automatic power control (APC), the optical power monitoring frequency, etc.), and bias current adjustment strategy (bias current adjustment step, etc.).

[0211] The strategy-directed description feature is a descriptive representation of the content of these strategies, and is used to guide the extraction of adjustment-related information from the module operation text data stream. For example, if the temperature adjustment strategy mentions increasing the cooling power of the TEC when the temperature rises, then in the module operation text data stream, when there is a record of temperature rise, the corresponding adjustment annotation text data may be "temperature rises, and the TEC cooling power needs to be adjusted according to the strategy." In this way, according to the module operation text data stream and the strategy-directed description feature, a number of adjustment annotation text data with a time sequence relationship are formed to form an adjustment annotation text data stream.

[0212] Then, based on the module operation text data stream, the module operation semantic entity topology stream is obtained through the first knowledge entity mapping algorithm included in the associated storage processing network. The first knowledge entity mapping algorithm is an algorithm that converts text data into semantic entity topology. For example, for the laser output power data in the module operation text data stream, the first knowledge entity mapping algorithm may map it into an entity topology structure containing semantic elements such as "laser", "output power", and "value". For each data item in the entire module operation text data stream, after being processed by the first knowledge entity mapping algorithm, a number of module operation semantic entity topologies are obtained, thereby constituting a module operation semantic entity topology stream.

[0213] In addition, based on the adjustment annotation text data stream, the adjustment annotation semantic entity topology stream is obtained through the second knowledge entity mapping algorithm included in the associated storage processing network. Similar to the first knowledge entity mapping algorithm, the second knowledge entity mapping algorithm converts the adjustment annotation text data into a semantic entity topology. For example, for the "temperature rises, the TEC cooling power needs to be adjusted according to the strategy" in the adjustment annotation text data, the second knowledge entity mapping algorithm may map it into an adjustment annotation semantic entity topology containing semantic elements such as "temperature", "increase", "TEC", "cooling power", and "adjustment". By processing each data item in the adjustment annotation text data stream, a number of adjustment annotation semantic entity topologies are obtained to form an adjustment annotation semantic entity topology stream.

[0214] It can be understood that the semantic entity topology flow is run based on the module, and a plurality of first one-hot linear codes are obtained through the first one-hot semantic coding layer included in the associative storage processing network. This process also includes multiple sub-steps.

[0215] For each module in the module running semantic entity topology stream, the first local down-sampling semantic entity topology is first obtained through the local down-sampling layer included in the first one-hot semantic encoding layer. The function of the local down-sampling layer is to perform local data simplification processing on the semantic entity topology, such as reducing some minor semantic elements or merging some semantic elements. Then, the first global down-sampling semantic entity topology is obtained through the global down-sampling layer included in the first one-hot semantic encoding layer. The global down-sampling layer samples the semantic entity topology as a whole to obtain more macroscopic semantic features.

[0216] Next, for each module running semantic entity topology, based on the first local down-sampled semantic entity topology and the first global down-sampled semantic entity topology, the first integrated semantic entity topology is obtained through the encoding feature integration layer included in the first one-hot semantic coding layer. The encoding feature integration layer integrates the features of the semantic entity topologies obtained by local and global downsampling to form a comprehensive semantic entity topology. Finally, based on the first integrated semantic entity topology and the module running semantic entity topology, the first one-hot linear encoding is obtained through the first global down-sampling layer included in the first one-hot semantic coding layer. Each first one-hot linear encoding corresponds to a module running semantic entity topology.

[0217] In detail, based on the adjusted annotated semantic entity topology stream, a plurality of second one-hot linear codes are obtained through a second one-hot semantic coding layer included in the associative storage processing network. This process is similar to obtaining the first one-hot linear code, and each second one-hot linear code corresponds to an adjusted annotated semantic entity topology.

[0218] A plurality of first one-hot linear codes and a plurality of second one-hot linear codes are fused to obtain a plurality of target one-hot linear codes, wherein each target one-hot linear code includes a first one-hot linear code and a second one-hot linear code. This fusion process may be to combine the two codes together according to certain rules, such as simple concatenation or combination according to weights.

