Distributed photovoltaic user-oriented acquisition terminal adaptive data acquisition method

Abstract

The invention discloses a distributed photovoltaic user-oriented acquisition terminal adaptive data acquisition method, which belongs to the field of distributed photovoltaic technology, and comprises the following steps: an acquisition master station establishes a configuration template library containing photovoltaic equipment static characteristics, an environmental response model and an operation strategy based on historical photovoltaic data; receiving a user demand through the visual interaction interface and generating a configuration scheme; the acquisition master station converts the scheme into a structured configuration file, encrypts the structured configuration file and issues the encrypted structured configuration file to the terminal; the terminal dynamically executes the collection task and compares a difference value between the current power data and a set normal threshold value in real time, and when the difference value is detected to exceed a first preset threshold value, configuration updating is triggered; and the master station generates an incremental configuration file based on a machine learning algorithm to complete closed-loop optimization. Through template configuration, intelligent matching and dynamic adjustment mechanisms, the technical problems of rigid configuration, low communication efficiency, lack of safe closed loop and high operation and maintenance cost can be solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of distributed photovoltaic, and particularly relates to a self-adaptive data collection method for a collection terminal of a distributed photovoltaic user. BACKGROUND

[0002] In recent years, distributed photovoltaic has shown explosive growth, and therefore a multi-layer power consumption information collection system of a master station-collection terminal-inverter / assembly has been built. Early systems rely on manual parameter issuance or factory solidification configuration, and can only complete fixed-period and fixed-precision power and electricity reading, and cannot adapt to differences in inverter brands, weather conditions and user individual needs. Subsequently, a configuration file remote upgrade based on JSON / XML was introduced, but it is still not perfect in nature.

[0003] There are four core defects in the existing distributed photovoltaic data collection technology: first, configuration rigidity, terminal functions and collection strategies are hard-coded in firmware, and adjustment of sampling intervals, data precision and other parameters requires recompilation and whole package upgrade, which cannot adapt to inverter differences, weather changes and user individual needs, and is long in cycle and high in risk; second, inefficient communication, configuration files are issued in plaintext, there is no breakpoint continuation and differential update mechanism, and in a narrowband channel such as 2G / 4G / NB, the transmission success rate is low, the retransmission rate is high, and the traffic consumption is large; third, lack of security and version management, configuration files have no digital signature, version check and local rollback mechanism, are easy to tamper with or misupgrade, and cannot meet the power grid secondary security requirements; and fourth, high operation and maintenance cost, terminal debugging and abnormal recovery rely on manual on-site operation, and it is difficult to adapt to the operation and maintenance needs of the rapid growth of distributed photovoltaic. SUMMARY

[0004] The purpose of the present application is to overcome the deficiencies of the prior art and provide a self-adaptive data collection method for a collection terminal of a distributed photovoltaic user, which solves the technical problems in the background art.

[0005] The purpose of the present application is achieved as follows: a self-adaptive data collection method for a collection terminal of a distributed photovoltaic user, comprising the following steps: Step S101, establishing a configuration template library based on historical photovoltaic power generation data; the configuration template library includes a photovoltaic device static characteristic parameter set, an environment response model library and an operation strategy knowledge base; the photovoltaic device static characteristic parameter set includes inverter rated power, voltage range of maximum power point tracking, component type and its efficiency curve; the environment response model library includes power output characteristics under different light intensities, temperature attenuation compensation coefficients, and collection strategies corresponding to typical weather patterns; the operation strategy knowledge base stores basic collection strategies, emergency strategies and optimization strategies, and the strategies include sampling intervals, data precision and abnormal handling rules.

[0006] Further, the step S101 comprises: extracting original power generation data, meteorological data and equipment account information from a historical photovoltaic power generation database; performing missing value filling, outlier elimination and time synchronization processing on the original power generation data, meteorological data and equipment account information to form a standard data set; performing feature extraction on the standard data set to obtain photovoltaic equipment static features, environmental dynamic features and operation features; clustering the original power generation data and environmental dynamic features by using an unsupervised clustering algorithm to establish a mapping relationship between weather patterns and power output and construct an environmental response model library; training an environmental response model based on the clustering results and corresponding environmental dynamic features by using a regression or neural network algorithm to output power output features, temperature attenuation compensation coefficients and prediction error distributions under different weather patterns; generating basic collection strategies, emergency strategies and optimization strategies according to equipment static features, the environmental response model library and business requirements to constitute an operation strategy knowledge base; and storing the photovoltaic equipment feature static parameter set, the environmental response model library and the operation strategy knowledge base in a configuration template library in a structured format, assigning a version number and a digital signature to each template.

[0007] Step S102, receiving collection requirements input by a user through a visual interactive interface and automatically generating a configuration scheme containing collection frequency, data accuracy and abnormal threshold based on the configuration template library.

[0008] Further, the step S102 comprises: visualizing and displaying selectable collection parameter items including collection frequency, data accuracy, abnormal threshold, alarm mode and reporting channel to a user in a man-machine interactive interface of a collection master station; receiving and verifying individualized collection requirements input by the user in real time through an interface control; performing similarity matching on the individualized collection requirements and the configuration template library to filter out matched candidate templates; calling a machine learning model based on the matched candidate templates to automatically generate a configuration scheme containing collection frequency, data accuracy and abnormal threshold and displaying a preview scheme to the user; receiving confirmation or fine-tuning instructions of the user on the preview scheme and generating a configuration file after confirming that the preview scheme is correct.

[0009] Step S103, converting the configuration scheme into a machine-readable structured configuration file and attaching a digital signature to the configuration file.

