Multi-energy cooperative control method and system with self-adaptive disturbance discrimination function

Abstract

The invention discloses a multi-energy cooperative control method and system with a self-adaptive disturbance discrimination function, and relates to the technical field of energy cooperative control, and the method comprises the following steps: constructing a multi-band phase fingerprint sequence based on the instantaneous voltage response of a photovoltaic array, and forming an initial energy transition baseline through extracting microsecond phase deviation tracks in different bands; and based on the initial energy transition baseline, constructing time series dynamic gradient mapping, converting the phase deviation trajectory into continuous power gradient distribution, and identifying a potential acceleration fragment in the power gradient distribution. According to the method, a high-sensitivity energy identification mechanism is constructed, pre-judgment is carried out in advance before photovoltaic output sharply rises, and power of different frequency bands is guided to enter a layered energy storage path based on a high-risk window, so that energy deposition according to needs is realized, local thermal stress concentration is avoided, and structural damage and energy storage attenuation of a heat storage body are effectively prevented; the stability and safety of electric heating are improved, and the service life of heat storage materials is prolonged.

Description

Technical Field

[0001] This invention relates to the field of energy collaborative control technology, specifically to a multi-energy collaborative control method and system with adaptive disturbance discrimination function. Background Technology

[0002] Photovoltaic direct-drive solid-state thermal storage and electric heating multi-energy coordinated control is a technical solution that integrates solar energy, thermal storage devices, and the power grid for management and optimized scheduling. Its core objective is to improve energy utilization and operational flexibility. This method utilizes photovoltaic power generation to directly drive solid-state thermal storage devices, enabling flexible adjustment of thermal storage and electric heating based on real-time fluctuations in photovoltaic output. Excess electricity can be converted into thermal energy for storage and released on demand in low-temperature environments, reducing dependence on traditional electricity. During operation, dynamic adjustments can be made based on weather conditions, load demand, and electricity price changes to maximize energy utilization. Through intelligent management strategies, the proportion and scheduling methods of different energy sources are adjusted in real time, which not only alleviates grid pressure and enhances heating self-sufficiency but also effectively reduces carbon emissions and operating costs, promoting efficient, safe, and sustainable energy utilization.

[0003] The existing technology has the following shortcomings:

[0004] In existing technologies, when photovoltaic output increases rapidly within a short period, the solid thermal storage medium is prone to triggering abnormal phase transition zones due to the excessively rapid temperature rise, leading to a sharp accumulation of localized thermal stress. This stress, after diffusing within the solid medium, often causes the material's microstructure to disintegrate and thermal cracks to propagate continuously, resulting in irreversible energy storage degradation. Furthermore, when cracks evolve into penetrating defects, the effective heat capacity of the thermal storage medium will significantly decrease, and in severe cases, it may even directly lose its energy storage function, causing the energy regulation process to fail and compromising the stability and continuity of the electric heating process.

[0005] 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

[0006] The purpose of this invention is to provide a multi-energy cooperative control method and system with adaptive disturbance discrimination function to solve the problems in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-energy cooperative control method with adaptive disturbance discrimination function, comprising the following steps:

[0008] A multi-band phase fingerprint sequence is constructed based on the instantaneous voltage response of the photovoltaic array. By extracting the microsecond-level phase shift trajectory in different frequency bands, an initial energy transition baseline is formed to capture potential triggering signals of rapid changes in photovoltaic power.

[0009] Based on the initial energy transition baseline, a time-series dynamic gradient mapping is constructed to transform the phase shift trajectory into a continuous power gradient distribution, and potential acceleration segments with nonlinear steep rise characteristics are identified in the power gradient distribution.

[0010] A cross-scale time backtracking curve is established within the potential acceleration segment, and key change features are enhanced by differential amplification. The key change features are coupled with the phase fingerprint sequence to generate an enhanced energy mutation curve.

[0011] Based on the enhanced energy mutation curve, a dynamic critical threshold index is set. Highly sensitive time segments that continuously cross the dynamic critical threshold index are screened in the enhanced energy mutation curve, and these highly sensitive time segments are marked as high-risk windows for a sharp increase in photovoltaic output.

[0012] An adaptive disturbance identification channel is constructed based on a high-risk window, which integrates the characteristics of the energy transition baseline, power gradient distribution and enhanced energy mutation curve to complete the identification and early warning output of the photovoltaic power output in real time.

[0013] When it is determined that the photovoltaic output is entering a state of rapid increase, the output spectrum of the photovoltaic array is selectively coupled, and the power of different wavelength bands is introduced into different energy storage paths (such as directing short-wavelength power to the direct heating path and directing long-wavelength power to the slow-release heating path) to achieve energy stratification and reduce the rapid thermal stress accumulation in a single area of ​​the solid thermal storage body.

[0014] Preferably, the steps for generating the initial energy transition baseline include:

[0015] The DC output voltage signal of the photovoltaic array is sampled with high precision and then divided into multiple logarithmic frequency ranges after removing noise below 10 Hz and above 500 kHz by bandpass filtering.

[0016] The voltage signals in each frequency range are subjected to Fast Fourier Transform and Hilbert Transform to extract the instantaneous phase angle, and the phase trajectory for at least 10 seconds is stored in a circular buffer.

[0017] The phase trajectories are subjected to noise suppression and normalization, and the phase trajectories of different frequency bands are mapped to the 0 to 1 interval. A phase combination vector is formed under the same timestamp to construct a multi-frequency band phase fingerprint sequence.

[0018] Based on the multi-band phase fingerprint sequence, key feature points with phase change amplitude greater than 5 degrees and duration exceeding 20 microseconds are identified, and the least squares method is used to fit and generate an initial energy transition baseline, which serves as a criterion and reference standard for rapid changes in photovoltaic power.

[0019] Preferably, when generating the initial energy transition baseline by least squares fitting, fitting is only performed when three or more key feature points with a phase change amplitude greater than 5 degrees and a duration of more than 20 microseconds appear consecutively, so as to improve the accuracy of the energy transition baseline in distinguishing rapid changes in photovoltaic power.

[0020] Preferably, the step of identifying potential acceleration segments includes:

[0021] After obtaining the initial energy transition baseline, the phase fingerprint sequence is time-series processed to align each phase sampling point with the initial energy transition baseline and keep it consistent with the corresponding timestamp.

[0022] After completing the time serialization process, the phase shift trajectory is subjected to first-order difference operation to obtain the instantaneous phase change rate, and the phase change rate is smoothed by weighted averaging.