[0219] Based on several target one-hot linear codes, the storage pattern matching score corresponding to the text data stream is obtained by running the storage pattern matching algorithm acquisition module included in the associative storage processing network. The storage pattern matching algorithm performs matching calculations with the preset storage pattern according to the characteristics of the target one-hot linear code to obtain a score indicating the degree of matching.

[0220] Finally, according to the target storage mode corresponding to the storage mode matching score, the module operation text data stream and the current state adjustment strategy are associated and stored. Different storage mode matching scores may correspond to different storage methods. For example, if the score is high, it means that the module operation text data stream and the current state adjustment strategy are highly associated, and a closely associated storage method may be used, such as storing the two in adjacent fields in the same database table or using a specific index structure for associated storage; if the score is low, a relatively loose storage method may be used, but still maintain a certain association relationship, so that relevant information can be obtained as needed in subsequent queries and analysis.

[0221] This design associates and stores the current operating data record of the target optoelectronic module with the current state adjustment strategy, which facilitates the subsequent tracing and analysis of the operating status of the optoelectronic module. In the process of obtaining the module operation text data stream and adjusting the annotation text data stream, accurate extraction of relevant information based on the policy-oriented description features helps to deeply understand the relationship between the operating data and the adjustment strategy. Through a series of semantic entity topology transformations and one-hot linear encoding processing, the storage mode matching score can be determined more accurately, so as to select the appropriate associated storage method. This not only improves the rationality and effectiveness of data storage, but also enables faster and more accurate acquisition of required information when querying and analyzing the relationship between the optoelectronic module operation data and the state adjustment strategy, providing strong support for the optimized operation, troubleshooting and performance improvement of the optoelectronic module.

[0222] Further, Figure 2 FIG. 2 is a schematic diagram of a state adjustment optimization system 200 provided by the present invention. Figure 2 The state adjustment optimization system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method of the present invention. Figure 2 As shown, the state adjustment optimization system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method of the present invention. The memory 230 may be a separate device independent of the processor 210, or may be integrated in the processor 210. Optionally, as Figure 2As shown, the state adjustment optimization system 200 may also include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices. Specifically, it may send information or data to other devices, or receive information or data sent by other devices. Optionally, the state adjustment optimization system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or the device deployed with the storage engine in each method of the present invention. For the sake of brevity, it will not be repeated here. It should be understood that the processor of the present invention may be an integrated circuit chip with signal processing capabilities. It can be understood that the memory in the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. It should be noted that the memory of the system and method described herein is intended to include, but is not limited to, a suitable type of memory.

[0223] Based on the above, a readable storage medium is provided, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the above method are implemented.

[0224] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the purpose of the present invention and the scope of protection of the present invention, all of which are within the protection of the present invention.