[0010] Further, the step S103 comprises: converting the configuration file into a standardized XML configuration file, a structured configuration file containing collection tasks, parameter thresholds and strategy rules; performing hash operation and private key signature on the structured configuration file by using an SM2 or RSA algorithm to generate a digital signature; packaging the digital signature and the structured configuration file into a signature file and delivering the signature file to a terminal through an encrypted communication channel Step S104, the structured configuration file is downloaded to the terminal through the encrypted communication channel, and the acquisition data returned by the terminal after verifying the digital signature is received.

[0011] Further, the step S104 comprises: the master station establishes an end-to-end encrypted channel based on TLS1.3, and downloads the signed structured configuration file to the terminal in a resume transmission manner; after the terminal receives the file, the digital signature of the structured configuration file is verified using the pre-set SM2 public key, if the verification is passed, the next step is continued, otherwise, the file is discarded and a signature error alarm is reported; the terminal performs format checking and integrity checking on the structured configuration file, and after the checking is passed, the configuration parameters are parsed according to the pre-set XML Schema, and the parsing result is written into the local configuration database.

[0012] The terminal writes the parsed configuration parameters into the corresponding function registers according to the register address mapping table, and the function registers at least include a sampling period register, a data precision register, a power threshold register and an alarm enable register; the terminal generates a data acquisition task according to the register values written, and the data acquisition task at least includes a sampling channel, a sampling frequency, a data format and a reporting period.

[0013] The terminal periodically acquires current power data according to the data acquisition task, and compares the current power data with the normal threshold in real time; if the absolute value of the difference continuously exceeds the first pre-set threshold for three times, the terminal triggers an optimization configuration scheme request, and encrypts and reports the real-time acquisition data to the acquisition master station.

[0014] Step S105, according to the returned acquisition data, an optimal acquisition strategy is predicted by using a machine learning algorithm, an incremental configuration file containing only changed parameters is generated, the incremental configuration file is downloaded to the terminal through the encrypted communication channel, and a closed-loop update is completed.

[0015] Further, the step S105 comprises: the acquisition master station receives and decrypts the real-time acquisition data, and uses a pre-trained machine learning model to perform prediction analysis on the real-time acquisition data to obtain an optimal acquisition strategy; the acquisition master station compares the optimal acquisition strategy with the current configuration to obtain an incremental configuration file composed of changed parameters; the acquisition master station downloads the incremental configuration file to the terminal through the encrypted channel, and the terminal receives and applies the incremental configuration file to complete the closed-loop update.

[0016] The collection master station appends a new version number and a secondary digital signature based on an SM2 algorithm to the incremental configuration file after generating the incremental configuration file, and writes the version number and the secondary digital signature to a file header of the incremental configuration file. After receiving the incremental configuration file, the terminal first verifies the secondary digital signature and the version number, and only when the verification is passed and the version number is higher than the local current configuration version number, the incremental update is performed, and the old version configuration is backed up locally. Once it is found that the new configuration has an exception during running, the terminal restores to the old version configuration.

[0017] The present application has the following advantages: based on historical power generation data and real-time operation data, a differentiated configuration file is automatically generated, incremental update, version rollback and security closed loop optimization are realized through an encrypted channel, and the problems of rigid configuration, low communication efficiency, lack of security and version management and high operation and maintenance cost in the process of collecting photovoltaic user information are solved.

[0018] First, the template dynamic configuration replaces the firmware hard coding, realizes the minute-level self-adaptive adjustment of the inverter difference, weather change and user individual demand without the need of whole package upgrade. Second, the combination of TLS encryption breakpoint continuation and incremental delivery significantly improves the transmission reliability in narrowband environment and reduces the traffic consumption. Third, the SM2 double signature plus version check and local rollback mechanism form a security closed loop, effectively resisting tampering and misupgrade risk. Fourth, relying on machine learning algorithm closed loop optimization and remote parameter self-adaptive adjustment, the artificial on-site debugging is converted into online adaptive operation and maintenance, which significantly reduces the long-term operation and maintenance cost. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a method step flow chart of the present application; Figure 2 is a schematic diagram of the implementation process equipment of the present application. DETAILED DESCRIPTION

[0020] The present application will be further described in detail below with reference to the accompanying drawings, and it should be pointed out that the following is only for more clearly illustrating and explaining the present application.

[0021] As shown in Figure 1 and 2 The present embodiment discloses a self-adaptive data collection method of a collection terminal for distributed photovoltaic users, comprising the following steps: Step S101, The configuration template library is established based on historical photovoltaic power generation data. The process of step S101 can specifically include the following steps: extracting original power generation data, meteorological data, and equipment account information from the historical photovoltaic power generation database, performing missing value filling, abnormal value elimination, and time synchronization processing on the original power generation data, meteorological data, and equipment account information to form a standard data set; performing feature extraction on the standard data set to obtain photovoltaic equipment static features, environmental dynamic features, and operation features; clustering the original power generation data and the environmental dynamic features by using an unsupervised clustering algorithm, establishing a mapping relationship between weather patterns and power output, and constructing an environmental response model library; based on the clustering results and the corresponding environmental dynamic features, training the environmental response model by using a regression or neural network algorithm, and outputting power output features, temperature attenuation compensation coefficients, and prediction error distributions under different weather patterns; generating basic collection strategies, emergency strategies, and optimization strategies according to the equipment static features, the environmental response model library, and business requirements, and constructing an operation strategy knowledge base; and storing the photovoltaic equipment feature static parameter set, the environmental response model library, and the operation strategy knowledge base in the configuration template library in a structured format, and assigning a version number and a digital signature to each template.