[0023] After obtaining the phase change rate, the phase change rate is superimposed on the initial energy transition baseline point by point, and then converted into a power gradient through integral normalization to form a continuous power gradient distribution curve.

[0024] After forming a continuous power gradient distribution curve, a second-order difference operation is performed on the distribution curve. When the power velocity exceeds five consecutive sampling points and is greater than three times the standard deviation of the power change rate, the segment is marked as a potential acceleration segment. The potential acceleration segment is then compared and confirmed point by point with the initial energy transition baseline and phase shift trajectory.

[0025] Preferably, the confirmation of potential acceleration segments includes: when the power velocity exceeds five consecutive sampling points and the corresponding phase change amplitude is greater than 5 degrees, the segment is determined to be an effective acceleration segment of the photovoltaic power rapid transition trend.

[0026] Preferably, the steps for generating the enhanced energy mutation curve include:

[0027] After identifying the effective acceleration segment, the effective acceleration segment is used as the core interval. The time series is traced back by no less than five times the segment length and extended backward by no less than two times the segment length. The phase shift trajectory and power gradient distribution are resampled at log intervals of microsecond, millisecond and second within the core interval to form a cross-scale time backtracking curve.

[0028] On the cross-scale time backtracking curve, second-order and third-order difference operations are performed on the phase offset trajectory and power gradient distribution, respectively. The composite enhanced signal is obtained by amplitude normalization and energy weighting. At the same time, wavelet threshold denoising method is used to filter out noise.

[0029] After obtaining the composite enhanced signal, it is coupled point by point with the multi-band phase fingerprint sequence, and the phase fingerprint vector is weighted by the key change amplitude after differential amplification, and finally arranged in time order to form an enhanced energy mutation curve.

[0030] Preferably, the steps of setting a dynamic critical threshold index and calibrating a high-risk window based on the enhanced energy mutation curve include:

[0031] After obtaining the enhanced energy mutation curve, a sliding window statistical analysis was performed on the curve to calculate the mean and standard deviation, and the dynamic critical threshold index was set to the mean plus twice the standard deviation.

[0032] After establishing the dynamic critical threshold index, the threshold index is compared point by point with the enhanced energy mutation curve. If the amplitude of three or more consecutive sampling points is greater than the corresponding threshold, the interval of consecutive sampling points is marked as an effective overthreshold segment.

[0033] After obtaining the effective overthreshold segment, all consecutive overthreshold points are time-aggregated and extended by 20% of the length of the preceding and following intervals to form a highly sensitive time segment, and pseudo segments with insufficient time length are removed.

[0034] After obtaining the set of highly sensitive time segments, their start point, end point, maximum amplitude point, and duration are used as risk indicators and compared with the phase fingerprint sequence. If the two correspond, the highly sensitive time segment is marked as a high-risk window for a sharp increase in photovoltaic power output.

[0035] Preferably, the steps for constructing an adaptive disturbance identification channel based on a high-risk window include:

[0036] After obtaining the high-risk window, the high-risk window is used as the starting point of the discrimination interval. Within this interval, the phase shift trajectory from the energy transition baseline, the power gradient distribution from the time series dynamic gradient mapping, and the mutation features from the enhanced energy mutation curve are called respectively, and aligned by a unified timestamp.

[0037] After feature alignment is completed, the energy transition baseline offset amplitude, the instantaneous slope of the power gradient distribution, and the amplitude of the enhanced energy mutation curve are used as the basic weights, respectively. After normalization, they are linearly weighted and combined to generate an adaptive perturbation feature curve.

[0038] After obtaining the adaptive disturbance characteristic curve, it is detected whether the curve continuously exceeds the threshold within the high-risk window. The threshold is the average amplitude of the steady phase plus three times the standard deviation. If the duration of the adaptive disturbance characteristic curve continuously exceeding the threshold exceeds one-third of the duration of the high-risk window, it is determined to be a state of rapid increase in photovoltaic power output, and an early warning signal including amplitude index, duration index and phase offset index is output.

[0039] Preferably, the step of selectively coupling the output spectrum of the photovoltaic array and achieving energy stratification deposition when it is determined that the photovoltaic output has entered a state of rapid increase includes:

[0040] After determining that the photovoltaic output has entered a state of rapid increase, the frequency domain decomposition of the DC output signal of the photovoltaic array is performed, the output spectrum is obtained by fast Fourier transform, and the spectrum is divided into short-wavelength band corresponding to transient power and long-wavelength band corresponding to continuous power.

[0041] After obtaining the band division of the photovoltaic output spectrum, the short-wavelength power is directed to the direct heating path to quickly consume instantaneous energy on the surface, and the long-wavelength power is directed to the slow-release heating path to gradually inject into the deep layer of the thermal storage body.

[0042] After completing the power stratification guidance, dynamic current limiting control is set for the direct heating path and progressive energy release control is set for the slow-release heating path, and the temperature difference between the surface and the deep layer is monitored in real time to keep it not exceeding 20 degrees Celsius.

[0043] During the energy stratification deposition process, the energy flow of the two energy storage paths is continuously corrected in a closed loop. When the surface temperature rises too quickly, the proportion of short-wavelength power is reduced, and when the temperature rise in the deep layer lags, the proportion of long-wavelength power is increased, thereby achieving adaptive stratification control.

[0044] The multi-energy collaborative control system with adaptive disturbance discrimination function includes a phase fingerprint construction module, a dynamic gradient mapping module, a feature enhancement coupling module, a dynamic threshold recognition module, an adaptive disturbance discrimination module, and an energy stratification deposition module;

[0045] The phase fingerprint construction module constructs a multi-band phase fingerprint sequence based on the instantaneous voltage response of the photovoltaic array. By extracting the microsecond-level phase shift trajectory in different frequency bands, an initial energy transition baseline is formed.

[0046] The dynamic gradient mapping module constructs a time-series dynamic gradient mapping based on the initial energy transition baseline, transforms the phase shift trajectory into a continuous power gradient distribution, and identifies potential acceleration segments in the power gradient distribution.

[0047] The feature enhancement coupling module establishes a cross-scale time backtracking curve within the potential acceleration segment and enhances key change features using differential amplification. It then couples the key change features with the phase fingerprint sequence to generate an enhanced energy mutation curve.

