Photovoltaic power station relay protection intelligent operation and maintenance system based on digital twinning

Abstract

The invention discloses a photovoltaic power station relay protection intelligent operation and maintenance system based on digital twinning, and relates to the technical field of photovoltaic power station operation and maintenance. The time sequence layered mapping module is used for establishing a time sequence layered mapping model based on a real-time operation signal of the photovoltaic power station, expanding the collected current signal and voltage signal into a continuous mapping band on multiple time scales, and extracting a response lag boundary under a high-load dynamic working condition; and the dynamic delay compensation module is used for constructing a dynamic delay compensation link according to a response lag boundary result of the time sequence layered mapping model, and carrying out sliding time window processing on a sampling sequence of the current signal and the voltage signal. According to the invention, through time sequence layered mapping, delay compensation and peak suppression cooperative control, time synchronization and simulation accuracy improvement of electrical signals of the photovoltaic power station are realized; meanwhile, through energy balance and self-adaptive threshold regulation and control, relay protection only responds to real faults, misoperation is avoided, and operation safety and intelligent operation and maintenance stability of the power station are guaranteed.

Description

Intelligent Operation and Maintenance System for Relay Protection of Photovoltaic Power Plants Based on Digital Twin Technical Field

[0001] This invention relates to the field of photovoltaic power plant operation and maintenance technology, specifically to a digital twin-based intelligent operation and maintenance system for photovoltaic power plant relay protection. Background Technology

[0002] Intelligent operation and maintenance of photovoltaic power plant relay protection based on digital twin technology refers to constructing a digital mapping model that is highly consistent with the actual operating state of the photovoltaic power plant. This model allows for synchronous simulation and dynamic linkage of the power plant's primary electrical equipment, secondary protection devices, communication networks, and operating environment in a virtual space, enabling intelligent monitoring, evaluation, and decision optimization of the relay protection system. This method collects multi-source data such as current, voltage, frequency, and fault signals in real time, performs time-series feature analysis and virtual fault reproduction in the digital twin model, and can identify potential risks such as protection setting drift, action delay, and maloperation or failure to operate in advance. Based on the dynamic calculation results of the virtual model, the system can achieve online verification and adaptive optimization of protection strategies, support protection coordination and intelligent alarms under multi-scenario grid-connected conditions, thereby realizing accurate diagnosis, predictive maintenance, and intelligent operation and maintenance management throughout the entire lifecycle of the photovoltaic power plant relay protection system.

[0003] Existing technologies have the following shortcomings: Under the dynamic operating conditions of high loads in photovoltaic power plants, the real-time simulation rate of digital twin models may experience a slight lag due to limitations in computation cycle and communication bandwidth. When this lag is superimposed on the measurement delays in the on-site current and voltage acquisition stages, the twin system generates electrical transient characteristics in virtual space that do not conform to the actual operating state, leading to an abnormal amplification of the virtual short-circuit current peak. This false peak is easily misjudged as a sudden short-circuit event in the relay protection logic, thereby triggering the overcurrent protection action of the circuit breaker and causing the non-faulty branch to be disconnected prematurely. Such anomalies not only cause abnormal disconnection of the combiner box circuit, disrupting the power balance between branches, but may also lead to frequent start-stop of downstream inverter units, forming a chain of malfunctions in the relay protection system, thus seriously affecting the grid connection stability and overall operational safety of the photovoltaic power plant.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent operation and maintenance system for relay protection of photovoltaic power plants based on digital twins, so as to solve the problems in the background art mentioned above.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a digital twin-based intelligent operation and maintenance system for photovoltaic power plant relay protection, comprising a time-series hierarchical mapping module, a dynamic delay compensation module, a virtual peak suppression module, a zoned energy balance module, and an adaptive threshold control module; the time-series hierarchical mapping module: establishes a time-series hierarchical mapping model based on the real-time operating signals of the photovoltaic power plant, expands the collected current and voltage signals into continuous mapping bands at multiple time scales, extracts the response lag boundary under high load dynamic conditions, and enables the simulation input of the digital twin model to have a unified and stable time reference; the dynamic delay compensation module: constructs a dynamic delay compensation link based on the response lag boundary results of the time-series hierarchical mapping model, performs sliding time window processing on the sampling sequences of current and voltage signals, corrects the signal input timing in a phase-precessing manner, and enables the digital twin model to respond to key electrical mutations in advance within the calculation cycle, thereby maintaining the time synchronization between the simulation output and the actual operating state; Virtual Peak Suppression Module: Utilizes the synchronous output of the dynamic delay compensation stage to generate a virtual peak suppression unit. During the simulation calculation of the digital twin model, it implements nonlinear attenuation control on the transient amplitude of the current to weaken the false short-circuit current peak caused by the accumulation of response lag, thereby maintaining the physical continuity of the simulation output signal. Partitioned Energy Balance Module: Based on the output of the virtual peak suppression unit, a partitioned energy balance loop is constructed. By performing time-weighted correction and dynamic allocation of the power change trends of each branch of the photovoltaic power plant, the energy distribution in the simulation space is maintained in a stable dynamic balance to prevent abnormal protection conditions triggered by false peaks. Adaptive Threshold Control Module: Based on the dynamic output of the partitioned energy balance loop, an adaptive action threshold control loop is established. It automatically adjusts the relay protection action threshold according to the energy gradient changes of each branch, ensuring that the relay protection logic only responds to real short-circuit events, thereby achieving malfunction suppression and intelligent operation and maintenance management of the photovoltaic power plant relay protection system.

[0007] Preferably, the steps for establishing a time-series hierarchical mapping model based on real-time operation signals of a photovoltaic power station include: Firstly, based on the current and voltage signals obtained from the acquisition terminals of various electrical equipment in the photovoltaic power station, the sampled data from the DC-side combiner box, inverter input terminal, inverter output terminal, and AC bus are uniformly organized. Secondly, time normalization and time synchronization comparison correction are used to align all signals on the time axis. Thirdly, based on the time-aligned signals, the current and voltage signals are divided into fast-changing, medium-changing, and slow-changing layers according to their variation characteristics in the time dimension. Finally, time window sliding is used to achieve inter-layer mapping. The continuous transition forms a continuous mapping band with multiple time scales. Based on the continuous mapping band, the time difference between current and voltage signals in each time level is analyzed to extract the response hysteresis boundary under high load dynamic conditions. The hysteresis range of the electrical response is determined according to the change law of different levels. Based on the response hysteresis boundary, the time base of the input signal of the digital twin model is adjusted, the sampling sequence is rearranged and the missing data is filled in on the time axis, so that the simulation input maintains a unified, continuous and stable time base in different levels, thereby ensuring the consistency between the digital twin simulation output and the actual operating state of the photovoltaic power station in the time dimension.

