Cable cross interconnect adaptive processing system

Abstract

This specification provides an adaptive processing system for cable cross-connection. This system, by constructing a complete processing system integrating environmental perception, output prediction, strategy pre-planning, and real-time closed-loop correction, achieves a fundamental shift in cable cross-connection from "passive response" to "active prediction and adaptation." The system can proactively adjust operating parameters, effectively mitigating the impact of photovoltaic output fluctuations, significantly reducing circulating current losses and fault risks. Simultaneously, its adaptive sealing and connection protection mechanisms greatly improve long-term operational stability and reliability under extreme environments, thereby enhancing overall power transmission efficiency and reducing maintenance intervention needs and costs.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of power technology, and in particular to an adaptive processing system for cable cross-connection. Background Technology

[0002] In photovoltaic power plants deployed in harsh environments such as deserts and Gobi, the stable operation of cable interconnection systems faces severe challenges. Existing solutions mainly rely on passive responses and post-event adjustments to the real-time operating status of the system, which are insufficient to address issues such as cable interface performance degradation, enclosure sealing failure, and a surge in transmission circulating current losses caused by strong winds, sandstorms, drastic temperature differences, and sudden changes in sunlight. This lagging, passive control mode not only results in low energy transmission efficiency but also makes it difficult to guarantee the long-term reliability and economical operation and maintenance of the system, becoming a key bottleneck restricting the efficient power transmission from renewable energy bases.

[0003] Therefore, a better solution is urgently needed. Summary of the Invention

[0004] In view of this, embodiments of this specification provide an adaptive processing system for cable cross-connection to address the technical deficiencies existing in the prior art.

[0005] According to a first aspect of the embodiments of this specification, an adaptive processing system for cable cross-connection is provided, comprising:

[0006] The data processing module is used to acquire weather forecast data, photovoltaic module temperature data, historical output data of photovoltaic system and internal status monitoring data of cable cross-interconnection box, and to preprocess the acquired data; The prediction module is used to input the preprocessed data into a deep learning-based time series prediction model and output the prediction curve of photovoltaic power output. The planning module is used to generate a timing table of operating parameter adaptation strategies for cable cross-connection systems for future time periods, based on the prediction curves and pre-established correlation models. The correction module is used to compare the actual operating parameters of the photovoltaic system with the predicted values ​​corresponding to the prediction curve in real time, and dynamically adjust the operating parameter adaptation strategy based on the fuzzy control algorithm when the deviation exceeds the set threshold. The execution and feedback module is used to control the actions of the execution units of the cable interconnection interface and cross-connection box according to the operating parameter adaptation strategy or the dynamically adjusted operating parameter adaptation strategy, and to collect the actual operating data of the system and feed it back to the data processing module.

[0007] In one possible implementation, the deep learning-based time series prediction model consists of a long short-term memory network layer and a gated recurrent unit layer connected in series. The photovoltaic power output prediction curve output by the prediction module includes a medium- to long-term prediction curve for the next 24 hours and a short-term prediction curve for the next few hours at 15-minute intervals.

[0008] In one possible implementation, the cable interconnect interface is a modular quick-connect interface, including a male connector and a female connector. The male connector is crimped and fixed to the cable core, and the female connector has built-in conductive contacts and a locking mechanism. The execution and feedback module controls the locking mechanism to achieve connection and disconnection.

[0009] In one possible implementation, the cross-connected enclosure includes an enclosure structure, an environmental sensing layer, an active sealing layer, and a controller. The environmental sensing layer includes flexible capacitive sensors embedded in the sealed parts of the enclosure, used to monitor changes in sealing gaps and the depth of wind and sand erosion in real time. The active sealing layer comprises a shape memory polymer-based composite sealant and a micro electric heating unit integrated therein; The execution and feedback module controls the start and stop of the micro electric heating unit based on the data monitored by the flexible capacitive sensor, so as to trigger the shape memory polymer-based composite sealant to recover its shape.

[0010] In one possible implementation, the correction module compares the actual output current value of the photovoltaic system with the predicted current value at the same time to calculate the current deviation rate, and inputs the current deviation rate and the trend of the current deviation rate into the fuzzy proportional-integral-derivative controller to adjust the target value of the grounding impedance in real time.

[0011] In one possible implementation, a miniature temperature sensing element is integrated inside the cable interconnect interface. The execution and feedback module continuously monitors the connection point temperature data collected by the miniature temperature sensing element and triggers an alarm when the temperature exceeds a set threshold.

[0012] In one possible implementation, the housing surface of the cable interconnect interface has a self-cleaning nano-coating with a thickness of 5-10 micrometers and a contact angle greater than or equal to 110 degrees.

[0013] In one possible implementation, the execution and feedback module sends the periodically collected actual system operation data to the data processing module, and the prediction module periodically uses the newly added actual operation data to retrain and optimize the parameters of the deep learning-based time series prediction model.

[0014] In one possible implementation, the cross-connected enclosure also includes a redundant protective layer disposed on the inside of the enclosure, the redundant protective layer containing expandable microcapsule buffer strips; when the execution and feedback module detects that the humidity or dust concentration inside the enclosure exceeds a set threshold, it triggers the expandable microcapsule buffer strips to expand to form a secondary sealing barrier, and simultaneously sends a remote alarm signal.

[0015] In one possible implementation, the data processing module preprocesses the acquired data by: filling missing values ​​in the data sequence using linear interpolation, identifying and removing outliers using the 3σ criterion, and normalizing the numerical data to map it to the [0,1] interval.

[0016] This specification provides an adaptive processing system for cable cross-connection. This system, by constructing a complete processing system integrating environmental perception, output prediction, strategy pre-planning, and real-time closed-loop correction, achieves a fundamental shift in cable cross-connection from "passive response" to "active prediction and adaptation." The system can proactively adjust operating parameters, effectively mitigating the impact of photovoltaic output fluctuations, significantly reducing circulating current losses and fault risks. Simultaneously, its adaptive sealing and connection protection mechanisms greatly improve long-term operational stability and reliability under extreme environments, thereby enhancing overall power transmission efficiency and reducing maintenance intervention needs and costs. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of an adaptive processing system for cable cross-connection provided in one embodiment of this specification. Detailed Implementation

[0018] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0019] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0020] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0021] This specification provides an adaptive processing system for cable crossover, which will be described in detail in the following embodiments.