[0048] The dynamic threshold identification module sets a dynamic critical threshold index based on the enhanced energy mutation curve, filters out highly sensitive time segments that continuously cross the dynamic critical threshold index in the enhanced energy mutation curve, and marks the highly sensitive time segments as high-risk windows for a sharp increase in photovoltaic output.

[0049] The adaptive disturbance discrimination module constructs an adaptive disturbance identification channel based on a high-risk window, and integrates the features of the energy transition baseline, power gradient distribution and enhanced energy mutation curve to complete the discrimination and early warning output of the photovoltaic power output rising sharply.

[0050] The energy stratification deposition module selectively couples the output spectrum of the photovoltaic array when it detects that the photovoltaic output is entering a state of rapid increase, and introduces the power of different bands into different energy storage paths to achieve energy stratification deposition.

[0051] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0052] This invention constructs a highly sensitive energy change identification mechanism that combines multi-band phase fingerprint sequences with dynamic gradient mapping. Combined with cross-scale backtracking, differential enhancement, and dynamic threshold screening, it can identify the rapid upward trend of photovoltaic power output before it triggers an actual temperature rise. Using high-risk windows as trigger conditions, it accurately guides power in different frequency bands into the corresponding energy storage paths, achieving on-demand layered energy deposition. This effectively avoids the local thermal stress accumulation caused by the instantaneous concentrated injection of energy into a single area, reduces the risk of microstructural damage, crack propagation, and energy storage attenuation caused by rapid heating of solid thermal storage materials, and improves the continuity, safety, and service life of electric heating operation and thermal storage materials. Attached Figure Description

[0053] 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.

[0054] Figure 1 This is a flowchart of the multi-energy cooperative control method with adaptive disturbance discrimination function of the present invention.

[0055] Figure 2 This is a schematic diagram of the multi-energy cooperative control system with adaptive disturbance discrimination function of the present invention.

[0056] Figure 3 This is a schematic diagram illustrating the steps for generating the initial energy transition baseline of this invention.

[0057] Figure 4 This is a schematic diagram illustrating the steps involved in generating the enhanced energy mutation curve of this invention. Detailed Implementation

[0058] 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.

[0059] like Figures 1 to 4 As shown:

[0060] This invention provides a multi-energy cooperative control method with adaptive disturbance discrimination function, comprising the following steps:

[0061] A multi-band phase fingerprint sequence is constructed based on the instantaneous voltage response of the photovoltaic array. By extracting the microsecond-level phase shift trajectory in different frequency bands, an initial energy transition baseline is formed to capture potential triggering signals of rapid changes in photovoltaic power.

[0062] To capture potential triggering signals from rapid changes in photovoltaic power, a technique based on constructing a multi-band phase fingerprint sequence using the instantaneous voltage response of the photovoltaic array is adopted. The specific process is as follows:

[0063] High-precision sampling of the instantaneous voltage signal of the photovoltaic array is performed. Specifically, a high-bandwidth voltage acquisition line is arranged at the DC output end of the photovoltaic array. This line consists of a high-speed analog-to-digital converter and a signal conditioning circuit, which can acquire the voltage signal in real time at a sampling frequency of no less than 1 MHz without distortion. The time resolution of signal acquisition is strictly controlled at the order of 1 microsecond to ensure the capture of transient voltage changes caused by sudden changes in solar irradiance or rapid cloud movement. The acquired raw voltage signal is first processed by a bandpass filter circuit to filter out DC drift components below 10 Hz and electromagnetic noise above 500 kHz, thereby retaining the effective voltage waveform related to the dynamic characteristics of photovoltaic output. Subsequently, the filtered voltage waveform is divided into multiple frequency intervals, with the bandwidth of each interval strictly controlled within a logarithmic distribution range, such as 10 Hz to 100 Hz, 100 Hz to 1 kHz, 1 kHz to 10 kHz, and 10 kHz to 100 kHz, to ensure coverage of the dynamic response characteristics of photovoltaic voltage across the entire frequency domain.

[0064] After dividing the voltage signal into frequency bands, phase extraction and trajectory tracking are performed on the voltage signals within each band. Specifically, each frequency band signal is input into a Fast Fourier Transform (FFT) operation to obtain the complex representation of the signal in the frequency domain, and the instantaneous phase angle is further calculated using a Hilbert Transform. To ensure the accuracy of the calculation, the phase angle resolution is set to 0.01 degrees, and it is updated in real time at each sampling point. In this way, the phase change trajectory of each frequency band over time on a microsecond-scale can be obtained. Since the voltage response of a photovoltaic array under strong light often manifests as a phase shift within a short period of time, tracking these trajectories can reveal abnormal signals before the power amplitude changes significantly. The phase trajectory is recorded using a circular buffer to ensure that at least 10 seconds of historical phase data can be saved during continuous operation, allowing for complete backtracking during the subsequent establishment of the energy transition baseline.

[0065] After obtaining the phase trajectories of multiple frequency bands, these trajectories are normalized and mapped to construct a unified multi-frequency band phase fingerprint sequence. Specifically, noise suppression is first applied to each phase trajectory, using a moving average window length of 50 microseconds to remove spike interference, and wavelet denoising algorithm is used to further eliminate high-frequency harmonic interference coupled from the grid side. Subsequently, the phase trajectories of different frequency bands are uniformly mapped to a normalized interval of 0 to 1 to ensure the comparability of energy amplitude across all frequency bands. After normalization, the phase trajectories of each frequency band are matched one-to-one according to time points, forming phase combination vectors at the same timestamp. As time progresses, these phase combination vectors are arranged sequentially, ultimately forming a multi-frequency band phase fingerprint sequence. This phase fingerprint sequence can simultaneously reflect low-frequency steady-state characteristics and high-frequency transient disturbances at the same time reference, thus forming a complete photovoltaic voltage dynamic response fingerprint.

[0066] After obtaining the multi-band phase fingerprint sequence, an initial energy transition baseline is established based on this sequence. Specifically, phase abrupt changes in the phase fingerprint sequence are identified, and points with phase changes greater than 5 degrees and durations exceeding 20 microseconds are selected as key feature points. After identifying the key feature points, the temporal distribution of these feature points is fitted using the least squares method to generate a smooth fitting curve, which is the initial energy transition baseline. This baseline not only records the transition trajectory of the photovoltaic array between steady state and disturbed state but also serves as a sensitive criterion for rapid power changes. When new sampling data is compared with this baseline, a significant phase deviation indicates that the photovoltaic output may be entering a rapid increase state, thus providing an early warning of power disturbances. The initial energy transition baseline established in this way can serve as a reference standard for the normal operation of the photovoltaic system and can be continuously updated and corrected during actual operation, maintaining high sensitivity and adaptability to the dynamic characteristics of the photovoltaic system over the long term.