[0008] Preferably, during the time base adjustment process, the sampling sequences of the fast-changing layer, the medium-changing layer, and the slow-changing layer are rearranged in segments according to the time range of the response lag boundary, and interpolation is performed between adjacent time layers to ensure that the current signal and the voltage signal maintain a continuous correspondence at multiple time scales. This forms a complete time-aligned sequence in the simulation input stage of the digital twin model, thereby achieving the continuity and stability of the simulation input signal.

[0009] Preferably, the steps for constructing a dynamic delay compensation stage based on the response lag boundary results of the time-series hierarchical mapping model include: after obtaining the response lag boundary of the time-series hierarchical mapping model, performing time offset analysis on the sampling sequences of current and voltage signals at each measurement point of the photovoltaic power station, and establishing a continuous sliding time window according to the time range of the lag boundary to capture the changing trend of the signal on the time axis; within the sliding time window, comparing the time deviations of the current and voltage signals, and sliding the corresponding signal forward on the time axis according to the lag direction, so that its change curve maintains a synchronous response relationship within the window range, and in the time domain... A smooth and continuous transition is achieved. After phase pre-adjustment, the time-corrected signal is recombined into a new continuous time-series data band. The boundaries of adjacent sliding windows are smoothly connected, and the signals of different measurement points are dynamically prioritized according to the real-time operating status of the photovoltaic power station, so that key electrical mutations are reflected in the simulation input first. The corrected continuous time-series data band is input into the digital twin simulation calculation process, so that it responds to key electrical mutations in actual operation in advance within the calculation cycle, thereby maintaining the synchronization between the simulation output and the on-site operating status in time and avoiding the virtual peak amplification phenomenon caused by delay superposition.

[0010] Preferably, the length of the sliding time window and the phase advance amplitude are dynamically adjusted according to the response hysteresis boundary of each level in the time-series hierarchical mapping model, so that the time sliding amplitude of the rapidly changing layer is smaller than that of the medium-changing layer and the slowly changing layer, thereby realizing the synchronous response of the current signal and the voltage signal at multiple time scales, and ensuring that the digital twin simulation input maintains a continuous and consistent time alignment relationship between different levels.

[0011] Preferably, the step of generating a virtual peak suppression unit using the synchronous output of the dynamic delay compensation stage includes: after the dynamic delay compensation stage completes time synchronization adjustment, acquiring the current signal and voltage signal sequences after phase preprocessing; performing a partitioned analysis on the transient amplitude change of the current signal under high load dynamics to identify the power rapid rise region, voltage drop region, and steady-state maintenance region, and determining the time interval where the false peak occurs; within the time interval of the false peak, performing a continuous rate of change analysis on the transient amplitude change of the current signal, and establishing a smooth transition channel based on the synchronous timing reference provided by the dynamic delay compensation stage to extend the peak change within a physically reasonable range. And maintain signal continuity; based on the current signal after smooth transition, nonlinear attenuation control is implemented in the false peak region, and the current amplitude is gradually reduced according to the peak change rate and duration, so that the false peak smoothly transitions to the adjacent normal signal segment and maintains the phase consistency with the original waveform; after the false peak attenuation adjustment is completed, the current signal processed by nonlinear attenuation control and the voltage signal after delay compensation are recombined to form a new simulation input signal set, so that the digital twin simulation process receives an input signal with stable energy and continuous time, thereby weakening the false short-circuit current peak caused by hysteresis accumulation and maintaining the physical continuity of the simulation output signal.

[0012] Preferably, during the nonlinear attenuation control process, a transition zone is set on the time axis for the current signal in the false peak region. The duration of the transition zone is adaptively adjusted according to the rate of change of the peak value, so that the current amplitude gradually and smoothly transitions during the attenuation process. This ensures that the current signal after the false peak is weakened remains continuous and consistent with the adjacent normal signal segment in terms of amplitude and phase, thereby further improving the physical stability and response accuracy of the simulation output signal.

[0013] Preferably, the step of constructing a partitioned energy balance loop based on the output of the virtual peak suppression unit includes: after the virtual peak suppression unit outputs current and voltage signals after nonlinear attenuation control, the operating status of each branch of the photovoltaic power station is divided into energy characteristics. Based on the electrical connection relationship of the combiner branch, inverter input side, inverter output side, and AC bus, the electrical network is divided into multiple energy distribution regions, and the time-series distribution characteristics of the power change rate of each branch are recorded. After completing the energy region division, the power change trend of each branch is corrected by time weighting, so that branches with faster power changes receive lower time weights, and branches with slower changes receive higher time weights. The simulation process involves several steps: First, the power distribution is weighted to maintain time balance and avoid deviations caused by lag effects. Second, the power output of each branch of the photovoltaic power station is dynamically allocated according to the magnitude of the change, achieving dynamic compensation of energy between different branches and ensuring continuous balance in overall energy transfer. Third, after dynamic energy allocation, the adjusted power change results of each branch are recombined into the overall energy distribution curve in the simulation space, maintaining dynamic balance in the energy distribution within the simulation space and preventing abnormal protection conditions triggered by false peaks. This ensures that the digital twin simulation output remains consistent with the actual operating state of the photovoltaic power station.

[0014] Preferably, during the time-weighted correction process, a fixed time window is used to continuously weight the power change curves of each branch, so that the time weight of branches with high power change rates decreases proportionally, while the time weight of branches with gentle changes increases proportionally. In the dynamic allocation stage, energy compensation is given to the power fluctuation edge branches based on the weighting results to ensure the continuity of the overall energy distribution and the dynamic balance stability in the simulation space.