[0022] See Figure 1 , Figure 1 This document illustrates a schematic diagram of an adaptive processing system for cable cross-interconnection according to an embodiment of this specification. Specifically, it includes a data processing module for acquiring weather forecast data, photovoltaic module temperature data, historical photovoltaic system output data, and internal status monitoring data of the cable cross-interconnection enclosure, and preprocessing the acquired data; a prediction module for inputting the preprocessed data into a deep learning-based time series prediction model and outputting a predicted curve for photovoltaic output; a planning module for generating a timing table of operating parameter adaptation strategies for the cable cross-interconnection system for future time periods based on the predicted curve and a pre-established correlation model; a correction module for comparing the actual operating parameters of the photovoltaic system with the predicted values ​​corresponding to the predicted curve in real time, and dynamically adjusting the operating parameter adaptation strategy based on a fuzzy control algorithm when the deviation exceeds a set threshold; and an execution and feedback module for controlling the actions of the execution units of the cable interconnection interface and the cross-interconnection enclosure according to the operating parameter adaptation strategy or the dynamically adjusted operating parameter adaptation strategy, and collecting actual system operating data to feed back to the data processing module.

[0023] The data processing module can refer to the unit responsible for aggregating and integrating information from multiple sources. For example, this module obtains 72-hour weather forecast data from a meteorological service API via a network interface, retrieves historical output logs from the photovoltaic power plant monitoring system, and receives real-time monitoring data from sensors installed in cable cross-connection boxes via wired or wireless means. Preprocessing refers to operations performed to improve data quality and usability, including filling in missing data points caused by communication interruptions in the data sequence, identifying and removing outlier data caused by momentary sensor failures, and converting raw data with different physical units and magnitudes to a unified numerical range for subsequent model processing. The prediction module can refer to the functional unit that uses machine learning algorithms to estimate future states. Its built-in time series prediction model infers the future power generation trend of the photovoltaic system by analyzing the temporal correlation patterns of multi-dimensional data such as weather, temperature, and historical output, and outputs the result as a curve. The planning module can refer to the decision-making unit that transforms abstract prediction results into specific executable operational instructions. This module internally stores physical or empirical models describing the mapping relationship between the system's electrical characteristics (such as current and temperature) and optimal control parameters (such as grounding impedance). Based on predicted future operating conditions, it searches for or calculates the optimal parameters matching the model and arranges them into a detailed strategy plan in chronological order. The correction module refers to the control unit responsible for addressing prediction uncertainties and ensuring the system's real-time stable operation. This module continuously compares predicted values ​​with actual measured values. When a significant deviation is detected, it activates a built-in intelligent control algorithm to quickly adjust the strategy issued by the planning module online to adapt to the current real-world situation. The execution and feedback module can be considered the "hand" and "eye" that interacts with the physical world. On one hand, this module translates digital control commands (such as "start heating" or "adjust impedance to a certain value") into physical operations on actuators such as cable interface locking mechanisms and enclosure heating units. On the other hand, it collects the system's actual response status (such as actual current, actual temperature, and sealing status) after command execution through a sensor network and sends this information back to the upstream modules of the system, forming an information closed loop.

[0024] The present invention will be further described below through a detailed embodiment: This adaptive cable cross-interconnection processing system was applied to the collector lines of a large photovoltaic power station in a desert region, aiming to improve the operational efficiency of its cable cross-interconnection section. After system startup, the data processing module begins operation. It periodically acquires detailed weather grid forecasts for the next three days from the local meteorological bureau's data interface, including solar irradiance, ambient temperature, wind speed, and wind direction data every 15 minutes. Simultaneously, it retrieves historical output records of all relevant photovoltaic arrays at 15-minute intervals from the past year through the power station's monitoring and data acquisition system, as well as the current backsheet temperature of each photovoltaic module. Furthermore, miniature sensors deployed within dozens of key cross-interconnection enclosures upload real-time measurements of humidity, dust concentration, and sealing gaps within the enclosures to the module via a wireless sensor network. Upon receiving these raw data streams, the data processing module first runs a data cleaning program. For sporadic data gaps caused by network fluctuations, linear interpolation of data from previous and subsequent times is used to fill in the gaps. Subsequently, the three-sigma criterion is applied to scan each type of data sequence, removing outliers that significantly deviate from the normal statistical distribution. Finally, all numerical data, such as irradiance, temperature, and current values, are normalized by linearly transforming them to between 0 and 1 to eliminate dimensional differences and complete the preprocessing.

[0025] The preprocessed, regularized data is pushed to the prediction module. This module loads a pre-trained deep learning model. The model first processes the input data sequence using a Long Short-Term Memory (LSTM) network layer, which effectively captures and remembers patterns and dependencies spanning long time periods, such as "sunny days are often followed by more sunny days" and "dust storms cause a sharp drop in power output." The output of the LSM layer is then fed into a gate-controlled recurrent unit (RCU) layer, which further extracts and optimizes the short-term fluctuation characteristics of the data, such as handling rapid changes in irradiance caused by passing clouds. The model ultimately outputs two prediction curves: one is a photovoltaic power output prediction curve for the next 24 hours at a 15-minute resolution; the other is a higher-precision power output prediction curve for the next 4 hours, with the error controlled at a low level.

[0026] The predicted curve is transmitted to the planning module in real time. This module calls its internally stored "impedance-current-temperature" correlation model. This model exists in the form of a three-dimensional data table or a fitting function, which clarifies the optimal grounding impedance value that minimizes the circulating current in the cable sheath and prevents overload under different combinations of load current and ambient temperature. The planning module, combining the predicted future ambient temperature with a temperature correction factor for cable current carrying capacity, dynamically calculates the allowable current carrying capacity of the cable at various future time points. Subsequently, based on the predicted output current curve and the calculated current carrying capacity, it looks up a table or performs calculations in the correlation model for each future 15-minute time point to determine a recommended target grounding impedance value, thereby generating a detailed 24-hour impedance adaptation strategy timing table. For example, the strategy table might indicate that the grounding impedance should be adjusted to a lower value before the predicted midday peak output period to reserve sufficient current carrying capacity margin.