[0067] Through the aforementioned sub-steps, the entire process from acquiring the instantaneous voltage response of the photovoltaic array to forming the initial energy transition baseline is realized. The entire process not only fully captures the full-frequency characteristics of the photovoltaic voltage signal but also amplifies and enhances microsecond-level changes at the phase level, ultimately forming a baseline curve that accurately characterizes the energy transition trend. This process provides a reliable basis for subsequent power gradient mapping, high-risk window identification, and energy stratification deposition, thereby effectively improving the power disturbance control capability in photovoltaic direct-drive solid thermal energy storage electric heating processes and preventing thermal stress concentration and crack propagation in the thermal storage body due to localized rapid heating.

[0068] Based on the initial energy transition baseline, a time-series dynamic gradient mapping is constructed to transform the phase shift trajectory into a continuous power gradient distribution, and potential acceleration segments with nonlinear steep rise characteristics are identified in the power gradient distribution.

[0069] To transform the phase-shifted trajectory obtained based on the initial energy transition baseline into a continuous power gradient distribution, and to identify potential acceleration segments with nonlinear steep rise characteristics within the power gradient distribution, a time-series dynamic gradient mapping solution is adopted. The specific process is as follows:

[0070] After obtaining the initial energy transition baseline, the phase fingerprint sequence is time-series processed based on this baseline. Specifically, using the initial energy transition baseline as a unified reference, each phase sampling point in the phase fingerprint sequence is time-aligned with this baseline, ensuring that all phase sampling points correspond one-to-one with their respective timestamps. This time-series processing ensures that the phase offset trajectory can be continuously unfolded under a unified time coordinate, thereby avoiding offset errors caused by sampling delays or noise between different frequency bands. After time alignment, the obtained phase offset trajectory can fully reflect the dynamic response characteristics of the photovoltaic array on a microsecond-level time scale and maintain consistency with the initial energy transition baseline.

[0071] After time-series processing, a first-order difference operation is performed on the phase shift trajectory to obtain the phase change rate. Specifically, the difference in phase angle is calculated for each adjacent time sampling point and divided by the corresponding time interval to obtain the instantaneous phase change rate. The phase change rate reflects the dynamic transition speed of the photovoltaic array voltage within that time period and is an important intermediate parameter for characterizing power change trends. To avoid interference from high-frequency noise amplification on the difference results, a sliding window weighted average method is used to smooth the results while calculating the phase change rate. The sliding window length is set to 20 microseconds, thereby effectively reducing the impact of random noise on the results while maintaining dynamic response sensitivity.

[0072] After obtaining the phase change rate, a time-series dynamic gradient mapping is established based on the initial energy transition baseline. Specifically, the phase change rate is superimposed point-by-point on the initial energy transition baseline, and the phase change rate is mapped to a power gradient through integral normalization. This integral normalization process uses the initial energy transition baseline as a reference, converting the phase change rate into an equivalent power change amplitude. Through this mapping method, the power gradient corresponding to the phase shift can be obtained at each time point, thus forming a complete continuous power gradient distribution curve. This distribution curve not only includes the trend of slow changes in photovoltaic power but also amplifies and quantifies the details of rapid power changes, making it possible to subsequently identify steep rise characteristics.

[0073] After forming a continuous power gradient distribution, nonlinear feature detection is performed on the distribution curve to identify potential acceleration segments. Specifically, the rate of change of the power gradient over time, i.e., power acceleration, is calculated by performing a second-order difference operation on the power gradient distribution. If the value of the power acceleration is consistently greater than a preset threshold (e.g., three standard deviations of the rate of change of power) within a certain time period, a nonlinear steep increase characteristic is determined to exist within that time period. To ensure the accuracy of identification, the threshold condition must be met for more than five consecutive time sampling points during the detection process before the segment can be marked as a potential acceleration segment. In this way, false judgments caused by instantaneous noise spikes can be effectively avoided, thereby ensuring the stability and reliability of the identification results.

[0074] After identifying potential acceleration segments, these segments are coupled with the original phase shift trajectory and the initial energy transition baseline to verify the consistency between the acceleration segments and changes in photovoltaic power output. Specifically, the power gradient in the potential acceleration segment is compared point-by-point with the anomalies in the phase shift trajectory. If the two highly overlap in time and the corresponding phase abrupt change amplitude exceeds a set threshold (e.g., 5 degrees), the segment is finally confirmed and labeled as an effective acceleration segment with a rapid photovoltaic power transition trend. The confirmation result of this acceleration segment will directly serve as the input for the subsequent construction of cross-scale time backtracking curves, providing a reliable basis for further amplifying key change features and generating enhanced energy abrupt change curves.

[0075] The above method achieves a complete process starting from the initial energy transition baseline, proceeding through phase trajectory time-series processing, first-order difference calculation of phase change rate, integral normalization to obtain power gradient distribution, second-order difference detection of nonlinear steep rise characteristics, and finally calibrating potential acceleration segments. The entire process not only discloses the mapping method from phase to power gradient but also proposes a nonlinear feature detection method based on second-order difference. Furthermore, joint verification with the initial energy transition baseline and phase shift trajectory ensures the accuracy and stability of the identification results. Therefore, it is possible to identify potential rapid upward trends in photovoltaic power before they manifest as significant output changes, thus providing solid data support for stratified energy deposition and stress mitigation control in solid thermal storage, significantly improving the safety and reliability of energy utilization in electric heating processes.

[0076] A cross-scale time backtracking curve is established within the potential acceleration segment, and key change features are enhanced by differential amplification. The key change features are coupled with the phase fingerprint sequence to generate an enhanced energy mutation curve.