[0022] Specifically, in the field of photovoltaic power generation data collection, the traditional configuration method has problems such as rigid parameter setting and poor environmental adaptability. The present application realizes a complete conversion process from raw data to application strategy by establishing an intelligent configuration template library. In the data collection link, the computer remote supervision control and data acquisition system obtains minute-level data such as inverter direct current side voltage, current, and power, and synchronously collects parameters such as irradiance and environmental temperature recorded by the weather station. The equipment account provides static information such as component model and rated parameters. These raw data are preprocessed to form a standard data set. The processing process includes using linear interpolation to fill in the missing irradiance values, eliminating voltage abnormal data based on the 3σ principle, and ensuring the time consistency of each data source by time stamp alignment.

[0023] The feature extraction stage converts the standard data set into three types of feature vectors. The photovoltaic equipment static features are derived from the equipment account and include inherent parameters such as component open-circuit voltage and short-circuit current. The environmental dynamic features are obtained by sliding window calculation and include 5-minute power change rate and frequency domain energy distribution obtained by FFT transformation. The operation features reflect the actual working state of the equipment, such as the MPPT voltage tracking range. These features provide structured input for subsequent modeling, in which the static features determine the basic performance boundary of the equipment, the dynamic features reflect the environmental change law, and the operation features represent the real-time state of the system.

[0024] The clustering algorithm processes the correlation between environmental dynamic characteristics and power generation data, uses an improved DBSCAN algorithm for pattern recognition of the irradiance-power curve, sets the parameter combination of neighborhood radius 0.15 and minimum sample number 20, and divides the weather conditions into six typical modes such as sunny and cloudy. Each weather mode corresponds to a specific power output characteristic, and a mapping relationship is established through an LSTM neural network. The network input layer receives a 24-hour historical data window containing 5-dimensional time series data such as irradiance and temperature, and the output layer predicts the power trend, temperature decay coefficient, and confidence interval for the future period. The model training uses the mean square error loss function and the Adam optimizer, and stabilizes after 100 iterations on the historical data set.

[0025] The environmental response model and the device static characteristics jointly constitute the basis for strategy generation. In sunny mode, a 5-minute sampling interval is used for monocrystalline silicon components, and in cloudy conditions, it is adjusted to 2 minutes. This differentiated configuration is derived from statistical analysis of historical data. The emergency strategy design considers abnormal conditions such as voltage sag, and automatically switches to second-level sampling when more than 10% fluctuation is detected. In the case of communication interruption, a local cache mechanism is enabled. All strategies are stored in MongoDB sharded clusters in JSON format, each template contains a version number and an SM2 digital signature, and the version number uses semantic numbering rules to reflect the iterative update process.

[0026] Step S102, The acquisition demand input by the user is received through the visual interactive interface, and a configuration scheme including acquisition frequency, data accuracy, and abnormal threshold is automatically generated based on the configuration template library. The process of executing step S102 can specifically include the following steps: in the human-computer interaction interface of the acquisition master station, the user is visually displayed with selectable acquisition parameter items, including acquisition frequency, data accuracy, abnormal threshold, alarm mode, and reporting channel; real-time reception and verification of user input personalized acquisition requirements are performed through interface controls; similarity matching of the personalized acquisition requirements with the configuration template library is performed to screen out matched candidate templates; based on the matched candidate templates, a machine learning model is called to automatically generate a configuration scheme including acquisition frequency, data accuracy, and abnormal threshold, and a preview scheme is displayed to the user; the user's confirmation or fine-tuning instruction for the preview scheme is received, and a configuration file is generated after confirmation.

[0027] Specifically, in the photovoltaic power generation data acquisition system, the traditional configuration method has problems such as complex operation and inflexible parameter setting. The application realizes intuitive configuration of collection requirements and automatic generation of schemes by combining a visual interactive interface with intelligent algorithms. The visual interactive interface is designed using the Vue.js framework and includes three core areas: a device type selection area, a parameter configuration area, and a scheme preview area. The device type selection area displays six component type options, including monocrystalline silicon and thin film, through a drop-down menu. The corresponding electrical parameter ranges in the template library are automatically associated in the background. When the user selects the "monocrystalline silicon" type, the system automatically loads constraints such as a voltage range of 200-800V and a current range of 0-10A. The parameter configuration area uses dynamic form technology to realize interactive logic. The collection frequency setting slider control range is 5-300 seconds, and the data accuracy options provide three levels of selection: 0.2, 0.5, and 1.0. The built-in verification rules in the abnormal threshold input box ensure that the input value is within the range of ±10%-±25% of the rated value. The user's operation triggers a real-time verification mechanism, and when the input value exceeds the rated parameters of the component, the interface displays a red warning border and prompts the valid range. The alarm mode options include SMS, email, and platform pop-up window notification methods, and the reporting channel settings support 4G private network and optical fiber transmission media selection.

[0028] The configuration requirements submitted by the user are converted into a structured query vector, including device type encoding and parameter weight coefficient, etc. The similarity matching engine uses an improved cosine similarity algorithm to retrieve candidate templates in the configuration template library. The algorithm adds a power curve shape weight factor, and the matching threshold is set to 0.85. For a 50MW photovoltaic power plant, the average retrieval time is controlled within 200ms, and 3-5 optimal candidate templates are returned.

[0029] The machine learning model receives the candidate template features and user demand vector for fusion processing. The LSTM neural network analyzes the pattern rules in the historical operation data and outputs optimized values for parameters such as collection frequency and data accuracy. The multi-objective optimization algorithm balances communication overhead and data value. The objective function considers transmission delay, energy cost, and information entropy weight. The NSGA-II genetic algorithm is used to solve the Pareto optimal solution set. The scheme generation stage automatically avoids resource conflicts. When high-frequency sampling and high-precision collection are detected simultaneously, the resource competition warning mechanism is automatically triggered.