[0027] The system enters the real-time operation phase. The correction module begins working, collecting the actual output current value at the photovoltaic array's combiner every 15 minutes and comparing it with the predicted current value output by the prediction module to calculate the current deviation rate. Under mostly stable weather conditions, the deviation rate is small, and the system defaults to executing the pre-planned strategy generated by the planning module. However, when unpredictable short-term strong sandstorms occur, the actual output may drop rapidly, causing the current deviation rate to increase sharply and exceed the set threshold of 8%. At this time, the correction module is immediately activated. It sends the current deviation rate and its rate of change as input to a fuzzy proportional-integral-derivative (PID) controller. The controller's embedded fuzzy rule base can reason like an experienced engineer: for example, "If the deviation is large and is increasing rapidly, a large and rapid adjustment is needed." Based on this, the controller dynamically adjusts the coefficients of the proportional, integral, and derivative terms and outputs an instantaneous correction amount for the target grounding impedance value. This newly calculated impedance value will overwrite the value in the original strategy table and be immediately distributed, thereby quickly reducing system circulating current losses and avoiding energy waste.

[0028] The execution and feedback module is the final executor of the strategy. It receives the final impedance target command from the planning or correction module and adjusts the grounding impedance of the corresponding cable segment via power electronic devices. Simultaneously, it controls the rapid locking and unlocking mechanism of the cable interconnect interface and the heating unit of the enclosure's active sealing layer. For example, when the enclosure sensor detects that the sealing gap exceeds the limit, the execution and feedback module drives the heating wire to work, causing the sealant to expand and fill the gap. This module also undertakes data acquisition tasks, continuously collecting data such as the adjusted actual system impedance, current, and internal environmental conditions, and packaging this valuable operational data as new samples to feed back to the data processing module. These new samples are used to periodically retrain the deep learning model of the prediction module, allowing its predictive ability to continuously evolve with accumulated operating time, forming a self-optimizing closed loop.

[0029] The beneficial effect of this embodiment lies in that, by constructing a processing system that integrates data fusion, intelligent prediction, forward-looking planning, real-time correction, and closed-loop feedback, the management mode of cable cross-interconnection is transformed from a passive "post-event remediation" to a proactive "pre-event prediction and dynamic adaptation." This system can effectively address the challenges of drastic power output fluctuations in desert photovoltaic power plants, optimize system parameters in advance, significantly reduce circulating current losses and fault risks, and ensure the reliability of physical connections and seals through intelligent actuators, thereby improving overall power transmission efficiency, system stability, and the level of intelligent operation and maintenance.

[0030] In the aforementioned adaptive processing system for cable cross-connection, the time series prediction model based on deep learning consists of a long short-term memory network layer and a gated recurrent unit layer connected in series. The photovoltaic power output prediction curve output by the prediction module includes a medium-to-long-term prediction curve for the next 24 hours and a short-term prediction curve for the next few hours at 15-minute intervals.

[0031] The Long Short-Term Memory (LSTM) network layer can refer to a special type of recurrent neural network structure. It incorporates mechanisms such as input gates, forget gates, and output gates, enabling it to selectively remember or forget historical information. This makes it highly suitable for processing time-series data with long-term dependencies, such as weather and power output. The gated recurrent unit (GRU) layer can refer to another type of recurrent neural network structure. Compared to the LTM network, it is simpler, has fewer parameters, and trains faster, while still effectively capturing short-term dependencies in sequential data. Connecting the two means using the output sequence of the LTM network layer as the direct input to the GRU layer. This structure combines the former's strength in remembering long-term patterns with the latter's strength in handling short-term fluctuations.

[0032] In the specific implementation of the prediction module, the fusion prediction model employs a hierarchical processing architecture. Preprocessed multidimensional time-series data (such as irradiance, temperature, and historical power output sequences over multiple days) are first input into the first layer—the Long Short-Term Memory (LSTM) network layer. This layer is configured with multiple memory units, and through learning, it can identify and remember macroscopic patterns and cyclical regularities that may span several days or even weeks, such as "power output stabilizes at a high level after a period of sunny weather" or "power output periodically declines during sandstorm seasons." The output of the LTM network layer is a high-dimensional sequence that extracts long-term features.

[0033] This sequence is then fed into the second layer—the gated cyclic unit layer. This layer receives feature sequences that already contain long-term regularity information and focuses on resolving finer, short-term details of change. For example, it can respond more sensitively to rapid fluctuations in irradiance over the next few hours caused by the passage of local clouds. The gated cyclic unit layer, with its efficient computational characteristics, rapidly optimizes and integrates these short-term fluctuation features.

[0034] The fully connected layer at the end of the model maps the final features output by the gated recurrent unit layer to specific power values. Through this cascaded structure, the model works collaboratively: the long short-term memory network layer lays an accurate "tone" and trend for prediction, while the gated recurrent unit layer refines the fluctuations in the near future based on this "tone." Ultimately, the system outputs two curves simultaneously: one is a medium- to long-term prediction curve depicting the overall trend of changes throughout the day, used to guide strategy pre-planning across the entire day; the other is a more refined short-term prediction curve focusing on the next few hours, whose higher temporal resolution (15-minute intervals) provides a more timely reference benchmark for the real-time correction module, ensuring a rapid response capability to sudden fluctuations.

[0035] The beneficial effect of this embodiment is that by adopting a series model architecture of long short-term memory network and gated recurrent unit, the complementary advantages of the two network structures are fully utilized. This not only ensures the accuracy of long-term trend prediction of photovoltaic power output, but also enhances the ability to capture short-term drastic fluctuations, thereby comprehensively improving the reliability and practicality of the prediction results and providing a solid and detailed data foundation for subsequent pre-planning and real-time correction.