[0077] To further reveal the implicit characteristics of rapid photovoltaic power transitions within the potential acceleration segment, this step establishes a cross-scale time-backtracking curve within the potential acceleration segment and enhances key change features through differential amplification. Finally, the key change features are coupled with a phase fingerprint sequence to generate an enhanced energy jump curve, effectively addressing this problem. The specific process is as follows:

[0078] After identifying potential acceleration segments, a cross-scale time backtracking curve is established using these segments as the core time window. Specifically, the time series is traced back at least five times the segment length from the start time of the potential acceleration segment, and then extended backward at least twice the segment length from the end time of the potential acceleration segment. This ensures that the constructed time backtracking interval covers both the antecedent and consequences of the potential acceleration segment. Within this time backtracking interval, the original phase shift trajectory and power gradient distribution are simultaneously preserved and resampled at different time scales. The resampling time scales include microseconds, milliseconds, and seconds, with a logarithmic interval between sampling points. This ensures that instantaneous details are captured at high time resolution, while the overall trend is depicted at low time resolution. This cross-scale backtracking method yields a time backtracking curve containing multi-level information at different time scales. This curve exhibits high-density sampling at potential acceleration segments and sparse sampling in regions far from these segments, thus balancing data processing accuracy and efficiency.

[0079] Based on the cross-scale time backtracking curve, a differential amplification method is employed to enhance key change features within the potential acceleration segment. Specifically, second-order and third-order differential operations are performed on the phase shift trajectory and power gradient distribution in the backtracking curve, respectively. The second-order differential operation highlights the acceleration characteristics of power changes, while the third-order differential operation reveals abrupt inflection points in power changes. After obtaining the differential results, the second-order and third-order differential curves are superimposed using amplitude normalization and energy weighting to form a composite enhanced signal. This composite enhanced signal amplifies key change features within the potential acceleration segment, clearly revealing subtle changes that were previously difficult to identify directly from the original trajectory. Simultaneously, to avoid amplifying random noise during the differential operation, wavelet thresholding denoising is used to filter the differential results, ensuring sufficient signal-to-noise ratio for the amplified key change features. In this process, differential amplification not only highlights instantaneous energy abrupt changes within the potential acceleration segment but also reveals the energy transfer relationships between different time levels in the cross-scale time backtracking curve, thereby achieving in-depth analysis of key change features.

[0080] After obtaining the key change features after differential amplification, these features are coupled point-by-point with the multi-band phase fingerprint sequence to generate an enhanced energy leap curve. Specifically, at each time point of the cross-scale time backtracking curve, the phase fingerprint vector corresponding to that time point is first extracted. This vector contains the normalized phase trajectory results from different frequency bands. Subsequently, the key change amplitude obtained from differential amplification is used as a weighting factor and applied to each component of the phase fingerprint vector to form a weighted phase fingerprint. After weighting, all weighted phase fingerprints are rearranged in chronological order to obtain a composite curve that integrates the phase features and the differential enhancement features. This composite curve is the enhanced energy leap curve. The enhanced energy leap curve not only retains the complete dynamic information of the phase fingerprint sequence in multiple frequency bands, but also significantly highlights the energy transition characteristics of potential acceleration segments through differential amplification, enabling the nonlinear steep rise phenomenon of photovoltaic power to be represented in a clearer and more intuitive form. Ultimately, this enhanced energy mutation curve will serve as the basis for subsequent dynamic threshold index setting and high-risk window discrimination, providing a highly sensitive predictive basis for energy regulation in photovoltaic direct-drive solid thermal energy storage heating.

[0081] Through the steps described above, this scheme achieves a complete process from identifying potential acceleration segments to establishing cross-scale time-backtracking curves, and then to differential amplification and coupling with phase fingerprints. This process not only fully utilizes the multi-level information provided by time backtracking but also amplifies key features through differential operations. Finally, the amplified features are combined with the phase fingerprint sequence to generate an enhanced energy mutation curve that clearly reflects the nature of energy mutations. Therefore, potential risks can be identified in the early stages of rapid photovoltaic power generation, effectively avoiding overheating and stress concentration problems caused by sudden energy injection into solid thermal storage.

[0082] Based on the enhanced energy mutation curve, a dynamic critical threshold index is set. Highly sensitive time segments that continuously cross the dynamic critical threshold index are screened in the enhanced energy mutation curve, and these highly sensitive time segments are marked as high-risk windows for a sharp increase in photovoltaic output.

[0083] To accurately identify the risk of rapid increases in photovoltaic power, a dynamic critical threshold index is set based on the enhanced energy mutation curve. Highly sensitive time segments that continuously cross the dynamic critical threshold index are then screened within the enhanced energy mutation curve. These highly sensitive time segments are ultimately identified as high-risk windows for a sharp increase in photovoltaic output. The specific process includes the following steps:

[0084] After obtaining the enhanced energy mutation curve, the initial range of the dynamic critical threshold index is determined based on the curve's statistical characteristics. Specifically, a sliding window statistical analysis is performed on the enhanced energy mutation curve, calculating the mean, standard deviation, and extreme value difference within each sliding window. This method allows the acquisition of the fluctuation characteristics of the enhanced energy mutation curve over a local time period. Then, based on the combination of the mean and standard deviation, the dynamic critical threshold index is set to the form of "mean plus twice the standard deviation" to ensure that the threshold adapts to change over time, rather than remaining fixed as a constant. This dynamic critical threshold index setting reflects the real-time changes in the curve, resulting in a lower threshold during stable periods and an automatically increasing threshold during periods of sharp fluctuations, thus balancing sensitivity and anti-interference capabilities.

[0085] After establishing a dynamic critical threshold index, the index is compared point-by-point with the enhanced energy mutation curve to detect whether the curve crosses the threshold. Specifically, at each sampling point, it is first determined whether the instantaneous amplitude of the enhanced energy mutation curve is greater than the dynamic critical threshold index at the corresponding time point. If this condition is met, the sampling point is marked as an "exceeding threshold point." During this process, to avoid false exceeding of the threshold caused by instantaneous noise, the exceeding of the threshold at adjacent sampling points is continuously detected. Only when three or more consecutive sampling points cross the threshold is it confirmed as a valid exceeding threshold segment. This method significantly improves the robustness of highly sensitive segment identification and avoids incorrect judgments due to random perturbations.

[0086] After obtaining valid overthreshold segments, these segments are time-aggregated to form highly sensitive time segments. Specifically, all overthreshold points that continuously cross the dynamic critical threshold index are merged into a single interval, and the start and end points of this interval are further extended by 20% of the interval length to ensure that the transition process before and after the overthreshold is included. The highly sensitive time segments formed in this way not only include the core interval of rapid power rise but also cover potential precursor signals and trailing effects, thus achieving more comprehensive coverage in terms of time range. Subsequently, all the formed highly sensitive time segments are sorted in chronological order, and by comparing them with the overall trend of the enhanced energy mutation curve, isolated pseudo-segments or those with a time length insufficient to meet the threshold are eliminated, ultimately resulting in a reliable set of highly sensitive time segments.