Claims

1. A method for optimizing the state adjustment of a photovoltaic module, characterized in that: The method is performed by a state adjustment optimization system, and the method comprises: The operation scenario description knowledge, state adjustment decision knowledge of several attention dimensions and X device performance feedback semantics of the reference optoelectronic module are obtained; wherein the state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics are interconnected to generate a knowledge feature relationship spectrum, each of the several attention dimensions corresponds to one or more state adjustment decision knowledge, the state adjustment decision knowledge is characterized by the terminal relationship entity of the knowledge feature relationship spectrum, the device performance feedback semantics is characterized by the linkage relationship entity of the knowledge feature relationship spectrum, and X is a positive integer; wherein the operation scenario description knowledge includes the device type, ambient temperature range, working optical fiber link length, and data transmission rate requirement of the optoelectronic module; the attention dimension includes optical power stability and temperature sensitivity, and the state adjustment decision knowledge corresponding to the optical power stability includes the feedback loop parameter setting of the automatic power control APC and the frequency setting of the optical power monitoring, and the state adjustment decision knowledge corresponding to the temperature sensitivity includes the maximum cooling power of the thermoelectric cooler TEC and the accuracy requirement of the temperature control; the device performance feedback semantics reflects the actual working efficiency of each device in the optoelectronic module, and the devices include lasers, detectors, and amplifiers; Based on the temporal commonality measurement value of the operation scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge of the several attention dimensions, the state adjustment decision knowledge of the several attention dimensions is synchronously optimized; Based on the entity semantic transmission characteristics between the terminal relationship entity and the linkage relationship entity, the state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics are synchronously optimized; wherein the state adjustment decision knowledge of the several attention dimensions after synchronous optimization is used to determine the current state adjustment strategy of the target optoelectronic module, and the current state adjustment strategy includes at least one of a temperature adjustment strategy, an optical power adjustment strategy, a bias current adjustment strategy, and a signal modulation optimization strategy. The temperature adjustment strategy is used to adjust the cooling power adjustment step and the temperature control proportional coefficient of the thermoelectric cooler TEC; the optical power adjustment strategy is used to adjust the proportional coefficient and the optical power monitoring frequency of the automatic power control APC; the X device performance feedback semantics after synchronous optimization are used for state adjustment verification; The synchronous optimization of the state adjustment decision knowledge of the several attention dimensions based on the temporal commonality measurement value of the operation scenario description knowledge of the reference optoelectronic module and the state adjustment decision knowledge of the several attention dimensions includes: Based on the temporal commonality measurement value of the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions, determine the target state adjustment decision knowledge, wherein the target state adjustment decision knowledge is the state adjustment decision knowledge with the largest temporal commonality measurement value with the operation scenario description knowledge of the reference photoelectric module among the state adjustment decision knowledge of the several attention dimensions; Synchronously optimizing the target state adjustment decision knowledge by weighted processing of the temporal commonality measurement values ​​of the operation scenario description knowledge of the reference photovoltaic module and the target state adjustment decision knowledge; The method further comprises: In the direction of converging the distance between the first time series commonality metric value and the second time series commonality metric value, the state adjustment decision knowledge of the several attention dimensions is synchronously optimized, wherein the first time series commonality metric value is the time series commonality metric value of the operation scenario description knowledge of the reference photoelectric module and any one of the state adjustment decision knowledge of the several attention dimensions, and the second time series commonality metric value is the time series commonality metric value of the operation scenario description knowledge of the corrected photoelectric module and any one of the state adjustment decision knowledge of the several attention dimensions, and the corrected photoelectric module is determined after the reference photoelectric module is troubleshooted; The method further includes: synchronously optimizing the operation scenario description knowledge of the reference photovoltaic module and the operation scenario description knowledge of the modified photovoltaic module in the direction of increasing the temporal commonality metric value of the operation scenario description knowledge of the modified photovoltaic module and the operation scenario description knowledge of the reference photovoltaic module; The synchronously optimizing the state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics based on the entity semantic transmission feature between the terminal relationship entity and the linkage relationship entity includes: Determine a first linkage relationship entity from the linkage relationship entities, the first linkage relationship entity represents a terminal relationship entity of a first state adjustment decision knowledge and a core upstream relationship entity of a terminal relationship entity representing a second state adjustment decision knowledge, the first state adjustment decision knowledge is the state conduction influence knowledge of the second state adjustment decision knowledge, the second state adjustment decision knowledge is the state conduction influence knowledge of the first state adjustment decision knowledge, and the first state adjustment decision knowledge and the second state adjustment decision knowledge both belong to the state adjustment decision knowledge; In the direction of increasing the temporal commonality metric value between the first state adjustment decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity, the state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics are simultaneously optimized; Wherein, determining the first linkage relationship entity from the linkage relationship entities includes: Determine the discrimination possibility of each linkage relationship entity, the discrimination possibility being a quantitative score of the linkage relationship entity being a core upstream relationship entity of a terminal relationship entity representing the first state adjustment decision knowledge and a terminal relationship entity representing the second state adjustment decision knowledge; Determine a confidence weight of each linkage relationship entity, wherein the confidence weight is a weighted result of the discrimination possibility and the dynamic quantitative score; determining the first linkage relationship entity from the linkage relationship entities based on the confidence weights of the linkage relationship entities; The method further comprises: Determine a second linkage relationship entity from the linkage relationship entities, the second linkage relationship entity being a core upstream relationship entity of a terminal relationship entity representing the first state regulation decision knowledge, a terminal relationship entity representing the second state regulation decision knowledge, and a terminal relationship entity representing a third state regulation decision knowledge, the third state regulation decision knowledge not belonging to the state conduction influence knowledge of the first state regulation decision knowledge, but belonging to the state conduction influence knowledge of the second state regulation decision knowledge; The state adjustment decision knowledge of the several attention dimensions and the X device performance feedback semantics are simultaneously optimized in the direction of weakening at least one of the following temporal commonality metrics: A temporal commonality metric value between the first state adjustment decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity; A temporal commonality metric value between the second state adjustment decision knowledge and the device performance feedback semantics represented by the second linkage relationship entity; The temporal commonality metric value between the third state adjustment decision knowledge and the device performance feedback semantics represented by the first linkage relationship entity; Among them, for any state adjustment decision knowledge, the step of calculating the first timing commonality metric value includes: calculating the timing commonality metric value Sim_{A1} of the operation scenario description knowledge of the reference optoelectronic module and the optical power adjustment knowledge A according to the DTW algorithm; the step of calculating the second timing commonality metric value includes: calculating the timing commonality metric value Sim_{A2} of the operation scenario description knowledge of the corrected optoelectronic module and the optical power adjustment knowledge A; the optical power adjustment knowledge A is any state adjustment decision knowledge; Among them, after calculating the confidence weights of all linkage relationship entities, the step of determining the first linkage relationship entity from the linkage relationship entities based on the confidence weights includes: assuming that there are three linkage relationship entities entity_A, entity_B, and entity_C; selecting the linkage relationship entity entity_A with the highest confidence weight as the first linkage relationship entity; Among them, the step of determining the second linkage relationship entity includes: setting the existence of multiple linkage relationship entities, and for each linkage relationship entity E_i, calculating the quantitative index of the relationship between it and the terminal relationship entities of the first, second, and third state adjustment decision knowledge; calculating the relationship function value for each linkage relationship entity, and taking the linkage relationship entity with the largest relationship function value as the second linkage relationship entity; wherein the relationship function includes a weighted combination of parameters such as the influence coefficient of temperature on internal quantum efficiency, the influence coefficient of optical power on internal quantum efficiency, and the influence coefficient of bias current on internal quantum efficiency.