[0030] The preview panel displays the simulation running effect of the configuration scheme, including the predicted daily data transmission volume, key parameter curve comparison chart and other visual elements. Before the user confirms, the parameters can be fine-tuned, and the adjusted values are updated in real time through the constraint propagation algorithm to ensure the internal consistency of the configuration scheme. The confirmed configuration scheme is serialized into a JSON format, including a header information, a parameter body and a security check code. The header records the scheme version number and the generation timestamp, and the parameter body is stored according to the collection strategy, communication strategy and security strategy.

[0031] Step S103, The configuration scheme is converted into a machine-readable structured configuration file, and a digital signature is attached to the configuration file. The process of performing step S103 can specifically include the following steps: converting the configuration file into a standardized XML configuration file, a structured configuration file containing collection tasks, parameter thresholds and policy rules; performing hash operation and private key signature on the structured configuration file using SM2 or RSA algorithm to generate a digital signature; the digital signature and the structured configuration file are packaged into a signature file and are issued to the terminal through an encrypted communication channel.

[0032] Specifically, in the photovoltaic power generation data collection system, the secure transmission and tamper-proofing of the configuration scheme is a key link to ensure the reliable operation of the system. After the configuration scheme is generated, it is converted into an XML format structured document through standardization processing. The document header contains scheme version number, generation timestamp and device type metadata, and the main body is filled with content according to three dimensions of collection parameters, control strategies and security rules. The collection parameter section records specific indicators such as direct current voltage sampling range 200-850V and current sampling accuracy 0.5 level; the control strategy section defines the sampling interval under different weather conditions, such as 300 second basic interval in sunny mode, and automatically switches to 60 second enhanced mode when power fluctuation exceeds 5%; the security rule section specifies that the data encryption method is AES-256, and specifies the check code generation algorithm as CRC-32.

[0033] The conversion process of the XML document adopts the DOM tree construction method, and each parameter item is mapped to a specific node. Numerical parameters such as voltage threshold are stored in the form of attributes, and composite policy rules are converted into nested node structures. The document structure follows the format specification defined in the power industry standard DL / T 698.45, and is compatible with the parser of the terminal device. The converted document is standardized, including character encoding conversion to UTF-8, white space compression and node sorting standardization, to ensure that documents with the same content have exactly the same binary representation, laying a foundation for subsequent signature verification.

[0034] The digital signature generation link adopts the national SM2 algorithm, and the signature process first performs SM3 hash operation on the XML document to generate 256-bit summary information. The signature key management system allocates a unique key pair to each photovoltaic power station, the private key is stored in the HSM hardware security area, and the signature operation is performed in a secure environment. The signature data includes three parts: document hash value, signature timestamp and issuer digital certificate serial number. These data are encoded through ASN.1 format, and then encrypted by using the private key. The generated signature block is packaged with the original XML document through the PKCS#7 standard to form a complete signature file, and the digital certificate chain issued by the CA is embedded in the file to verify the legality of the signature.

[0035] The file transmission stage adopts a double encryption mechanism, the application layer uses the SM4 algorithm to encrypt the signature file, and the key is dynamically generated through the SM2 key exchange protocol; the transmission layer establishes a TLS 1.3 secure channel, and uses the ECDHE_SM4_SM3 cipher suite. After receiving the file, the terminal device verifies the validity of the TLS certificate chain, decrypts the signature file to obtain the XML document and the signature data. The verification process uses the built-in CA root certificate to verify the identity of the signer, and uses the public key to decrypt the signature data to obtain the document hash value, and simultaneously recalculates the SM3 hash of the received document for comparison. Only the configuration file that passes the integrity check will be loaded and executed, and the document that fails the check will trigger a security alarm and return to the master station.

[0036] Step S104, The structured configuration file is downloaded to the terminal through the encrypted communication channel, and the acquisition data returned by the terminal after verifying the digital signature is received. The process of performing step S104 can specifically include the following steps: the master station establishes an end-to-end encrypted channel based on TLS1.3, and downloads the signed structured configuration file to the terminal in a breakpoint resume manner; after the terminal receives it, the pre-set SM2 public key is used to verify the digital signature of the structured configuration file, and if the verification is passed, the next step is continued, otherwise the file is discarded and a signature error alarm is reported; the terminal performs format checking and integrity checking on the structured configuration file, and after the checking is passed, the configuration parameters are parsed according to the pre-set XML Schema, and the parsing result is written into the local configuration database.

[0037] Specifically, in the photovoltaic power generation data collection system, the reliable transmission and safe execution of the configuration file directly affect the stability of the power station operation. When the collection master station establishes a communication connection, it initiates a TLS 1.3 handshake protocol based on the photovoltaic terminal device identification code, negotiates a session key using the ECDHE_SM4_SM3 cipher suite, and embeds a power-specific CA root certificate in the certificate verification link to ensure that only authenticated terminals can establish a connection. During the channel establishment process, the master station checks whether the OU field of the terminal device certificate contains the "PV-Collector" identifier, and simultaneously verifies the certificate validity period and CRL revocation status. If the handshake fails, the connection is immediately terminated and a security event is recorded.