[0087] To eliminate isolated pseudo-segments with durations below the threshold, the threshold setting process is adaptively determined based on the temporal characteristic distribution of the enhanced energy mutation curve. Specifically, firstly, the duration of all identified highly sensitive time segments is statistically analyzed, and their mean and standard deviation are calculated. Then, the threshold is set as the mean minus the standard deviation, ensuring that most physically meaningful time segments are retained while abnormally short pseudo-segments are eliminated. To further improve robustness, the threshold is required not to fall below a preset lower limit, preferably 50 microseconds, to ensure that the length of the time segment covers at least two phase sampling periods. For example, if several highly sensitive time segments are identified in a single run, with an average duration of 500 microseconds and a standard deviation of 120 microseconds, the threshold calculation result is 380 microseconds. In this case, all segments with durations shorter than 380 microseconds will be eliminated and no longer included in the highly sensitive time segment set, while segments with durations greater than or equal to 380 microseconds will be retained as reliable results. The threshold set by this method can not only automatically adapt to the time characteristics under different operating conditions, but also effectively prevent short segments caused by noise or instantaneous spikes from entering the final set, thereby improving the accuracy and stability of high-risk window identification.

[0088] After obtaining the set of highly sensitive time segments, these are identified as high-risk windows for a sharp increase in photovoltaic power output. Specifically, within each highly sensitive time segment, the start point, end point, maximum amplitude point, and duration of the span (i.e., the cumulative duration of crossing the threshold) are recorded, and these parameters are used as components of the risk indicator. Subsequently, these high-risk windows are compared again with the phase fingerprint sequence. If a corresponding phase abrupt change point exists in the phase fingerprint sequence within a high-risk window, then that high-risk window is confirmed as the final risk segment for a rapid increase in photovoltaic power. Through this identification method, highly sensitive time segments no longer rely solely on the amplitude of the abrupt change curve but combine phase and duration characteristics, thus ensuring the high reliability and stability of the identified high-risk windows. Ultimately, the obtained high-risk windows will serve as the input basis for subsequent adaptive disturbance identification, directly providing a real-time reference for energy stratification and deposition scheduling in photovoltaic direct-drive solid-state thermal energy storage heating.

[0089] Through the above steps, a complete process is achieved, from obtaining the enhanced energy surge curve to setting the dynamic critical threshold index, screening highly sensitive time segments, and finally calibrating the high-risk window. The entire method not only solves the problem of insufficient adaptability of fixed thresholds through dynamic threshold indexing, but also improves the robustness of identification through continuous detection and time aggregation, and incorporates phase information in the final calibration, significantly enhancing the accuracy of high-risk window determination. Therefore, high-risk periods can be identified in the early stages of rapid photovoltaic power increase, effectively preventing structural failure and energy storage degradation of solid thermal storage during sudden energy injection.

[0090] An adaptive disturbance identification channel is constructed based on a high-risk window, which integrates the characteristics of the energy transition baseline, power gradient distribution and enhanced energy mutation curve to complete the identification and early warning output of the photovoltaic power output in real time.

[0091] To further improve the response speed and accuracy of detecting risks associated with rapid increases in photovoltaic (PV) output, an adaptive disturbance identification channel is constructed based on a high-risk window. This channel integrates multi-source features of the energy transition baseline, power gradient distribution, and enhanced energy mutation curve, ultimately enabling real-time detection and early warning output for PV output surges. The specific process includes the following steps:

[0092] After obtaining the high-risk window, it is used as the starting point of the discrimination interval to establish the input channel for the adaptive disturbance identification channel. Specifically, within the high-risk window, three types of basic features are simultaneously invoked: first, the phase offset trajectory from the initial energy transition baseline, which provides a transition benchmark for the photovoltaic array between steady-state and disturbed states; second, the power gradient distribution from the time-series dynamic gradient mapping, which reflects the speed and trend of power changes; and third, the enhanced energy mutation curve from the coupling analysis of potential acceleration segments, which amplifies and highlights the key nodes of rapid power transitions. During this process, the three types of features are aligned using a unified timestamp to ensure a one-to-one correspondence in the time dimension, enabling interactive analysis of data from different sources under the same reference coordinates. By using the high-risk window as the input range, the operation of the adaptive disturbance identification channel is concentrated within the most potentially threatening time period, thereby reducing redundant computation and improving real-time performance.

[0093] After aligning the three types of features, a weighted fusion approach is used to synthesize the energy transition baseline, power gradient distribution, and enhanced energy mutation curve. Specifically, the offset magnitude of the energy transition baseline is used as the basic weight, assigning it a benchmark function between steady state and disturbance; the instantaneous slope of the power gradient distribution is used as the dynamic weight to characterize the upward trend of photovoltaic power; and the mutation amplitude of the enhanced energy mutation curve is used as the enhancement weight to highlight potential abnormal energy injection signals. These three weights are normalized and applied to their respective feature values, and then linearly weighted at each time point to obtain an adaptive disturbance feature curve. This curve not only inherits the long-term stability of the energy transition baseline but also possesses the trend sensitivity of the power gradient distribution and the sudden response capability of the enhanced energy mutation curve. This weighted fusion approach adaptively adjusts the feature contribution under different operating conditions, ensuring that the disturbance identification results are both stable and highly sensitive to abnormal upward trends.

[0094] After obtaining the adaptive disturbance characteristic curve, it is evaluated in real time, and an early warning output for a sharp increase in photovoltaic power output is generated. Specifically, within each high-risk window, it is first detected whether the adaptive disturbance characteristic curve exceeds a set threshold within a short period of time. The threshold is the average amplitude of the curve in the stable phase plus three standard deviations. When the evaluation result shows that the adaptive disturbance characteristic curve continuously exceeds the threshold and the duration reaches more than one-third of the high-risk window duration, the high-risk window is confirmed as a state of sharp increase in photovoltaic power output. After confirming the sharp increase state, the evaluation result and the corresponding timestamp are output together to form an early warning signal. This early warning signal not only includes a binary determination of whether the increase state has been triggered, but also includes the amplitude index, duration index, and corresponding phase offset index of the power increase, thereby providing a refined control basis for subsequent energy stratification deposition and electric heating scheduling. Through the above real-time evaluation and early warning output, the energy distribution strategy can be triggered in advance before the sudden injection of photovoltaic power has fully acted on the solid thermal storage, avoiding thermal stress concentration and structural damage to the solid thermal storage due to energy impact.