[0015] Preferably, the steps for establishing an adaptive action threshold control loop based on the dynamic output of the partitioned energy balance loop include: after the partitioned energy balance loop completes the time-weighted correction and dynamic energy allocation of the power changes of each branch of the photovoltaic power station, obtaining the balanced energy output result, and matching the power change rate with the current signal and voltage signal to form the branch energy gradient curve, which is used to reflect the power change trend of each branch under dynamic operating conditions; after obtaining the energy gradient curve of each branch, performing correlation analysis on the energy gradients of different branches, determining the threshold adjustment benchmark based on the time-weighted power data, raising the action threshold when the branch energy gradient continues to rise, and lowering the threshold when the energy gradient rises sharply and... As the voltage drops, the action threshold is lowered to ensure that the threshold adjustment process is consistent with changes in energy distribution. After the initial threshold adjustment is completed, the threshold change results are dynamically smoothed by connecting the threshold curves of each branch in adjacent time periods to maintain the continuity and stability of threshold changes. The time smoothing span is automatically adjusted according to the operating status of the photovoltaic power station. After the dynamic smoothing adjustment of the adaptive action threshold is completed, the updated threshold parameters are input into the relay protection logic. The protection sensitivity and action delay are automatically adjusted according to the real-time changes in the energy gradient, so that the relay protection logic only responds to real short-circuit events, thereby achieving malfunction suppression and intelligent operation and maintenance management.

[0016] The technical effects and advantages provided by this invention in the above technical solution are as follows: By introducing a synergistic mechanism of time-series layered mapping, dynamic delay compensation, and virtual peak suppression in the digital twin simulation process, this invention enables the electrical signals of a photovoltaic power station to remain continuous and consistent in the time dimension under high-load dynamic operating conditions, thereby effectively avoiding the occurrence of false current peaks caused by the superposition of simulation delay and sampling lag. Through this mechanism, the digital twin model can respond to key electrical mutations in advance within the calculation cycle, and the simulation output is highly synchronized with the actual operating state, making the input basis for relay protection judgment more realistic and reliable, and significantly improving the accuracy and stability of the protection system in identifying real short-circuit events.

[0017] This invention constructs a partitioned energy balance loop and an adaptive action threshold control loop to maintain a dynamic balance in the power distribution of a photovoltaic power plant operating in a multi-branch environment, ensuring continuous stability of the energy transfer process in the simulation space. The relay protection action threshold can be automatically adjusted according to changes in the energy gradient of each branch, ensuring that the protection logic responds only to real fault conditions, fundamentally eliminating malfunction chains caused by false signals. This scheme achieves intelligent dynamic management of photovoltaic power plant relay protection, improving the safety of system operation and the reliability of maintenance. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 is a schematic diagram of the modules of the intelligent operation and maintenance system for relay protection of photovoltaic power plants based on digital twins according to the present invention. Detailed Implementation

[0020] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0021] This invention provides a digital twin-based intelligent operation and maintenance system for photovoltaic power plant relay protection, as shown in Figure 1. It includes a time-series hierarchical mapping module, a dynamic delay compensation module, a virtual peak suppression module, a zoned energy balance module, and an adaptive threshold control module. The time-series hierarchical mapping module establishes a time-series hierarchical mapping model based on the real-time operating signals of the photovoltaic power plant. It expands the collected current and voltage signals into continuous mapping bands across multiple time scales, extracting the response hysteresis boundary under high-load dynamic conditions, thus providing a unified and stable time reference for the simulation input of the digital twin model. The specific implementation of this step is as follows: Based on the real-time operating signals obtained from the acquisition terminals of various electrical devices in the photovoltaic power plant, the current and voltage signals from the DC-side combiner box, inverter input terminal, inverter output terminal, and AC busbar are uniformly processed. Each signal is periodically acquired using a high-precision sampling device, with the sampling time interval typically set in the millisecond range. Due to differences in the sampling period and signal transmission path at different acquisition points, there are slight temporal offsets between the signals. To achieve time synchronization, each sampling sequence is first reordered according to its acquisition timestamp, and the time interval between adjacent data points is calculated. Time normalization is then used to map all sequences onto a unified time coordinate system. During this process, the sampling time of each measuring point is compared with the unified time signal for the entire station. If a deviation exists, time shift correction is performed based on the time deviation to ensure complete alignment of the current and voltage signals on the time axis. After normalization and time alignment, a continuous time-series signal set covering the entire photovoltaic power station's electrical circuits is obtained. This set contains voltage and current waveform data from different electrical locations, with consistent time intervals, reflecting the instantaneous electrical state of the entire power station at any given moment. This unified processing enables the subsequently established time-series hierarchical mapping to accurately describe the response relationships between signals from different locations under the same time reference.

[0022] After obtaining the complete time-series signal set, the current and voltage signals are layered and unfolded along the time dimension according to their variation characteristics. Specifically, the time scale is divided into three levels, corresponding to the fast-changing layer, the medium-changing layer, and the slow-changing layer. The fast-changing layer mainly covers data with sampling periods ranging from milliseconds to tens of milliseconds, used to reflect transient characteristics such as short-term impacts, arc discharges, and current surges; the medium-changing layer covers the range from seconds to tens of seconds, reflecting the short-term fluctuations and dynamic adjustment processes of the inverter's output power; the slow-changing layer covers the range from minutes to hours, used to reflect the steady-state response trends under the influence of changes in irradiance, temperature, and external power grid fluctuations. Through this layered unfolding method, the fluctuation patterns of the same signal at different time scales are separated, allowing each time layer to independently display the electrical behavior of a specific frequency band. In this process, a continuous time window sliding method is used to smoothly transition the signals of each level, so that the signals between different levels are seamlessly connected in time, forming a continuous mapping band from fast to slow. By unfolding the data in this multi-timescale layered manner, the electrical response changes of a photovoltaic power station under high-load dynamic conditions can be fully presented on the same time reference, so that the signals at each level remain continuously correlated on the time axis.

[0023] After forming a continuous mapping band, the correspondence between current and voltage signals at different time levels is used to identify and extract the response hysteresis boundary under high-load dynamic conditions. Specifically, in each mapping band, adjacent sampling point pairs of voltage and current are selected, their time variation trends are analyzed, and the time difference between the current signal response and voltage change is determined. For the rapidly changing layer, the transient response hysteresis range is determined by comparing millisecond-level signal changes; for the moderately changing layer, the time hysteresis of the electrical response during power regulation is obtained by analyzing the response time of the inverter output current to changes in bus voltage; for the slowly changing layer, the steady-state response hysteresis range is extracted by examining the deviation of the long-term variation trend of load power and bus voltage. The hysteresis ranges of the three levels are combined in chronological order to obtain the complete response hysteresis boundary. This hysteresis boundary reflects the temporal differences between various signals during high-load operation of the photovoltaic power plant and is an important parameter describing the inertia and dynamic adaptability of the electrical system. The formation process of the hysteresis boundary not only considers the time variation of a single signal but also comprehensively considers the energy transfer path between signals at different levels. For example, when a sudden increase in illumination causes a rise in current and a short-term drop in voltage, the hysteresis boundary can accurately reflect the time interval between the rise in current and the recovery of voltage, thus revealing the dynamic response characteristics of the system.