[0038] The configuration file transmission adopts an improved breakpoint resume mechanism, packaging XML documents and signature data into 512 KB data blocks, each block with a CRC-32 check code. The transmission protocol design uses a sliding window flow control, with the window size dynamically adjusted according to the terminal signal strength. The default window size is set to 8 data blocks in a 4G network environment, and the fiber channel is expanded to 16 blocks. When the transmission is interrupted, the terminal records the hash sequence of the received blocks, sends a BITMAP retransmission request after the connection is restored, and the master station accurately supplements the missing blocks according to the request, avoiding repeated transmission. After receiving the data blocks, the terminal reassembles the original file in PKCS#7 format and verifies the CRC values of each block to ensure the correctness of the reassembly.

[0039] In the signature verification link, the terminal reads the pre-installed SM2 public key from the security chip and parses the signature file to obtain the ASN.1 encoded signature data. The verification process first extracts the XML document to calculate the SM3 hash value, uses the public key to decrypt the signature to obtain the original hash, and compares the consistency of the two. The judgment criteria for the validity of the signature include: hash matching, timestamp within the valid period, and certificate chain verification of the issuer passed. The three-level response mechanism is triggered upon verification failure: the file is retransmitted upon the first failure, the certificate is disabled upon the second failure, and the terminal configuration interface is locked and a security alarm is uploaded upon the third failure. The alarm information includes terminal positioning data and failure reason code.

[0040] Before configuration file parsing, format verification is performed. The verifier loads the XML Schema defined in DL / T 698.45 standard, checks whether the document structure conforms to the nesting rules of collection parameters, control strategies, and safety rules. The integrity check focuses on detecting the range of key parameters, with the DC voltage threshold must be between 200-850V, the sampling interval must not be less than 5 seconds, and the abnormal threshold floating range is limited to ±10%-±25%. The verification process uses stream parsing technology, which verifies while parsing, and immediately aborts the process upon discovering a violation.

[0041] The version management mechanism is established when the parameter is written into the configuration database, the new configuration obtains an incremental version number, and the old configuration retains a historical version for emergency rollback. The database operation enables transaction processing to ensure atomic update of the parameter, and the writing process records the operation log, including the modification time, operator identification, and parameter change summary. Before the configuration takes effect, a confirmation request is sent to the master station, the master station compares the consistency of the current configuration version of the terminal and the version of the issued configuration, and authorizes the terminal to activate the new configuration after confirming that there is no error.

[0042] For better effect, the process of performing step S104 can further include the following steps: The terminal writes the parsed configuration parameters into the corresponding function registers according to the register address mapping table, and the function registers at least include a sampling period register, a data precision register, a power threshold register, and an alarm enable register; the terminal generates a data collection task in real time according to the written register values, and the data collection task at least contains a sampling channel, a sampling frequency, a data format, and a reporting period.

[0043] Specifically, in the photovoltaic power generation data collection terminal, dynamic mapping of configuration parameters and task generation are the core links to realize adaptive collection. After receiving the verified configuration file, the terminal parses and extracts the parameter nodes in the XML document, and performs parameter allocation according to the predefined register address mapping table. The sampling period parameter is written into the 32-bit register group starting from 0x2000, the data precision parameter is configured into the 8-bit register starting from 0x2100, and the power threshold parameter is filled into the 16-bit floating-point register space at the 0x2200 address segment. The register writing process adopts a double buffering mechanism, the parameters are first temporarily stored in the shadow register area, and after verification, they are synchronized to the working register through the memory barrier instruction to avoid conflicts in the parameter update process.

[0044] After the register configuration is completed, the task generation engine scans the activated register area to construct a collection task descriptor. The sampling channel parameter is converted into a physical ADC channel selection code, such as DC voltage corresponding to ADC1 channel and environmental temperature corresponding to ADC4 channel. The sampling frequency parameter is calculated by PLL clock division, and is converted into a specific timer trigger interval. When the configuration requires a 300-second interval, the terminal divides the 32.768kHz reference clock into a 109Hz trigger signal. The data format parameter controls the preprocessing mode of the ADC sampling result. 12-bit raw data is subjected to sliding average filtering according to the precision register value, 8 times of superposition average is performed for 0.5-level precision, and 4 times of average is performed for 1.0-level precision.

[0045] The communication resource is allocated according to the reporting period parameter, 5-minute reporting period corresponds to the low-power mode of the GPRS channel, and 1-minute period enables the high-speed transmission of the 4G channel. The alarm enable register controls the abnormal detection process, when the difference value of the power threshold register meets the trigger condition, the main processor is directly interrupted, and the preset emergency sampling sequence is started. The task descriptor is finally packaged as a binary instruction stream, transmitted to the acquisition coprocessor through the DMA controller, and the coprocessor performs sampling, preprocessing and caching operations according to the instruction sequence, without occupying the main processor resources.

[0046] The real-time performance of parameter mapping and task generation directly affects the data quality, the register write delay control is within 50 microseconds, and the task descriptor generation time does not exceed 200 microseconds. The terminal maintains a register version mark, and the version number is incremented each time it is updated, so that the main station can confirm the configuration effective state by querying the version number. The abnormal processing mechanism monitors the register write result, automatically rolls back to the last valid configuration in case of verification failure, and reports error logs through a secure channel, which includes the difference analysis of register address, expected value and actual value For better effect, the process of executing step S104 can further include the following steps: The terminal periodically collects current power data according to the data collection task, and performs real-time difference comparison between the current power data and the normal threshold value; if the absolute value of the difference value exceeds the first preset threshold value for three consecutive times, the terminal triggers an optimization configuration scheme request and reports the real-time collected data to the collection master station after encryption.

[0047] Specifically, in the photovoltaic power generation data collection system, a real-time monitoring and dynamic response mechanism is established when the terminal device executes the data collection task. The terminal initializes the sampling task according to the configuration parameters, the ADC converter starts voltage and current measurement according to the set sampling period, and the measurement results are filtered by an IIR digital filter to eliminate high-frequency noise, and then the current power value is calculated by a DSP processor. The power data is stored in a circular buffer, and the buffer is designed in a three-level structure: the raw data layer stores ADC sampling values, the processed data layer saves filtered effective data, and the analysis data layer records power calculation results.