[0036] In the aforementioned adaptive processing system for cable cross-connection, the cable interconnection interface is a modular quick-connect interface, including a male connector and a female connector. The male connector is crimped and fixed to the cable core, while the female connector has built-in conductive contacts and a locking mechanism. The execution and feedback module achieves quick connection and disconnection by controlling the locking mechanism.

[0037] Modular quick-connect interfaces can refer to pre-assembled, standardized connection components designed to simplify field operations. A male connector refers to the plug portion of the interface, typically pre-fitted securely to the conductor end of the cable via mechanical crimping. A female connector refers to the socket portion of the interface, fixedly installed at the end of equipment or another cable, containing electrical contacts that mate with the male connector. Conductive contacts refer to the metal components inside the female connector used to establish an electrical connection with the male connector conductor; they are typically made of highly conductive and elastic materials to ensure low and stable contact resistance. A locking mechanism refers to a mechanical or electromechanical device used to secure the male connector after insertion into the female connector, preventing loosening due to vibration or external force. It can be operated manually (e.g., by pressing a button) or controlled by a micro-motor or electromagnet driven by an actuation and feedback module.

[0038] In the implementation of this system, the cable interconnect interface is designed as a plug-and-play component. For two cable segments that need to be interconnected, the installers first use specialized crimping tools in the factory or on-site pretreatment area to firmly crimp the male connector to the conductors of both cables. This process ensures a one-time, low-resistance electrical connection. The male connector is encased in a high-strength, environmentally resistant shell.

[0039] The female connector is pre-installed inside the designated inlet port of the cross-connect enclosure or as a separate connector housing. Inside the female connector, a set of elastic conductive springs are precisely arranged as conductive contacts. An electrically operated locking mechanism, controlled by an execution and feedback module, is integrated into the female connector housing. When connection is required, the installer simply aligns the male cable end with the female connector and pushes it into the female connector socket. Once the male connector is in place, the execution and feedback module receives a position sensor signal and automatically drives a micro-motor within the locking mechanism to push the locking tongue or rotate the latch, firmly locking the male connector inside the female connector. This ensures that the conductive contacts and the male conductor reach the preset contact pressure, completing the electrical connection. The entire process eliminates the need for various wrenches and torque control tools required for traditional bolted connections, and requires no multiple people; a single person can complete the process quickly, greatly improving installation efficiency and reducing the skill requirements for installers. When maintenance or replacement is needed, the execution and feedback module can reverse the locking mechanism to release the male connector, achieving rapid separation.

[0040] The beneficial effects of this embodiment are that, through modular design and integrated controllable locking mechanism, it achieves fast, convenient and reliable connection of cable interconnection interface, significantly shortens on-site installation and maintenance operation time, reduces reliance on complex tools and manual operation intensity, improves the level of construction standardization and operation safety, and is particularly suitable for application in desert areas with harsh environment and short construction window.

[0041] In the aforementioned adaptive processing system for cable cross-connection, the cross-connection enclosure includes an enclosure structure, an environmental sensing layer, an active sealing layer, and a controller. The environmental sensing layer includes a flexible capacitive sensor embedded in the enclosure's sealing area, used to monitor changes in the sealing gap and the depth of wind and sand erosion in real time. The active sealing layer contains a shape memory polymer-based composite sealant and a micro electric heating unit integrated therein. The execution and feedback module controls the start and stop of the micro electric heating unit based on the data monitored by the flexible capacitive sensor, thereby triggering the shape memory polymer-based composite sealant to recover its shape.

[0042] The enclosure structure can refer to the main shell constructed from welded or assembled metal sheets, providing physical protection and a mounting base for internal electrical components. The environmental sensing layer refers to a functional layer deployed on the enclosure to detect the degree of external environmental damage, specifically composed of a series of sensors. Flexible capacitive sensors refer to sensing elements manufactured using flexible circuit boards. Their measurement principle is based on the change in capacitance value with variations in the medium between the plates (such as air or dust) or the distance between the plates (sealing gap). Due to their flexibility and bendability, they can be fitted to complex curved surfaces or gaps. An active sealing layer refers to a sealing structure with self-healing capabilities, rather than a traditional static seal. Shape memory polymer-based composite sealant refers to a gel or paste-like material with a shape memory polymer as the matrix, possibly mixed with reinforcing fibers, conductive particles, and other functional fillers. It can recover from a temporary shape to a pre-set permanent shape at a specific temperature (phase transition temperature). A micro-electric heating unit refers to a very low-power resistance heating element, such as an extremely fine metal heating wire or heating plate, which can be precisely controlled to generate localized heat.

[0043] Part Three: In the system's enclosure design, the enclosure structure is constructed of corrosion-resistant steel plate with an anti-corrosion surface treatment. The area between the enclosure cover and the base, as well as the cable inlet, are critical weak points in the sealing. The core of the environmental sensing layer—a flexible capacitive sensor—is tightly installed inside the sealing grooves or seams at these locations. This sensor, in the form of a thin film, is attached to the enclosure wall; its extremely thin thickness (no more than 0.5 mm) ensures that it does not affect the installation and performance of the original sealing structure. It continuously monitors the capacitance parameters at its installation location, which are directly related to the thickness of the sand and dust accumulation on the sensor surface (i.e., the erosion depth) and the minute gap between the sensor and the opposite enclosure structure (i.e., the sealing gap). The monitoring data is transmitted to the controller inside the enclosure via thin wires.

[0044] The active sealing layer serves as the primary sealing barrier. During enclosure assembly, shape memory polymer-based composite sealant is filled or pre-formed into the grooves or joints requiring sealing, placing it in a compressed "temporary shape." Miniature electric heating units (such as nichrome wires) are pre-embedded or adhered to the sealant or its vicinity.