[0095] Through the specific implementation steps described above, adaptive disturbance identification based on a high-risk window was achieved, along with a complete process from feature invocation and weighted fusion to real-time discrimination and early warning output. The entire process fully integrates the steady-state benchmark of the energy transition baseline, the dynamic trend of the power gradient distribution, and the sudden response of the enhanced energy mutation curve, constructing a disturbance identification channel capable of adaptive adjustment and real-time discrimination. Therefore, accurate identification can be performed in the early stages of a sharp increase in photovoltaic output, and reliable data support can be provided for the energy regulation of thermal storage through an early warning mechanism.

[0096] When it is determined that the photovoltaic output is entering a state of rapid increase, the output spectrum of the photovoltaic array is selectively coupled to guide different wavelength power into different energy storage paths (such as directing short-wavelength power to the direct heating path and long-wavelength power to the slow-release heating path) to achieve energy stratification deposition, thereby reducing the rapid thermal stress accumulation in a single area of ​​the solid thermal storage body.

[0097] To avoid localized overheating and thermal stress concentration caused by directly injecting photovoltaic power into the solid thermal storage body during a rapid increase, selective coupling of the photovoltaic array's output spectrum is performed after the photovoltaic output enters a state of rapid increase. Power in different wavelength bands is then directed to different energy storage paths, thereby achieving stratified energy deposition and reducing the risk of rapid energy accumulation in a single area of ​​the solid thermal storage body. The specific process includes the following steps:

[0098] After the adaptive disturbance identification channel determines that the photovoltaic (PV) output has entered a state of rapid increase, the DC output signal of the PV array is decomposed in the frequency domain to obtain the band components of the PV output spectrum. Specifically, the PV output voltage and current signals are sampled simultaneously, and their amplitude distribution in the frequency domain is calculated using a Fast Fourier Transform (FFT). Subsequently, the entire spectrum is divided into short-wavelength and long-wavelength bands according to time response characteristics: the short-wavelength band corresponds to high-frequency components, reflecting the transient power of the PV array during rapid disturbances; the long-wavelength band corresponds to low-frequency components, reflecting the continuous power of the PV array under steady-state regulation. This decomposition method allows for the clear differentiation of energy components at different time scales before the overall PV power injection, laying the foundation for subsequent selective coupling.

[0099] After obtaining the band division of the photovoltaic output spectrum, selective coupling relationships are established for short-wavelength and long-wavelength power respectively. Specifically, short-wavelength power, due to its rapid change rate and large amplitude fluctuations, is prone to inducing local thermal stress in the solid thermal storage body. Therefore, short-wavelength power is preferentially guided to the direct heating path. This direct heating path contacts the surface of the solid thermal storage body, rapidly consuming instantaneous energy at the surface and preventing the formation of a high temperature gradient inside. Conversely, long-wavelength power, due to its stable energy release and slow change, is more suitable for injection into the deeper layers of the thermal storage body through a slow-release heating path. This slow-release heating path uses thermal storage units with larger heat capacity to disperse the energy deposition rate, thereby reducing local thermal shock. Through this selective coupling method, short-wavelength and long-wavelength power can be guided to the most suitable energy storage path respectively, achieving stratified energy deposition in space and time.

[0100] After power stratification is completed, the energy injection rates of the two energy storage paths are coordinated in real time to ensure that the overall temperature rise of the thermal storage body remains controllable. Specifically, in the short-wavelength power-guided direct heating path, dynamic current limiting control is set to ensure that short-term energy injection does not exceed the allowable heat capacity of the thermal storage body's surface; in the long-wavelength power-guided slow-release heating path, progressive energy release control is set to ensure that energy is gradually conducted into the interior of the thermal storage body at a linear or quasi-linear rate. During this process, by monitoring the temperature distribution at different locations within the thermal storage body in real time, the temperature difference between the surface and deep layers is controlled within a preset range (e.g., not exceeding 20 degrees Celsius), thereby avoiding thermal stress concentration caused by excessive temperature differences. This combination of stratified deposition and dynamic coordination enables smooth energy transfer during rapid photovoltaic power injection, ensuring the structural stability of the thermal storage body.

[0101] During the energy stratification deposition process, a closed-loop correction is continuously applied to the energy flow of the direct heating path and the slow-release heating path to form a dynamic adaptive control mechanism. Specifically, when monitoring results indicate that the surface temperature rise rate of the thermal storage exceeds a set threshold, the injection ratio of short-wavelength power in the direct heating path is immediately reduced, and more power is transferred to the long-wavelength slow-release path. When monitoring results indicate that the deep temperature response of the thermal storage is lagging, resulting in insufficient overall temperature rise, the injection rate of long-wavelength power is appropriately increased to balance the energy distribution. Through this closed-loop correction method, dynamic switching and ratio adjustment between short-wavelength and long-wavelength power can be achieved, making the energy deposition process not only stratified but also adaptive. Ultimately, the injection of photovoltaic power is rationally decomposed and allocated, effectively reducing the energy impact in a single area, thereby reducing the risk of crack propagation and energy storage degradation caused by rapid temperature rise in the solid thermal storage.

[0102] Through the above steps, a complete process is achieved, from band division of the photovoltaic output spectrum to selective coupling of different power components, and then to energy stratification deposition and closed-loop correction. This method not only solves the energy concentration problem caused by the rapid increase in photovoltaic power, but also achieves a balanced distribution of the temperature field of the thermal storage body through an adaptive control mechanism.