[0048] The normal threshold value range and the first preset threshold value are read from the configuration register, and the normal threshold value is dynamically generated according to the rated parameters of the photovoltaic module, such as setting the normal operation interval of a 300kW string to 270-330kW. A sliding window algorithm is used, and the window size is fixed at 3 consecutive sampling points. Each time new data arrives, the absolute difference between the current power and the normal threshold value is calculated, and when the difference exceeds the first preset threshold value for three consecutive times, an optimization configuration scheme request is triggered. In order to prevent false triggering caused by instantaneous interference, a hysteresis comparison mechanism is introduced, which requires that the fluctuation amplitude of the three over-limit difference values does not exceed 2%.

[0049] After triggering the optimization configuration scheme request, the terminal starts the data packaging process, collects power data, environmental temperature and device state information in the last 30 minutes, and packages them into an optimization request message. The message organizes data in TLV format, including metadata fields such as timestamp, device identification, and exception type. Encryption uses the SM4 algorithm to encrypt the message, and the key is dynamically generated by the security chip. The encrypted data is attached with an SM3 hash value as an integrity check code. The transmission layer establishes a dedicated message channel and reports to the master station through the MQTT protocol with QoS level 1, ensuring that the message is delivered at least once.

[0050] Step S105, According to the returned collection data, an optimal collection strategy is predicted using a machine learning algorithm, an incremental configuration file containing only changed parameters is generated, and the incremental configuration file is downloaded to the terminal through the encrypted communication channel to complete the closed-loop update. The process of executing step S105 can specifically include the following steps: The collection master station receives and decrypts the real-time collection data, uses a pre-trained machine learning model to perform predictive analysis on the real-time collection data, and obtains an optimal collection strategy. The collection master station compares the optimal collection strategy with the current configuration, extracts an incremental configuration file composed of changed parameters, and downloads the incremental configuration file to the terminal through the encrypted channel. The terminal receives and applies the incremental configuration file to complete the closed-loop update.

[0051] Specifically, in the photovoltaic power generation data collection system, the intelligent processing and closed-loop update of the terminal reported data by the collection master station constitute the core link of system self-optimization. After receiving the encrypted message from the terminal, the collection master station performs SM4 decryption and checks the SM3 hash value to ensure data integrity. The decrypted data is stored in a time series database, including power value, environmental temperature, device state, etc. The data is stored at 10-second intervals and time index is established. The data preprocessing link performs standardization processing on the original data, fills in the missing values using linear interpolation of adjacent time points, and removes the outliers by isolated forest algorithm, forming a regular data set meeting the model input requirements.

[0052] The pre-trained LSTM network architecture is loaded using a machine learning model. The network input layer receives a 120-time step historical data window containing 6-dimensional features such as power fluctuation sequence, temperature trend, etc. The model inference process is executed in a GPU-accelerated environment. The inference engine inputs standardized data into the network, which is processed through 3 layers of LSTM units and 2 layers of fully connected layers. The output is a 15-minute power prediction curve and confidence interval. Based on the prediction results, a collection strategy is generated. When the power fluctuation is predicted to intensify, the sampling interval is recommended to be shortened to 1 / 3 of the original value, and the abnormal threshold is tightened by 20%. During stable operation, the sampling interval is extended to reduce energy consumption. The strategy is evaluated by a multi-objective optimization algorithm, which considers data value, communication overhead, and device loss in three dimensions. The NSGA-II algorithm is used to solve the Pareto optimal solution set.

[0053] The system compares the optimized strategy with the current configuration, establishes a parameter change mapping table, and only marks configuration items with a value change of more than 5%. The generator reads the mapping table, constructs a simplified XML incremental configuration file according to the DL / T 698.45 standard, and retains the version number and change timestamp in the file header. The main part only contains the parameter nodes that need to be modified. The SM2 algorithm is used to set the digital signature, and the private key is stored in the encryption machine. The signature data includes the document hash value and the timestamp. The signature process introduces random numbers to enhance security.

[0054] The incremental configuration file is delivered through a channel that multiplexes a TLS 1.3 encrypted link. The transmission layer implements breakpoint continuation and priority scheduling, and the key configuration update packet is marked as the highest priority. After receiving the incremental file, the terminal verifies the validity of the signature, and the parser applies the changes in order of version number to modify the corresponding parameters in the local configuration database. Before the configuration takes effect, a pre-validation is performed to check whether the new parameters are within the allowed range of the device, such as the ADC sampling interval cannot be lower than 5 seconds, and the voltage threshold must be between 200-850V. The validated configuration takes effect immediately, and the terminal sends a confirmation message to the master station, forming a complete control loop.

[0055] To achieve better results, the process of step S105 can further include the following steps: After generating the incremental configuration file, the collection master station appends a new version number and a secondary digital signature based on the SM2 algorithm to it, and writes the version number and secondary digital signature into the file header of the incremental configuration file; After receiving the incremental configuration file, the terminal first verifies the secondary digital signature and version number. Only when the verification is passed and the version number is higher than the current local configuration version number, the incremental update is performed and the old version configuration is backed up locally. Once an abnormality is found in the new configuration during operation, the terminal can immediately restore to the old version configuration.