[0045] The enclosure controller acts as a local agent for the execution and feedback module within the enclosure. It receives data from flexible capacitive sensors in real time. When analysis reveals that the monitored sealing gap exceeds a preset safety threshold (e.g., 0.2 mm), the controller determines that the sealing performance has begun to decline. At this point, it immediately sends a start command to the miniature electric heating unit. Upon power-up, the heating unit rapidly generates heat, raising the temperature of the surrounding shape memory polymer-based composite sealant. When the temperature reaches its specific phase transition temperature range (e.g., 60-80 degrees Celsius), the sealant is "activated," its molecular chains begin to move, and the material strives to recover from its compressed temporary shape to its original, uncompressed "permanent shape." This recovery process generates outward expansion force, thereby tightly filling the sealing gap that has gradually increased due to material aging, vibration, or temperature deformation, achieving automatic repair of the sealing performance. Once the sensor detects that the gap has returned to a safe range, the controller shuts off the heating unit, stopping heating. Through this proactive "sensing-triggering-recovery" logic, the enclosure can automatically repair itself in the early stages of sealing failure, nipping the problem in the bud.

[0046] Part Four: The beneficial effect of this embodiment lies in the fact that by integrating high-precision flexible sensing technology with intelligent active sealing materials, the cable cross-interconnection box is endowed with the ability of "self-sensing" and "self-repair." This design can monitor the sealing status in real time and automatically initiate the repair procedure in the early stage of failure, which greatly improves the long-term sealing reliability and protection level of the box in harsh environments such as strong winds, sandstorms, and large temperature differences. It effectively prevents damage to internal components caused by sand and moisture intrusion, and significantly reduces the operation and maintenance costs and power outage risks caused by sealing failure.

[0047] In the aforementioned adaptive processing system for cable cross-interconnection, the correction module compares the actual output current value of the photovoltaic system with the predicted current value at the same time to calculate the current deviation rate, and inputs the current deviation rate and the trend of the current deviation rate into the fuzzy proportional-integral-derivative controller to adjust the target value of the grounding impedance in real time.

[0048] The current deviation rate refers to the ratio of the absolute value of the difference between the actual measured current value and the predicted current value to the predicted current value, expressed as a percentage. It is used to quantify the degree of deviation between the prediction and the actual value. The trend of the current deviation rate refers to how the deviation rate changes over time, such as whether it is increasing, decreasing, or remaining stable. This reflects the speed and direction of the system's deviation from the predicted state. A fuzzy proportional-integral-derivative (PID) controller can refer to an advanced controller that combines fuzzy logic control with traditional proportional-integral-derivative (PID) control. It first uses a fuzzy inference system to infer the fuzzy quantities that adjust the proportional, integral, and derivative control parameters based on the fuzzy description of the input variables (such as "large deviation rate" or "positive trend") and a fuzzy rule base established by expert experience. After defuzzification, the precise parameter adjustment quantities are obtained, thereby dynamically changing the response characteristics of the entire controller.

[0049] Part Three: During the real-time operation of the system, the correction module acts as a "corrector." At fixed short intervals (e.g., 15 minutes), it reads the instantaneous or average value of the actual output current at the photovoltaic array's combiner from the power station's current transformer. Simultaneously, it obtains the predicted current value corresponding to the same moment from the prediction module. The core algorithm of the correction module calculates the absolute difference between these two values, divides it by the predicted current value, and obtains the instantaneous current deviation rate.

[0050] In addition to the current deviation rate value, the algorithm also calculates the change in the deviation rate over the most recent periods to determine the trend of the deviation, such as whether it is "rapidly increasing", "slowly decreasing" or "basically stable".

[0051] Subsequently, the two key inputs—the magnitude of the current deviation rate and its changing trend—are fed into a fuzzy proportional-integral-derivative (PID) controller. This controller has a pre-set set of fuzzy rules simulating human expert decision-making. For example, one rule might be: "If the deviation rate is large and the trend is rapidly increasing, then the proportional and integral actions of the controller need to be significantly enhanced." The controller converts precise input values ​​into fuzzy language descriptions such as "large," "medium," "small," "positive," and "negative." By applying these rules through a fuzzy inference engine, it derives fuzzy conclusions on how the proportional, integral, and derivative coefficients should be adjusted, and then defuzzifies these conclusions into specific coefficient adjustment amounts.

[0052] Based on the newly adjusted control coefficients, the fuzzy proportional-integral-derivative (FID) controller quickly calculates a new, more suitable target grounding impedance value for the current instantaneous operating conditions, taking into account the actual current, ambient temperature, and system model. This new target value replaces the value at the corresponding time point in the original pre-planned strategy and is immediately issued to the execution and feedback modules for implementation. For example, when unexpected cloud cover causes the actual current to suddenly fall below the predicted value, the control system quickly increases the grounding impedance to reduce unnecessary circulating current losses; conversely, when sunlight suddenly intensifies, it quickly decreases the impedance to provide sufficient current-carrying margin. Through this dynamic adjustment based on fuzzy reasoning, the system can smoothly and intelligently cope with various unforeseen fluctuations, maintaining a consistently efficient and safe operating state.

[0053] The beneficial effect of this embodiment lies in the introduction of fuzzy proportional-integral-derivative control based on the current deviation rate and its changing trend, enabling the system's real-time correction mechanism to possess intelligent judgment and flexible adjustment capabilities similar to human experience. This method can adaptively change the control strength and strategy according to the severity and development trend of the deviation, achieving a rapid, accurate, and stable response to sudden operating conditions. It effectively compensates for the shortcomings of predictive models, ensuring the dynamic optimization and stability of the cable cross-connection system under complex all-weather operating conditions.

[0054] In one possible implementation, a miniature temperature sensing element is integrated inside the cable interconnect interface. The execution and feedback module continuously monitors the connection point temperature data collected by the miniature temperature sensing element and triggers an alarm when the temperature exceeds a set threshold.

[0055] Miniature temperature sensing elements refer to temperature sensors that are extremely small and can be embedded in a compact space, such as negative temperature coefficient thermistors or miniature digital temperature sensor chips. Connection point temperature refers to the temperature of the actual contact area between the male and female conductive contacts inside the cable interconnect interface. This temperature directly reflects the contact resistance and connection quality; abnormal temperature rise is usually an early sign of poor contact or overload.