[0024] Based on the extracted response hysteresis boundaries, the time base of the input signals of the digital twin model is uniformly adjusted to ensure stable consistency of the simulation input under high-load dynamic conditions. Specifically, the sampling sequences are rearranged according to the time range of the hysteresis boundaries to ensure that signals at different levels correspond to the same physical state at the same point in time. When the current signal lags behind the voltage signal, the time alignment order of the input signals is adjusted to compensate for this hysteresis in the simulation input, enabling the digital twin model to receive data that matches the actual power plant operating state at the calculation time. Simultaneously, signal segments in the continuous mapping band are interpolated to ensure a complete current-voltage correspondence at any point in time. This uniform adjustment of the time base maintains the continuity, stability, and traceability of the entire simulation input space. Furthermore, during high-load dynamic changes, when the power plant output power fluctuates rapidly, this time-series hierarchical mapping model can maintain the time consistency of the simulation input in real time, avoiding input misalignment problems caused by asynchronous responses of signals at different levels. In this way, when the digital twin model performs dynamic calculations, its input signals can accurately reflect the time-series relationships of on-site operation, ensuring that the simulation output results are consistent with actual operation in the time domain. The time-series hierarchical mapping model established through this method can continuously provide input signals with time continuity and hierarchical response characteristics during the operation of the photovoltaic power plant, providing a stable time reference for digital twin simulation. This allows subsequent delay compensation and dynamic response analysis to be based on a true and reliable time sequence, thus providing precise time support for intelligent operation and maintenance of relay protection.

[0025] By collecting and organizing current and voltage signals from photovoltaic power plants, performing time normalization, hierarchical expansion, extracting lag boundaries, and uniformly adjusting the time base, a complete time-series hierarchical mapping model was established. This model enables the digital twin model to obtain input signals that are highly synchronized with actual operation under high-load dynamic conditions of photovoltaic power plants, thereby accurately reproducing the dynamic changes of the electrical system in the simulation space.

[0026] The dynamic delay compensation module constructs a dynamic delay compensation stage based on the response lag boundary results of the time-series hierarchical mapping model. It performs time window processing on the sampling sequences of current and voltage signals, correcting the signal input timing in a phase-precessing manner. This allows the digital twin model to respond to key electrical abrupt changes in advance within the calculation cycle, thus maintaining time synchronization between the simulation output and the actual operating state. The specific implementation of this step is as follows: After obtaining the response lag boundary of the time-series hierarchical mapping model, time offset analysis is performed on the sampling sequences of current and voltage signals at each measurement point of the photovoltaic power station. By comparing the timestamp of each sampling point with the reference time interval in the lag boundary, the response delay direction and delay amplitude of the current signal relative to the voltage signal under different operating states are determined. In this process, the current and voltage signals are partitioned according to the time range defined by the lag boundary, ensuring that the signal within each time period has a clear response difference range. Subsequently, a continuous signal sliding sequence is established within each time period, maintaining a fixed length of time window between adjacent sampling points and gradually sliding to capture the changing trend of electrical signals on the time axis. By establishing this sliding time window, the original sampled data can be reorganized according to the time delay characteristics without disrupting the signal continuity, thus providing a precise time positioning basis for subsequent time pre-adjustment.

[0027] After forming a sliding time window, the current and voltage signal sequences within the window are synchronized in timing. By comparing the time deviations of the current and voltage sampling points within the same time window, the direction of response lag within that window is determined. If the current signal is delayed relative to the voltage signal, the current sampling point is shifted forward on the time axis to align its time position with the voltage signal's trend; if the voltage signal lags relative to the current signal, the voltage sampling point is shifted forward on the time axis to maintain a synchronous response relationship within the window. During this process, the magnitude of each time forward shift is limited by the allowable range of the lag boundary, ensuring that the signal's time adjustment does not disrupt its true physical response characteristics. Through this phase-forward adjustment method, the change curves of the current and voltage signals within the same time window tend to synchronize, thus forming a smooth and continuous response transition in the time domain. After continuous processing through multiple sliding windows, the current and voltage signals achieve dynamic alignment on the overall time axis, gradually canceling out the lag effect throughout the entire time range.

[0028] After phase pre-adjustment, all time-corrected signals are recombined into a new continuous time-series data band, which serves as the time reference for the digital twin simulation input. To ensure the integrity of the simulation input data, signal boundaries between different sliding windows are smoothly connected, ensuring seamless temporal transitions between data segments output from each window and avoiding jumps or gaps. During this process, the signal priorities of different measurement points are dynamically prioritized based on the real-time operating status of the photovoltaic power plant. For example, the current signal at the inverter output should be prioritized during rapid power fluctuations, with a slightly higher time-prep ratio than the current signal at the combiner box, ensuring that critical electrical abrupt changes are reflected in the simulation input in advance. This hierarchical signal recombination ensures that the electrical signals from different parts exhibit a coordinated and consistent change relationship on the time axis. The signal sequence, after dynamic delay compensation, accurately reflects the real-time response of the photovoltaic power plant during the simulation input stage, enabling the digital twin to obtain the trend of key electrical state changes in advance within the simulation calculation cycle.

[0029] After signal reconstruction, the corrected continuous time-series data is input into the digital twin simulation calculation process, enabling it to respond in advance to key electrical anomalies during actual operation within the calculation cycle. Specifically, when a photovoltaic power station experiences transient changes such as a sudden voltage drop or current surge during high-load operation, the simulation input signal has been phase-forwarded through dynamic delay compensation. This allows the simulation process to predict the state before the actual fault occurs, ensuring that the simulation output is consistent with the on-site state in time. When illumination conditions suddenly change or the inverter undergoes dynamic switching, the simulation process still responds according to the pre-corrected timing sequence, ensuring that the electrical transient waveforms in the virtual space correspond to the time characteristics of the actual sampled data. Throughout the entire simulation calculation cycle, the digital twin, using the corrected signal as input, maintains a state evolution process synchronized with the field in the time dimension, thus avoiding the amplification of virtual peaks caused by delay superposition. Through this advance response mechanism, the digital twin's simulation output maintains a high degree of consistency with the on-site operation process in key electrical parameters such as current, voltage, and power changes, enabling subsequent relay protection decisions to be based on realistic and effective simulation results.