[0056] Specifically, in the configuration update process of the photovoltaic power generation data collection system, a multi-verification strategy is adopted in the security interaction mechanism between the master station and the terminal to ensure the reliability of the update. After receiving the operation data reported by the terminal, the collection master station performs SM4 decryption through the security chip, and the decrypted data is stored in the time series database after integrity verification. These data include key parameters such as power value and environmental temperature, and each data point is marked with a timestamp accurate to milliseconds and a unique terminal device identifier. The data analysis engine loads the pre-trained LSTM neural network model, the model input layer receives a 120-time step data window containing 6-dimensional features, and after processing through 3-layer LSTM units and 2-layer fully connected layers, the output is the power prediction curve and confidence interval range of the future period.

[0057] Based on the prediction results, a collection parameter adjustment scheme is generated, and when the power fluctuation is detected to be intensified, new sampling intervals and abnormal threshold values are automatically calculated. The optimization strategy is compared with the current configuration item by item, and only the parameter items with a change amplitude exceeding 5% are extracted to generate an incremental XML configuration file. During file generation, the version control service assigns a semantic version number to the file, uses the format of major version number.minor version number.revision number to identify the change level, and simultaneously calls the HSM hardware security module to perform SM2 secondary signature, jointly signs the file content and version number, and the signature data is written into the file header special field through ASN.1 encoding.

[0058] The terminal receives and verifies the validity of the secondary signature, and the signature verification process uses the preset security certificate, while strictly checking that the version number must comply with the version evolution rules and be greater than the local current version. The version management service automatically backs up the current configuration to the secure storage area before applying the update, and the backup file is attached with an operation timestamp and a terminal state snapshot.

[0059] The terminal real-time monitors the running state of the new configuration, and automatically triggers the rollback mechanism when detecting 3 consecutive collection failures or parameter out-of-range. The rollback process restores the configuration of the last stable version from the backup area, and generates a security event log to report to the master station. The entire update process establishes a complete audit trail, records the timestamp, execution result and related parameters of each operation step, and these log data are synchronized to the master station analysis platform through the secure channel regularly.

[0060] As Figure 2As shown, the relationship between the parts in the implementation process is shown. According to the illustrated architecture, the system realizes the adaptive collection and optimization of photovoltaic data through a closed-loop control mechanism. The terminal device continuously obtains real-time data such as original power generation data, meteorological data, and equipment account information from the photovoltaic panel. According to the sampling frequency and accuracy set in the structure configuration file issued by the collection master station, the collection task is executed. During the collection process, the built-in monitoring algorithm in the terminal compares the current power data with the preset normal threshold value in real time. When the absolute value of the power difference is detected to exceed the first threshold value for 3 times in a row, an optimization configuration request is immediately sent to the collection master station through an encrypted channel. The request message contains complete collection data, environmental parameters, and equipment state information during the abnormal period.

[0061] After the collection master station receives the optimization request, the optimal collection strategy is recalculated through a machine learning model combined with historical operation data and real-time reporting information. The strategy optimization engine generates an incremental configuration file containing only the parameters that need to be adjusted. The new configuration file is attached with a version number and a digital signature and then transmitted to the terminal through the structure configuration file distribution channel. After the terminal verifies the validity of the signature, the new configuration is applied, and the old version is backed up. If an abnormality is detected during operation, it can automatically roll back.

[0062] The visualization interface receives the collection requirements input by the user in this process and also displays the power data reported by the terminal and the configuration parameters issued by the master station in real time. The interface displays the relationship between power fluctuations and threshold values in the form of a curve graph and highlights the alarm information when the configuration optimization event is triggered. Through this closed-loop mechanism, the system enables the collection strategy to dynamically adapt to changes in the output characteristics of the photovoltaic panel, optimizing the use of communication resources while ensuring data collection quality.

[0063] This architecture realizes end-to-end automated management from user requirements to data collection. The components are interconnected through standardized interfaces, and the separation of data flow and control flow design ensures system scalability. In actual operation, the master station can handle hundreds of concurrent requests from terminals per second, with configuration update delay controlled at the second level, meeting the real-time monitoring requirements in a distributed photovoltaic scenario. The system continuously optimizes the collection strategy through machine learning algorithms, ensuring data quality while reducing communication overhead. Typical applications can significantly reduce the amount of invalid data transmission.

[0064] The application collects a configuration template library containing static characteristics of photovoltaic equipment, environmental response model and operation strategy of the photovoltaic equipment based on historical photovoltaic data of a master station; receives user demand through a visual interactive interface and generates a configuration scheme; the master station converts the scheme into a structured configuration file and encrypts and issues the file to a terminal; the terminal dynamically executes a collection task and compares a difference between current power data and a set normal threshold value in real time, and triggers configuration updating when detecting that the difference exceeds a first preset threshold value; the master station generates an incremental configuration file based on a machine learning algorithm to complete closed-loop optimization. Through the template configuration, intelligent matching and dynamic adjustment mechanism, the technical problems of rigid configuration, low communication efficiency, lack of safety closed loop and high operation and maintenance cost are completely solved.

[0065] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can make equivalent replacement or change according to the technical scheme and concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for adaptive data acquisition of a collection terminal for distributed photovoltaic users, characterized in that, The method comprises the following steps: Step S101, establishing a configuration template library based on historical photovoltaic power generation data; Step S102, receiving user input collection requirements through a visual interactive interface, and automatically generating a configuration scheme including collection frequency, data accuracy, and abnormal threshold in combination with the configuration template library; Step S103, converting the configuration scheme into a machine-readable structured configuration file, and attaching a digital signature to the configuration file; Step S104, issuing the structured configuration file to a terminal through an encrypted communication channel, and receiving collection data returned by the terminal after verifying the digital signature; Step S105, predicting an optimal collection strategy according to the returned collection data, generating an incremental configuration file containing only changed parameters, and issuing the incremental configuration file to the terminal through the encrypted communication channel to complete closed-loop updating.