[0056] In the specific design of the modular high-speed cable interconnect interface, to achieve direct and accurate monitoring of the connection status, a miniature temperature sensing element is encapsulated in a critical location inside the interface—typically near the area where the conductive contacts of the male and female connectors are crimped or in contact. This element is extremely small, and its leads are carefully arranged to avoid affecting the mechanical strength and electrical insulation of the interface.

[0057] This miniature temperature sensing element transmits the acquired analog or digital temperature signals to the execution and feedback module (or its subordinate local data acquisition unit) via a thin wire or by using an interface housing as a communication channel. The execution and feedback module has built-in logic that reads the temperature value at a certain frequency (e.g., once per second).

[0058] The system presets a temperature alarm threshold, which is typically set based on a combination of the cable's rated current carrying capacity, ambient temperature, and the safe operating temperature of the materials, such as 80 degrees Celsius. The execution and feedback module continuously compares the real-time connection point temperature with this threshold. As long as the temperature remains below the threshold, the system considers the interface to be working normally. Once the temperature exceeds the set threshold, the execution and feedback module immediately determines it to be in an abnormal state. This triggers an alarm event, which can manifest in various ways: for example, illuminating a warning light on the local enclosure controller, sending an alarm message containing the interface location and over-temperature data to a remote central monitoring platform via the communication network, or even linking the alarm signal with the power plant's protection system. This allows maintenance personnel to receive timely warnings before overheating could cause interface burnout or lead to more serious failures, enabling them to locate and address the issue effectively, thus ensuring preventative maintenance is implemented.

[0059] The beneficial effect of this embodiment is that by integrating a miniature temperature sensor inside the cable interconnect interface, direct, online monitoring of the operating status of the core electrical connection components is achieved. This design can promptly detect potential overheating hazards caused by increased contact resistance, overload, or abnormal environment, and issue early warnings. It transforms the fault handling mode from reactive maintenance to proactive early warning, greatly enhancing the safety and reliability of interface operation and providing effective protection against fires or power outages caused by overheating at connection points.

[0060] In one possible implementation, the housing surface of the cable interconnect interface has a self-cleaning nano-coating with a thickness of 5-10 micrometers and a contact angle greater than or equal to 110 degrees.

[0061] Self-cleaning nano-coatings can refer to a special functional coating applied to the surface of a material. Their microstructure or chemical properties make them difficult for contaminants to adhere to, or easily carried away by natural forces such as wind and rain. The contact angle is a physical quantity that measures the wetting performance of a liquid on a solid surface. A contact angle greater than or equal to 110 degrees usually indicates that the surface is strongly hydrophobic (to water) or oleophobic, causing droplets to spherically roll off and carry away surface dust particles.

[0062] To address the severe dust accumulation problem in desert regions and extend the interface maintenance cycle, the housings of the modular cable interconnect interfaces (including the male and female connector shells) undergo special surface treatment during manufacturing. A self-cleaning nano-coating is uniformly sprayed or plated onto its outermost layer. The thickness of this coating is precisely controlled between 5 and 10 micrometers to ensure functionality without affecting the interface's mechanical dimensions and assembly precision.

[0063] This nano-coating endows the shell surface with superhydrophobic or superoleophobic properties by altering its microstructure and surface energy. Its key performance indicator—the contact angle with water—is designed to be greater than or equal to 110 degrees. This means that when fine sand particles (which typically adsorb trace amounts of moisture or oil) attempt to adhere to the interface shell, their adhesion to the coating surface is weak. Under natural wind, slight thermal expansion and contraction of the shell due to diurnal temperature variations, or occasional raindrops, the loosely adhered sand particles easily detach from the shell surface. Even in desert environments without rainfall, strong winds can blow away most of the loose sand, preventing significant accumulation and hardening of sand on the interface surface, especially in the gaps of the locking mechanism. This "lotus leaf effect"-like self-cleaning ability significantly reduces the risk of sand intruding into the interface and also reduces the possibility of sand covering affecting heat dissipation or hindering manual operation (such as scanning QR codes or pressing buttons).

[0064] The beneficial effect of this embodiment is that by applying a self-cleaning nano-coating with specific parameters to the surface of the interface housing, the outer surface of the interface housing is endowed with continuous dustproof and anti-adhesion capabilities. This effectively alleviates the adverse effects of sand and dust accumulation on interface performance and maintenance operations in desert environments, reduces the frequency and workload of cleaning and maintenance, helps maintain the interface in good working condition and visually identifiable appearance, and improves the interface's adaptability and long-term reliability in harsh environments from another dimension.

[0065] In the aforementioned adaptive processing system for cable cross-connection, the execution and feedback module sends the periodically collected actual system operation data to the data processing module, and the prediction module periodically uses the newly added actual operation data to retrain and optimize the parameters of the deep learning-based time series prediction model.

[0066] Periodic data collection refers to the regular collection of data at fixed time intervals or under certain triggering conditions. New operational data refers to newly generated measurement data containing the latest operating conditions after the system is put into operation. This data was not available during the initial model training phase and contains the latest system characteristics and environmental patterns.

[0067] The system is designed with a continuous self-improving feedback optimization mechanism. The execution and feedback module is not only the executor of commands but also the collector of the system's operational "health report." It packages the massive amounts of actual operational data collected over a period of time on a daily or weekly basis. These data packages are rich in content, including: the actual photovoltaic output current at various time points, the actual ambient temperature, the actual wind speed, the actual temperature and humidity inside the enclosure, and the actual grounding impedance value of the system.

[0068] These data packets are sent back to the data processing module via the communication network. The data processing module performs preprocessing operations such as cleaning and formatting on these new, actual operational data, just like it does on the initial data, and marks them with the corresponding timestamps.