[0030] This step, by applying the response lag boundary in the time-series hierarchical mapping model, combined with the establishment of a sliding time window, adjustment of signal phase pre-positioning, reorganization of the corrected signal, and time synchronization of the simulation input, realizes the entire process of dynamic delay compensation. It can effectively offset the time misalignment caused by computational lag, communication delay, and measurement deviation under the high-load dynamic operation conditions of photovoltaic power plants, enabling the digital twin to have the ability to respond in advance during the simulation process, thereby ensuring the consistency between the simulation output and the actual operating state in time.

[0031] Virtual Peak Suppression Module: This module utilizes the synchronous output of the dynamic delay compensation stage to generate a virtual peak suppression unit. During the simulation calculation of the digital twin model, it implements nonlinear attenuation control on the transient amplitude of the current to weaken the spurious short-circuit current peak caused by the accumulation of response lag, thereby maintaining the physical continuity of the simulation output signal. The specific implementation of this step is as follows: After the dynamic delay compensation stage completes time synchronization adjustment, a sequence of current and voltage signals after phase preprocessing is obtained. At this point, the signals are continuously aligned on the time axis, accurately reflecting the transient change trend of the photovoltaic power station. Based on this synchronous output, the transient amplitude change of the current signal under high load dynamics is analyzed by partitioning the signal, dividing each continuous signal segment according to its time change rate and amplitude fluctuation range. This partitioning allows for the identification of characteristic regions of the current response under different operating conditions, such as the rapid power rise region, voltage drop region, and steady-state maintenance region. Within each characteristic region, the peak change trend of the current signal is recorded and time-series correlated with the original signal before delay compensation. This time-series comparison can identify spurious peak regions caused by accumulated response lag—those current abrupt changes that still exhibit unreasonable increases or decreases after compensation. These regions often appear during voltage recovery or at the moment of inverter load redistribution, and are the main source of false short-circuit detections. Identifying these regions provides a basis for determining the effective range and timing of subsequent nonlinear attenuation control.

[0032] After identifying the time interval where spurious peaks occur, a continuous rate of change analysis is performed on the transient amplitude of the current within this interval to determine the transition trend of the peak change. In this process, the dynamically delayed compensated current signal is used as the input signal source, and its fluctuation intensity in the time dimension is smoothed to maintain signal continuity. By observing the rate of current rise or fall in time intervals, non-physical growth caused by the accumulation of hysteresis effects can be identified, i.e., abnormal sudden increases in magnitude that occur without actual fault current support. To address this characteristic, a smooth transition channel is established in the time dimension, allowing peak changes to extend within a physically reasonable range. This smooth transition ensures that the current signal does not experience unnatural, abrupt fluctuations in a very short time during electrical transients, thus avoiding the accumulation of spurious peaks. This process relies on the synchronous timing reference provided by the dynamic delay compensation stage, enabling the signal change process to accurately correspond to the actual operating time distribution.

[0033] Based on the obtained smooth-transition current signal, nonlinear attenuation control is implemented in the region where spurious peaks are located. Specifically, within each spurious peak region, the attenuation ratio of the current amplitude during that time period is determined according to the rate and duration of peak change. By gradually reducing the peak amplitude along the time axis, the spurious peak smoothly transitions to the adjacent normal signal segment. To maintain the natural continuity of the signal, the nonlinear attenuation process should cover a certain time range before and after the peak, ensuring that the current signal remains in phase with the original waveform after the peak is eliminated. The core of this attenuation control lies in using the synchronous output after dynamic delay compensation as a reference, ensuring that the current change at each moment corresponds to the real physical response process, thereby weakening non-physical fluctuations caused by hysteresis accumulation. When a photovoltaic power station experiences rapid load transfer or grid-connected power fluctuations, this nonlinear attenuation process can automatically adjust the transient amplitude of the current signal, making its amplitude change more continuous and preventing the generation of spurious short-circuit current peaks in digital twin simulation. After this process, the amplitude distribution of the current signal is closer to the energy transfer law in actual operating conditions, maintaining the authenticity and timing integrity of the simulation input.

[0034] After adjusting for the attenuation of spurious peaks, the current signal processed by nonlinear attenuation control is recombinated with the voltage signal after delay compensation to form a new set of simulation input signals. This signal set is consistent with the synchronous output of dynamic delay compensation on the time axis, and non-physical peaks caused by hysteresis superposition are removed from the amplitude. Through this combination, the digital twin simulation process can receive physically continuous and energy-stable input signals during operation. During the simulation calculation, when the photovoltaic power station experiences power surges, voltage drops, or short-term inverter start-up and shutdown, the current changes reflected in the simulation no longer exhibit abnormal amplification, but instead show transition characteristics consistent with the real physical process. Because the peaks are smoothly attenuated, the current and voltage waveforms output by the simulation remain coordinated in time and maintain energy balance in space, making the calculation results of the digital twin continuous and stable. In this way, no virtual short-circuit current peaks are generated during the simulation, and the conditions for malfunction of relay protection are not triggered. At the same time, the current waveform output by the simulation maintains similar timing characteristics to the field measurement signal, making the output signal of the digital twin model consistent with the actual operating state in the time dimension, thereby achieving the maintenance of physical continuity and the realistic reproduction of dynamic response.

[0035] This process, through the synchronous output of the dynamic delay compensation stage, identifies, smooths, nonlinearly attenuates, and reassembles the transient amplitude changes of the current signal, realizing the entire process of virtual peak suppression. It can effectively weaken the false short-circuit current peak caused by the accumulation of response lag, maintain the continuity of the digital twin simulation output signal in time and amplitude, and make the electrical simulation process of photovoltaic power plants under high load dynamic conditions more in line with the actual physical response law.

[0036] The partitioned energy balance module constructs a partitioned energy balance loop based on the output of the virtual peak suppression unit. By performing time-weighted correction and dynamic allocation of the power change trends of each branch of the photovoltaic power station, the energy distribution in the simulation space is maintained in a stable dynamic balance to prevent abnormal protection conditions triggered by false peaks. The specific implementation of this step is as follows: After the virtual peak suppression unit outputs current and voltage signals after nonlinear attenuation control, the operating status of each branch of the photovoltaic power station is divided into energy characteristic regions. Using the suppressed signals as the input basis, the entire electrical network is divided into multiple energy distribution regions according to the electrical connection relationships of each busbar, inverter input side, inverter output side, and AC busbar in the photovoltaic power station. Each region contains several parallel branches, and their energy exchange relationship is reflected by the instantaneous changes in voltage and current signals. By continuously recording the power change rate of each branch, the time-series distribution characteristics of power output in different regions can be obtained. Since the virtual peak suppression unit has eliminated false current surges caused by response lag, the power change trend at this time can more realistically reflect the actual energy transfer process. In this way, the energy transfer network inside the photovoltaic power station is divided into zones in the simulation space, so that each energy zone has independent power change characteristics and time response laws, thereby providing a clear range of action for subsequent dynamic balance adjustment.