2. The method according to claim 1, characterized in that The method further comprises: The state adjustment decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics are simultaneously optimized in a direction of weakening the temporal commonality metric value of the state adjustment decision knowledge represented by the terminal relationship entity and the device performance feedback semantics represented by the top-level relationship entity, wherein the top-level relationship entity is one of the linkage relationship entities.

3. The method according to claim 1, characterized in that The method further comprises: The state regulation decision knowledge of the plurality of attention dimensions and the X device performance feedback semantics are simultaneously optimized in a direction of weakening the temporal commonality metric value between the state regulation decision knowledge represented by any two of the terminal relationship entities.

4. The method according to any one of claims 1 to 3, characterized in that: The method further comprises: Acquiring a priori state adjustment strategy of the reference optoelectronic module; Analyzing the temporal commonality measurement value between the operation scenario description knowledge of the reference photoelectric module and the state adjustment decision knowledge of the several attention dimensions based on the long short-term memory model, and determining the state adjustment training strategy of the reference photoelectric module; The long short-term memory model is debugged according to the prior state adjustment strategy of the reference optoelectronic module and the state adjustment training strategy of the reference optoelectronic module.

5. The method according to claim 1, characterized in that The method further comprises: Acquire the target optoelectronic module; Based on the temporal commonality measurement value of the target optoelectronic module and the state adjustment decision knowledge of several attention dimensions, the current state adjustment strategy of the target optoelectronic module is determined.

6. A state adjustment optimization system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 5.

Images