[0069] Subsequently, according to a pre-defined optimization plan (e.g., every Sunday morning), the prediction module initiates an incremental learning or retraining process for the model. It mixes newly pre-processed real-world data with existing historical data to create a new training dataset. The prediction module then invokes its internal deep learning framework to retrain the model using this updated and more comprehensive dataset, starting with the current model parameters. During training, the optimization algorithm fine-tunes thousands of parameters in the model, such as neuron connection weights, based on the latest patterns contained in the new data (e.g., the impact of a newly emerging weather pattern on power output).

[0070] Through this periodic retraining, the deep learning-based time series prediction model can continuously "review and learn new things," enabling its predictive capabilities to keep pace with the times and better adapt to long-term, gradual changes such as the performance degradation of photovoltaic power plant equipment over the years and subtle variations in local climate patterns. This allows the system's prediction accuracy to remain at a high level over the long term, and even gradually improve with data accumulation, ensuring the effectiveness of the entire predictive adaptive system lifecycle.

[0071] The beneficial effect of this embodiment lies in establishing a periodic feedback optimization loop from actual operating data to the predictive model, enabling the entire system to continuously learn and self-evolve. This breaks the limitation of traditional control system parameters being fixed once set, allowing the predictive model to dynamically track and adapt to long-term changes in the power plant itself and the external environment. This ensures that the system maintains high-precision predictive performance and superior adaptive control effects throughout its entire life cycle, achieving long-term intelligent and optimized operation.

[0072] In one possible implementation, the cross-connected enclosure also includes a redundant protective layer disposed on the inside of the enclosure, the redundant protective layer containing expandable microcapsule buffer strips; when the execution and feedback module detects that the humidity or dust concentration inside the enclosure exceeds a set threshold, it triggers the expandable microcapsule buffer strips to expand to form a secondary sealing barrier, and simultaneously sends a remote alarm signal.

[0073] Among them, redundant protection layers can refer to backup protection structures set up in addition to the main protection measures, designed to provide supplementary protection when the main protection fails. Expandable microcapsule buffer strips can refer to a strip-shaped material in which a large number of micron-sized capsules are uniformly dispersed inside. The capsule walls are filled with a low-boiling-point liquid foaming agent. When heated, the capsules rupture, the foaming agent rapidly vaporizes and expands, causing the entire material volume to increase significantly.

[0074] In the design of the actively sealed cross-connection enclosure, to cope with extreme situations (such as complete failure of the active seal, sensor malfunction, or instantaneous penetration by extreme wind and sand), an additional redundant protective layer is added to the inside of the enclosure, adjacent to the active seal. This layer is specifically manifested as a ring of expandable microcapsule buffer tape adhered to the inner wall of the enclosure cover and the inside of all cable inlets. This buffer tape is soft and adhesive under normal conditions, approximately a few millimeters thick, and will not affect the installation and heat dissipation of components inside the enclosure.

[0075] The enclosure controller continuously monitors environmental sensor data within the enclosure, including humidity and laser particulate matter (dust) concentration sensors. The system has set higher alarm thresholds for these two parameters, such as relative humidity ≥80% and dust concentration ≥10 mg / m³. These thresholds indicate a significant deterioration in the enclosure's environment, suggesting that the active sealing layer may have failed to effectively prevent intrusion.

[0076] When the execution and feedback module (via the enclosure controller) detects that either the humidity or dust concentration inside the enclosure exceeds its set high-risk threshold, it determines that the active sealing layer may have failed, indicating an emergency. At this point, the module immediately activates its redundancy protection plan: it first sends a start command to the micro-heating wires embedded in the expandable microcapsule buffer strip (or heats the entire buffer strip through other means). Heating causes the buffer strip temperature to rise to 40-50 degrees Celsius within a short time. At this temperature, the uniformly distributed microcapsule wall material inside the buffer strip softens and ruptures, causing the liquid foaming agent encapsulated inside to rapidly evaporate into gas, resulting in a dramatic volume expansion.

[0077] This process causes the entire buffer zone to expand to 3 to 5 times its original size within tens of seconds, transforming into a soft and dense porous foam that tightly fills and blocks all possible intrusion channels, such as gaps in the enclosure lid and cable inlets, forming a physical secondary sealing barrier. This forcibly blocks further intrusion of wind, sand, and moisture, buying valuable time for emergency repairs. Simultaneously, the execution and feedback module immediately sends the highest-level remote alarm signal to the remote maintenance monitoring center via its GPRS or wireless private network module, clearly indicating "Active sealing of enclosure XX has failed, redundant protection has been activated, and urgent maintenance is needed," ensuring that maintenance personnel can be informed of the serious fault and locate and handle it immediately.

[0078] The beneficial effect of this embodiment lies in the fact that by setting a redundant protective layer based on expandable microcapsules inside the enclosure and linking it with the internal environmental monitoring, an ultimate protection mechanism of "active and passive combined, double insurance" is constructed for the cable cross-interconnection enclosure. Even if the active sealing system fails unexpectedly, this redundant layer can respond quickly, automatically forming an emergency sealing barrier to prevent the accident from escalating to the greatest extent, and immediately notify personnel for maintenance. This greatly improves the resilience and safety of the system under extreme failure scenarios and avoids significant losses caused by the destruction of all internal components of the enclosure due to a single point of failure.

[0079] In the aforementioned adaptive processing system for cable cross-connection, the data processing module performs the following preprocessing on the acquired data: filling missing values ​​in the data sequence using linear interpolation, identifying and removing outliers using the three sigma criterion, and normalizing numerical data to map it to the [0,1] interval.

[0080] Linear interpolation can be a simple data imputation method that assumes the data changes linearly between two known data points, thus estimating the value of missing points using a straight line. The three-sigma criterion is an outlier identification method based on the assumption of a normal data distribution. It assumes that the vast majority of data (approximately 99.7%) falls within the range of the mean plus or minus three standard deviations; data points outside this range are considered low-probability events, i.e., outliers. Normalization refers to scaling data proportionally to fit it into a specific interval (e.g., 0 to 1) to eliminate differences in units and value ranges between different features.