[0037] After dividing the energy regions, time-weighted correction is applied to the power change trends of each branch. Continuous analysis of the synchronous output of current and voltage signals in the time dimension determines the power change rate of each branch within different time periods. When the power increase rate of a branch is higher than that of other branches, it indicates that the energy input of that branch in the simulation space is relatively concentrated. Without time-weighted correction, this would lead to uneven energy distribution. Therefore, the power change curve of each branch is weighted according to a time window, giving lower time weights to branches with faster power changes within the same time period, and higher time weights to branches with slower changes, thus achieving time balance in the overall energy distribution. This time-weighted correction process relies on the stable output provided by the virtual peak suppression unit, ensuring continuous signal continuity and avoiding time offset caused by lag effects. After time-weighted correction, the power changes of each branch tend to be coordinated in time, and the energy transfer process exhibits mutual balance among multiple branches, laying the foundation for the next step of dynamic allocation.

[0038] After time-weighted correction, the corrected power change trend is dynamically allocated to maintain overall energy balance within the simulation space. Specifically, the power output of each branch of the photovoltaic power station is proportionally adjusted according to its change amplitude after time-weighted correction, ensuring that the power distribution of different branches meets the energy conservation condition. In this process, branches at the edge of power fluctuations are prioritized, gradually transferring their excess energy to branches with decreasing power, achieving dynamic compensation between regions. When a branch experiences a sudden power increase due to enhanced sunlight or localized temperature rise, the dynamic allocation process distributes this increased energy to adjacent branches in chronological order, ensuring a continuous transition in overall energy changes and preventing localized power overload. Simultaneously, when a branch experiences a brief power decrease due to inverter switching or external grid fluctuations, the power of other branches is compensated through time-weighted adjustment, smoothing the overall power curve. This dynamic allocation method ensures a continuous and coordinated energy flow process within the simulation space, eliminating abnormal protection conditions that might be triggered by sudden energy changes in a single branch.

[0039] After completing the dynamic energy allocation, the adjusted power change results of each branch are recombined into an overall energy distribution curve in the simulation space. This curve reflects the energy transfer pattern of the entire photovoltaic power station under high load dynamic operation. The power fluctuations in each time period are time-weighted and dynamically allocated, reflecting the real energy response process. During simulation operation, when the power change of a branch is too rapid or the energy fluctuation is too large, the partitioned energy balance loop will automatically adjust the energy of that branch according to the aforementioned dynamic allocation results, so that its power change curve returns to the balance range, thereby preventing false peaks from interfering with the relay protection logic. In this way, the energy distribution in the simulation space is always maintained in a dynamic balance state, and the energy transfer between branches remains stable and continuous even under high load fluctuation conditions. At the same time, since the energy distribution has been dynamically balanced, local energy accumulation or virtual overload phenomena no longer occur during the simulation, enabling the simulation output of the digital twin to accurately reflect the real operating characteristics of the photovoltaic power station. Finally, the overall power distribution in the simulation space is consistent with the actual operating state of the power station, and the relay protection logic will only respond to real power changes when determining the electrical state, thereby effectively preventing malfunctions caused by false peaks.

[0040] This step, using the output of the virtual peak suppression unit, sequentially completes energy characteristic partitioning, time-weighted correction, dynamic energy allocation, and overall energy reorganization, constructing a partitioned energy balance loop capable of adaptively maintaining energy stability. Under the high-load dynamic operating conditions of a photovoltaic power plant, it can continuously adjust the energy distribution of each branch, maintaining dynamic balance in the power transfer process within the simulation space. This fundamentally eliminates the risk of energy imbalance caused by false peaks, providing a continuous and reliable energy foundation for the stable operation of relay protection logic.

[0041] Adaptive Threshold Control Module: Based on the dynamic output of the partitioned energy balance loop, an adaptive action threshold control loop is established. This loop automatically adjusts the relay protection action threshold according to the energy gradient changes of each branch, ensuring that the relay protection logic only responds to real short-circuit events. This achieves suppression of malfunctions and intelligent operation and maintenance management of the photovoltaic power plant's relay protection system. The specific implementation of this step is as follows: After the partitioned energy balance loop completes the time-weighted correction and dynamic energy allocation of the power changes in each branch of the photovoltaic power plant, the balanced energy output result is obtained. This output result reflects the power change trend and energy gradient distribution of each branch within the current operating cycle, where the energy gradient represents the rate of power change per unit time. By continuously tracking these balanced energy data, the power change trajectory of each branch under high-load dynamic conditions can be clearly depicted. To achieve dynamic adaptive control of the relay protection action threshold, it is first necessary to establish a correspondence between the power change data output by the energy balance loop and the relay protection action parameters. In this process, the real-time power change rate of each branch is matched with its corresponding current and voltage signals to form the branch energy gradient curve. This curve not only reflects the temporal trend of branch energy change, but also demonstrates the branch's response sensitivity to electrical disturbances. When the energy gradient of a branch suddenly increases without triggering a real short-circuit event, this non-fault fluctuation can be identified through the continuous trend of the energy gradient curve, providing a reference for subsequent threshold adjustments.

[0042] After obtaining the energy gradient curves of each branch, a correlation analysis is performed on the energy gradients between different branches to determine the initial benchmark for threshold adjustment. By comparing the time-weighted power data output by the zoned energy balance loop, the relative energy change amplitude of each branch within the same time period can be obtained. When the energy gradient of some branches continues to rise while the energy change of other branches remains stable, it indicates that the area may be affected by load concentration or local grid fluctuations. To avoid triggering relay protection actions by such non-fault factors, the protection action threshold of that branch needs to be adjusted accordingly. Conversely, when the energy gradient of a branch rises sharply in a very short time, accompanied by a momentary drop in voltage, it indicates that there may be a real short circuit. In this case, the action threshold of that branch should be lowered to make the protection action more sensitive. During this process, the magnitude and direction of the threshold adjustment are updated in real time based on the dynamic energy distribution data output by the energy balance loop. Through this energy gradient-driven adjustment method, the threshold change process can continuously follow the dynamic changes in energy distribution, thereby maintaining consistency with the actual operating state.