2. The adaptive data acquisition method of the distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The configuration template library comprises a photovoltaic device static characteristic parameter set, an environment response model library, and a running strategy knowledge base; The photovoltaic device static characteristic parameter set comprises inverter rated power, voltage range of maximum power point tracking, component type, and efficiency curve thereof; The environment response model library comprises power output characteristics under different light intensities, temperature attenuation compensation coefficients, and collection strategies corresponding to typical weather patterns; The running strategy knowledge base stores basic collection strategies, emergency strategies, and optimization strategies, and the strategies comprise sampling interval, data accuracy, and abnormal handling rules.

3. The adaptive data acquisition method of distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The step S101 comprises: Extracting original power generation data, meteorological data, and device account information from a historical photovoltaic power generation database, performing missing value filling, abnormal value elimination, and time synchronization processing on the original power generation data, meteorological data, and device account information to form a standard data set; Extracting features from the standard data set to obtain photovoltaic device static characteristics, environment dynamic characteristics, and running characteristics; Using an unsupervised clustering algorithm to cluster the original power generation data and environment dynamic characteristics, establishing a mapping relationship between weather patterns and power output, and constructing an environment response model library; Based on the clustering results and corresponding environment dynamic characteristics, training an environment response model using a regression or neural network algorithm, and outputting power output characteristics, temperature attenuation compensation coefficients, and prediction error distribution under different weather patterns; Generating basic collection strategies, emergency strategies, and optimization strategies according to device static characteristics, the environment response model library, and business requirements, and constructing a running strategy knowledge base; Storing the above photovoltaic device static characteristic parameter set, environment response model library, and running strategy knowledge base in the configuration template library in a structured format, and assigning a version number and a digital signature to each template.

4. The adaptive data acquisition method of distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The step S102 comprises: In the human-computer interaction interface of a collection master station, visualizing and displaying selectable collection parameter items to a user, the collection parameter items comprising collection frequency, data accuracy, abnormal threshold, alarm mode, and reporting channel; Real-time receiving and verifying user input personalized collection requirements through an interface control; Matching the personalized collection requirements with the configuration template library to screen out matched candidate templates; Based on the matched candidate template, a machine learning model is called to automatically generate a configuration scheme containing collection frequency, data accuracy and abnormal threshold, and a preview scheme is shown to the user; The user's confirmation or fine-tuning instruction for the preview scheme is received, and the configuration file is generated after the confirmation is correct.

5. The adaptive data acquisition method of distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The step S103 comprises: The configuration file is converted into a standardized XML configuration file, including a structured configuration file of collection task, parameter threshold and strategy rule; The structured configuration file is subjected to hash operation and private key signature by using SM2 or RSA algorithm to generate a digital signature; The digital signature and the structured configuration file are packaged into a signature file and are issued to the terminal through an encrypted communication channel.

6. The adaptive data acquisition method of distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The step S104 comprises: The collection master station establishes an end-to-end encrypted channel based on TLS1.3 to issue the signed structured configuration file to the terminal in a breakpoint resume manner; After the terminal receives it, the digital signature of the structured configuration file is verified using the preset SM2 public key, and if the verification is passed, the next step is continued, otherwise the file is discarded and a signature error alarm is reported; The terminal performs format verification and integrity verification on the structured configuration file, and after the verification is passed, the configuration parameters are parsed according to the preset XMLSchema, and the parsing result is written into the local configuration database.

7. The adaptive data acquisition method of the distributed photovoltaic user-oriented acquisition terminal according to claim 6, characterized in that, The step S104 further comprises: The terminal writes the parsed configuration parameters into the corresponding functional registers according to the register address mapping table, and the functional registers at least include a sampling period register, a data accuracy register, a power threshold register and an alarm enable register; The terminal generates a data collection task according to the written register values, and the data collection task at least contains a sampling channel, a sampling frequency, a data format and a reporting period.

8. The adaptive data acquisition method of the distributed photovoltaic user-oriented acquisition terminal according to claim 7, characterized in that, The step S104 further comprises: The terminal periodically collects the current power data according to the data collection task, and compares the current power data with the normal threshold in real time; If the absolute value of the difference continuously exceeds the first preset threshold for three times, the terminal triggers an optimization configuration scheme request and encrypts the real-time collection data to report to the collection master station. 9.The adaptive data acquisition method for distributed photovoltaic user-oriented acquisition terminal according to claim 1, characterized in that, The step S105 comprises: The collection master station receives and decrypts the real-time collection data, and uses a pre-trained machine learning model to perform prediction analysis on the real-time collection data to obtain an optimal collection strategy; The collection master station compares the optimal collection strategy with the current configuration to extract an incremental configuration file composed of changed parameters; The collection master station issues the incremental configuration file to the terminal through the encrypted channel, and the terminal receives and applies the incremental configuration file to complete the closed-loop update.

10. The adaptive data acquisition method of the distributed photovoltaic user-oriented acquisition terminal according to claim 9, characterized in that, The step S105 further comprises: After generating the incremental configuration file, the collection master station appends a new version number and a secondary digital signature based on the SM2 algorithm to the incremental configuration file, and writes the version number and the secondary digital signature into the file header of the incremental configuration file. After receiving the incremental configuration file, the terminal first verifies the secondary digital signature and the version number. Only when the verification is passed and the version number is higher than the local current configuration version number, the terminal performs the incremental update and keeps the old version configuration as a backup in the local. Once it is found that the new configuration has an exception during running, the terminal restores to the old version configuration.

Images