[0081] In the data processing module's workflow, preprocessing is a crucial step in ensuring data quality. After acquiring raw data streams from multiple sources, including meteorological APIs, power plant monitoring systems, and enclosure sensors, the module first checks the data's integrity. Due to potential network transmission interruptions or occasional sensor malfunctions, data points may occasionally be missing in the time series data. For these missing points, the module uses linear interpolation. Specifically, it finds the two closest valid data points before and after the missing time point, assumes the data change between these two points is uniform, and then calculates and fills in the missing point's estimated value according to the time ratio, thus ensuring the continuity of the time series.

[0082] After filling in missing values, the module then performs data cleaning to remove "noise." It uses the three-sigma criterion to identify outliers. For each type of data sequence (such as a light intensity sequence), the module first calculates the arithmetic mean and standard deviation of all valid data in the sequence. Then, it expands the range of data by three times the standard deviation both upwards and downwards from the mean, forming a reasonable range. Any data point falling outside this range is considered an outlier due to transient interference, sensor false alarms, etc., and is discarded. The discarded positions will again form a "missing value," which is then filled using linear interpolation.

[0083] Even after cleaning, the various features (such as illumination, temperature, and current values) still have different physical units and numerical ranges. Directly inputting these into the model would lead to features with large dimensions dominating the training process. Therefore, the final step is to normalize all numerical features. The data processing module calculates the minimum and maximum values ​​for each feature across the entire training dataset (or a sliding time window). Then, for each data point's value for that feature, the formula (current value - minimum value) / (maximum value - minimum value) is used to linearly map it to a closed interval between 0 and 1. After this step, all features are transformed to the same scale, laying a solid foundation for the stable and efficient training of the subsequent deep learning model.

[0084] The beneficial effect of this embodiment lies in its effective handling of common problems in the original data, such as missing values, anomalies, and inconsistencies in units, through a complete and rigorous data preprocessing workflow, including linear interpolation imputation, three-sigma criterion cleaning, and maximum-minimum normalization. This significantly improves the data quality and consistency input to the prediction model, ensuring accurate and reliable learning results for the model, and is a crucial prerequisite for the entire intelligent processing system to perform effectively. It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the embodiments in this specification. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0085] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0086] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. An adaptive processing system for cable cross-connection, characterized in that, include: The data processing module is used to acquire weather forecast data, photovoltaic module temperature data, historical output data of photovoltaic system and internal status monitoring data of cable cross-interconnection box, and to preprocess the acquired data; The prediction module is used to input the preprocessed data into a deep learning-based time series prediction model and output the prediction curve of photovoltaic power output. The planning module is used to generate a timing table of operating parameter adaptation strategies for cable cross-connection systems for future time periods based on the predicted curves and pre-established correlation models. The correction module is used to compare the actual operating parameters of the photovoltaic system with the predicted values ​​corresponding to the prediction curve in real time, and dynamically adjust the operating parameter adaptation strategy based on the fuzzy control algorithm when the deviation exceeds a set threshold. The execution and feedback module is used to control the actions of the execution units of the cable interconnection interface and the cross-connection box according to the operation parameter adaptation strategy or the dynamically adjusted operation parameter adaptation strategy, and to collect the actual operation data of the system and feed it back to the data processing module.

2. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The deep learning-based time series prediction model consists of a long short-term memory network layer and a gated recurrent unit layer connected in series. The photovoltaic power output prediction curve output by the prediction module includes a medium-to-long-term prediction curve for the next 24 hours and a short-term prediction curve for the next few hours at 15-minute intervals.

3. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The cable interconnection interface is a modular quick-connect interface, including a male connector and a female connector. The male connector is crimped and fixed to the cable core, and the female connector has built-in conductive contacts and a locking mechanism. The execution and feedback module controls the locking mechanism to achieve connection and disconnection.

4. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The cross-connected enclosure includes an enclosure structure, an environmental sensing layer, an active sealing layer, and a controller; The environmental sensing layer includes a flexible capacitive sensor embedded in the sealed part of the box, used to monitor changes in the sealing gap and the depth of wind and sand erosion in real time; The active sealing layer comprises a shape memory polymer-based composite sealant and a micro electric heating unit integrated therein; The execution and feedback module controls the start and stop of the micro electric heating unit based on the data monitored by the flexible capacitive sensor, so as to trigger the shape memory polymer-based composite sealant to recover its shape.

5. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The correction module compares the actual output current value of the photovoltaic system with the predicted current value at the same time to calculate the current deviation rate, and inputs the current deviation rate and the trend of the current deviation rate into the fuzzy proportional-integral-derivative controller to adjust the target value of the grounding impedance in real time.

6. The adaptive processing system for cable cross-connection according to claim 3, characterized in that, The cable interconnection interface integrates a miniature temperature sensing element. The execution and feedback module continuously monitors the connection point temperature data collected by the miniature temperature sensing element and triggers an alarm when the temperature exceeds a set threshold.

7. The adaptive processing system for cable cross-connection according to claim 3, characterized in that, The housing surface of the cable interconnection interface has a self-cleaning nano-coating with a thickness of 5-10 micrometers and a contact angle greater than or equal to 110 degrees.

8. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The execution and feedback module sends the periodically collected actual system operation data to the data processing module, and the prediction module periodically uses the newly added actual operation data to retrain and optimize the parameters of the deep learning-based time series prediction model.

9. The adaptive processing system for cable cross-connection according to claim 4, characterized in that, The cross-connected enclosure also includes a redundant protective layer disposed on the inside of the enclosure, the redundant protective layer containing expandable microcapsule buffer strips; when the execution and feedback module detects that the humidity or dust concentration inside the enclosure exceeds a set threshold, it triggers the expandable microcapsule buffer strips to expand to form a secondary sealing barrier, and simultaneously sends a remote alarm signal.

10. The adaptive processing system for cable cross-connection according to claim 1, characterized in that, The data processing module performs the following preprocessing on the acquired data: filling missing values ​​in the data sequence using linear interpolation, identifying and removing outliers using the 3σ criterion, and normalizing the numerical data to map it to the [0,1] interval.

Images