[0043] After the initial threshold adjustment, a dynamic smoothing transition is performed on the results to ensure the continuity and stability of the adaptive action threshold over time. Specifically, the threshold change curves of each branch in adjacent time periods are connected to eliminate abrupt changes in threshold values ​​at different time points. This smooth transition prevents relay protection logic from jumping its action when threshold changes occur. This process relies on the dynamic output of the zoned energy balance loop to ensure that the boundary conditions for time smoothing are consistent with the changing trend of energy distribution. When the photovoltaic power station is operating under high load, the threshold smoothing transition time span is relatively shortened to quickly respond to power changes; while under relatively stable load conditions, the smoothing time span is relatively extended to maintain system stability. Through this time smoothing control, threshold changes can both quickly respond to dynamic operating conditions and maintain a continuous and stable trend overall. After this process, the relay protection action threshold is no longer a fixed value, but a dynamic parameter that is continuously adjusted over time according to changes in energy distribution, thus making the protection action more consistent with the actual operating conditions.

[0044] After completing the dynamic smooth adjustment of the adaptive action threshold, the updated threshold parameters are input into the relay protection logic to achieve adaptive response in action determination. When a sudden electrical disturbance occurs in the photovoltaic power station, the relay protection logic first determines the nature of the disturbance based on the current energy gradient. If the energy change is a slow transition and does not exceed the adjusted threshold, it is determined to be a normal fluctuation, and the protection action remains stationary. If the energy gradient rises sharply and exceeds the threshold value, the corresponding protection action is triggered, achieving a rapid response to a real short-circuit event. Throughout the operation, the threshold control loop continuously receives dynamic output from the zoned energy balance loop and continuously fine-tunes the threshold value according to the real-time changes in the energy gradient. When the photovoltaic power station load gradually returns to stability, the threshold value automatically returns to the equilibrium state, allowing the protection logic to maintain its normal sensitivity again. Through this adaptive dynamic adjustment mechanism, the relay protection logic can automatically adjust its sensitivity and action delay under different load conditions, ensuring safety while avoiding malfunctions. When non-fault energy fluctuations occur, the adaptive upward adjustment of the threshold can effectively prevent false signals from triggering circuit breakers; when a real short-circuit event occurs, the rapid downward adjustment of the threshold can ensure the timeliness of protection actions, thus achieving a dynamic balance between the two.

[0045] Based on the above steps, the energy gradient extraction, threshold benchmark determination, dynamic smooth transition, and relay protection logic input are sequentially completed through the dynamic output of the partitioned energy balance loop. This achieves the construction of an adaptive action threshold control loop, which can automatically adjust the relay protection action threshold according to the real-time changes in the energy gradient of each branch of the photovoltaic power plant. This ensures that the relay protection logic only responds to real short-circuit events, thereby effectively suppressing malfunctions and achieving intelligent operation and maintenance management under complex dynamic conditions. In this way, the photovoltaic power plant relay protection system possesses adaptive adjustment capabilities synchronized with the operating status, and can continuously maintain stable and accurate operating characteristics in high-load dynamic environments.

[0046] This invention introduces a synergistic mechanism of time-series hierarchical mapping, dynamic delay compensation, and virtual peak suppression during digital twin simulation. This ensures that the electrical signals of a photovoltaic power plant remain continuous and consistent in the time dimension under high-load dynamic operating conditions, effectively avoiding false current peaks caused by the superposition of simulation delay and sampling lag. Through this mechanism, the digital twin model can respond to key electrical abrupt changes in advance within the calculation cycle, achieving a high degree of synchronization between the simulation output and the actual operating state. This makes the input basis for relay protection judgments more realistic and reliable, significantly improving the accuracy and stability of the protection system in identifying real short-circuit events.

[0047] This invention constructs a partitioned energy balance loop and an adaptive action threshold control loop to maintain a dynamic balance in the power distribution of a photovoltaic power plant operating in a multi-branch environment, ensuring continuous stability of the energy transfer process in the simulation space. The relay protection action threshold can be automatically adjusted according to changes in the energy gradient of each branch, ensuring that the protection logic responds only to real fault conditions, fundamentally eliminating malfunction chains caused by false signals. This scheme achieves intelligent dynamic management of photovoltaic power plant relay protection, improving the safety of system operation and the reliability of maintenance.

[0048] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A photovoltaic power plant relay protection intelligent operation and maintenance system based on digital twins, characterized in that, It includes a time-series hierarchical mapping module, a dynamic delay compensation module, a virtual peak suppression module, a partitioned energy balance module, and an adaptive threshold control module; The time-series hierarchical mapping module establishes a time-series hierarchical mapping model based on the real-time operation signals of the photovoltaic power station. It unfolds the collected current and voltage signals into continuous mapping bands at multiple time scales and extracts the response lag boundary under high load dynamic conditions. The dynamic delay compensation module constructs a dynamic delay compensation link based on the response lag boundary results of the time-series hierarchical mapping model. It performs sliding time window processing on the sampling sequence of current and voltage signals and corrects the signal input timing in a phase-precessing manner, so that the digital twin model can respond to key electrical mutations in advance within the calculation cycle. The virtual peak suppression module generates a virtual peak suppression unit using the synchronous output of the dynamic delay compensation link. During the simulation calculation of the digital twin model, it implements nonlinear attenuation control on the transient amplitude of the current, weakening the false short-circuit current peak caused by the accumulation of response lag. Partitioned Energy Balance Module: Based on the output of the virtual peak suppression unit, a partitioned energy balance loop is constructed. By performing time-weighted correction and dynamic allocation on the power change trend of each branch of the photovoltaic power station, the energy distribution in the simulation space is kept in a stable dynamic balance. Adaptive Threshold Control Module: Based on the dynamic output of the partitioned energy balance loop, an adaptive action threshold control loop is established. The relay protection action threshold is automatically adjusted according to the energy gradient changes of each branch, so that the relay protection logic only responds to real short-circuit events.

2. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 1, characterized in that, The steps for establishing a time-series hierarchical mapping model based on real-time operation signals of a photovoltaic power plant include: First, based on the current and voltage signals obtained from the acquisition terminals of various electrical equipment in the photovoltaic power plant, the sampled data from the DC-side combiner box, inverter input terminal, inverter output terminal, and AC bus are uniformly organized and aligned on the time axis through time normalization and time synchronization comparison correction. Second, based on the time-aligned signals, the current and voltage signals are divided into fast-changing, medium-changing, and slow-changing layers according to their variation characteristics in the time dimension. A continuous transition between layers is achieved through time window sliding, forming a multi-time-scale continuous mapping band. Third, based on the continuous mapping band, the time difference between the current and voltage signals in each time layer is analyzed to extract the response lag boundary under high-load dynamic conditions, and the lag range of the electrical response is determined according to the variation law of different layers. Fourth, based on the response lag boundary, the time base of the input signal of the digital twin model is adjusted, the sampling sequence is rearranged, and missing data is filled in on the time axis.

3. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 2, characterized in that, During the time base adjustment process, the sampling sequences of the fast-changing layer, medium-changing layer and slow-changing layer are rearranged in segments according to the time range of the response lag boundary, and interpolation is performed between adjacent time layers to ensure that the current signal and the voltage signal maintain a continuous correspondence at multiple time scales.

4. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 2, characterized in that, The steps for constructing a dynamic delay compensation stage based on the response lag boundary results of the time-series hierarchical mapping model include: after obtaining the response lag boundary of the time-series hierarchical mapping model, performing time offset analysis on the sampling sequences of current and voltage signals at each measuring point of the photovoltaic power station, and establishing a continuous sliding time window according to the time range of the lag boundary to capture the changing trend of the signal on the time axis; within the sliding time window, comparing the time deviations of the current and voltage signals, and sliding the corresponding signals forward on the time axis according to the lag direction, so that their change curves maintain a synchronous response relationship within the window range and form a smooth and continuous transition in the time domain; after completing the phase pre-adjustment, recombining the time-corrected signals into a new continuous time-series data band, smoothly connecting the boundaries of adjacent sliding windows, and dynamically prioritizing the signals of different measuring points according to the real-time operating status of the photovoltaic power station, so that key electrical mutations are reflected in the simulation input first; inputting the corrected continuous time-series data band into the digital twin simulation calculation process, so that it responds in advance to key electrical mutations in actual operation within the calculation cycle, maintaining the synchronization between the simulation output and the on-site operating status in time.

5. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 4, characterized in that, The length of the sliding time window and the phase advance amplitude are dynamically adjusted according to the response hysteresis boundary of each level in the time-series hierarchical mapping model, so that the time sliding amplitude of the rapidly changing layer is smaller than that of the medium-changing layer and the slowly changing layer.

6. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 4, characterized in that, The steps for generating a virtual peak suppression unit using the synchronous output of the dynamic delay compensation stage include: after completing time synchronization adjustment in the dynamic delay compensation stage, acquiring current and voltage signal sequences after phase pre-processing; performing zone analysis on the transient amplitude change of the current signal under high load dynamics to identify the rapid power rise region, voltage drop region, and steady-state maintenance region, and determining the time interval where the false peak occurs; within the false peak time interval, performing continuous rate of change analysis on the transient amplitude change of the current signal, establishing a smooth transition channel based on the synchronization timing reference provided by the dynamic delay compensation stage, so that the peak change extends within a physically reasonable range and maintains signal continuity; based on the current signal after smooth transition, implementing nonlinear attenuation control in the false peak region, gradually reducing the current amplitude according to the peak change rate and duration, so that the false peak smoothly transitions to the adjacent normal signal segment and maintains phase consistency with the original waveform; after completing the false peak attenuation adjustment, recombining the current signal processed by nonlinear attenuation control and the voltage signal after delay compensation to form a new simulation input signal set, so that the digital twin simulation process receives an input signal with stable energy and continuous time, weakening the false short-circuit current peak caused by hysteresis accumulation.

7. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 6, characterized in that, In the nonlinear decay control process, a transition zone is set on the time axis for the current signal in the false peak region. The time length of the transition zone is adaptively adjusted according to the peak change rate, so that the current amplitude gradually and smoothly transitions during the decay process.

8. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 6, characterized in that, The steps for constructing a partitioned energy balance loop based on the output of the virtual peak suppression unit include: after the virtual peak suppression unit outputs current and voltage signals after nonlinear attenuation control, the operating status of each branch of the photovoltaic power station is divided into energy characteristics. Based on the electrical connection relationship of the combiner branch, inverter input side, inverter output side, and AC bus, the electrical network is divided into multiple energy distribution regions, and the time-series distribution characteristics of the power change rate of each branch are recorded. After completing the energy region division, the power change trend of each branch is corrected by time weighting, so that branches with fast power changes receive low time weights and branches with slow changes receive high time weights. After completing the time weighting correction, the corrected power change trend is dynamically allocated, and the power output of each branch of the photovoltaic power station is adjusted proportionally according to the change amplitude to achieve dynamic compensation of energy between different branches, so that the overall energy transfer remains continuously balanced. After completing the dynamic energy allocation, the adjusted power change results of each branch are recombined into the overall energy distribution curve of the simulation space, so that the energy distribution in the simulation space remains dynamically balanced.

9. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 8, characterized in that, During the time-weighted correction process, a fixed time window is used to continuously weight the power change curves of each branch, so that the time weight of branches with high power change rates decreases proportionally, while the time weight of branches with gentle changes increases proportionally. In the dynamic allocation stage, energy compensation is given priority to branches with power fluctuation edges based on the weighting results.

10. The intelligent operation and maintenance system for photovoltaic power station relay protection based on digital twin as described in claim 8, characterized in that, The steps for establishing an adaptive action threshold control loop based on the dynamic output of the partitioned energy balance loop include: after the partitioned energy balance loop completes the time-weighted correction and dynamic energy allocation of the power changes of each branch of the photovoltaic power station, the energy output result after balance processing is obtained, and the power change rate is matched with the current signal and voltage signal to form the branch energy gradient curve; after obtaining the energy gradient curve of each branch, the energy gradient of different branches is correlated and analyzed, and the threshold adjustment benchmark is determined based on the time-weighted power data. When the branch energy gradient continues to rise, the action threshold is raised; when the energy gradient rises sharply and is accompanied by a voltage drop, the action threshold is lowered, so that the threshold adjustment process is consistent with the energy distribution change; after completing the initial threshold adjustment, the threshold change result is dynamically smoothed, the threshold curves of each branch in adjacent time periods are connected in time, and the time smoothing span is automatically adjusted according to the operating status of the photovoltaic power station; after completing the dynamic smoothing adjustment of the adaptive action threshold, the updated threshold parameters are input into the relay protection logic, and the protection sensitivity and action delay are automatically adjusted according to the real-time change of the energy gradient, so that the relay protection logic only responds to real short-circuit events.

Images