[0103] This invention constructs a highly sensitive energy change identification mechanism that combines multi-band phase fingerprint sequences with dynamic gradient mapping. This mechanism enables multi-dimensional real-time monitoring of microsecond-level phase shifts in the output voltage of photovoltaic arrays. Based on an energy transition baseline, it extracts the dynamic evolution characteristics of the power gradient, capturing potential nonlinear transition trends before significant power output. Furthermore, it combines cross-scale time backtracking and multi-order differential enhancement methods to accurately amplify key perturbation features. Through deep coupling with the phase fingerprint sequence, it generates an enhanced energy mutation curve reflecting multi-temporal and spatial scale abrupt changes. Subsequently, a dynamic critical threshold index is constructed based on this mutation curve to filter out highly sensitive time segments that continuously cross the threshold, designating them as high-risk windows to achieve early identification of the risk of a sharp increase in photovoltaic power. After identifying the high-risk window, an adaptive perturbation identification channel is established by fusing the energy transition baseline, power gradient distribution, and enhanced mutation features, enabling rapid identification and dynamic early warning output of sudden increases in photovoltaic output. Ultimately, guided by the discrimination results, selective coupling is achieved based on the photovoltaic output spectrum. High-frequency short-wavelength power is directed to a fast-response direct heating path, while low-frequency long-wavelength power is directed to a slow-release gradual heating path, forming a spatially and temporally decoupled energy stratification deposition strategy. This strategy effectively avoids short-term concentrated energy injection into a single area of ​​the solid thermal storage body, significantly reduces the instantaneous peak value of local thermal stress, inhibits the initiation and propagation of microstructural cracks, and slows down the thermal decay process of the thermal storage material. This improves the operational stability, thermal response uniformity, and multi-cycle operating life of the thermal storage unit in photovoltaic direct-drive electric heating scenarios, promoting the efficient, safe, and long-term application of clean energy in the field of low-carbon heating.

[0104] The multi-energy collaborative control system with adaptive disturbance discrimination function includes a phase fingerprint construction module, a dynamic gradient mapping module, a feature enhancement coupling module, a dynamic threshold recognition module, an adaptive disturbance discrimination module, and an energy stratification deposition module;

[0105] The phase fingerprint construction module constructs a multi-band phase fingerprint sequence based on the instantaneous voltage response of the photovoltaic array. By extracting the microsecond-level phase shift trajectory in different frequency bands, an initial energy transition baseline is formed.

[0106] The dynamic gradient mapping module constructs a time-series dynamic gradient mapping based on the initial energy transition baseline, transforms the phase shift trajectory into a continuous power gradient distribution, and identifies potential acceleration segments in the power gradient distribution.

[0107] The feature enhancement coupling module establishes a cross-scale time backtracking curve within the potential acceleration segment and enhances key change features using differential amplification. It then couples the key change features with the phase fingerprint sequence to generate an enhanced energy mutation curve.

[0108] The dynamic threshold identification module sets a dynamic critical threshold index based on the enhanced energy mutation curve, filters out highly sensitive time segments that continuously cross the dynamic critical threshold index in the enhanced energy mutation curve, and marks the highly sensitive time segments as high-risk windows for a sharp increase in photovoltaic output.

[0109] The adaptive disturbance discrimination module constructs an adaptive disturbance identification channel based on a high-risk window, and integrates the features of the energy transition baseline, power gradient distribution and enhanced energy mutation curve to complete the discrimination and early warning output of the photovoltaic power output rising sharply.

[0110] The energy stratification deposition module selectively couples the output spectrum of the photovoltaic array when it detects that the photovoltaic output is entering a state of rapid increase, and introduces the power of different bands into different energy storage paths to achieve energy stratification deposition.

[0111] The multi-energy cooperative control method with adaptive disturbance discrimination function provided in this embodiment of the invention is implemented through the multi-energy cooperative control system with adaptive disturbance discrimination function described above. For details of the specific method and process of the multi-energy cooperative control system with adaptive disturbance discrimination function, please refer to the embodiment of the multi-energy cooperative control method with adaptive disturbance discrimination function described above, which will not be repeated here.

[0112] 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 above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the present invention.

Claims

1. A multi-energy collaborative control method of adaptive disturbance discrimination function, characterized in that, The method comprises the following steps: Based on the instantaneous voltage response of the photovoltaic array, a multi-band phase fingerprint sequence is constructed, the microsecond phase shift trajectory in different frequency bands is extracted, and an initial energy transition baseline is formed; Based on the initial energy transition baseline, a time series dynamic gradient mapping is constructed, the phase shift trajectory is converted into a continuous power gradient distribution, and potential acceleration segments are identified in the power gradient distribution; A cross-scale time backtracking curve is established in the potential acceleration segment, and the key change characteristics are strengthened using a differential amplification method, the key change characteristics are coupled with the phase fingerprint sequence, and an enhanced energy mutation curve is generated; Based on the enhanced energy mutation curve, a dynamic critical threshold index is set, high-sensitivity time segments that continuously cross the dynamic critical threshold index are screened in the enhanced energy mutation curve, and the high-sensitivity time segments are marked as high-risk windows of photovoltaic output sharp rise; Based on the high-risk window, an adaptive disturbance recognition channel is constructed, the features of the energy transition baseline, the power gradient distribution and the enhanced energy mutation curve are fused, and the discrimination and early warning output of the photovoltaic output sharp rise state are completed; When it is discriminated that the photovoltaic output enters the sharp rise state, the output spectrum of the photovoltaic array is selectively coupled, the power of different wave bands is guided into different energy storage paths, and energy is deposited in layers.

2. The method of claim 1, wherein, The generation step of the initial energy transition baseline comprises: The direct current output voltage signal of the photovoltaic array is sampled and divided into multiple frequency intervals with logarithmic distribution; The voltage signals in each frequency interval are subjected to fast Fourier transform and Hilbert transform to extract instantaneous phase angles, and at least 10 seconds of phase trajectories are saved in a circular buffer; The phase trajectories are subjected to noise suppression and normalization processing, the phase trajectories in different frequency bands are mapped to the interval of 0 to 1, and phase combination vectors are formed at the same time stamp to construct a multi-band phase fingerprint sequence; Based on the multi-band phase fingerprint sequence, key feature points with a phase change amplitude greater than 5 degrees and a duration of more than 20 microseconds are identified, and the initial energy transition baseline is generated by least square fitting.

3. The method of claim 2, wherein, When the initial energy transition baseline is generated by least square fitting, only when the key feature points with a phase change amplitude greater than 5 degrees and a duration of more than 20 microseconds appear continuously for three or more times, the fitting processing is performed.

4. The method of claim 2, wherein, The step of identifying potential acceleration segments comprises: After obtaining the initial energy transition baseline, the phase fingerprint sequence is subjected to time series processing, so that each phase sampling point is aligned with the initial energy transition baseline and keeps consistent with the corresponding time stamp; After completing the time series processing, a first-order differential operation is performed on the phase shift trajectory to obtain an instantaneous phase change rate, and the phase change rate is smoothed by using a weighted average method; After obtaining the phase change rate, the phase change rate is superimposed with the initial energy transition baseline point by point, and is converted into a power gradient by integral normalization to form a continuous power gradient distribution curve; After forming the continuous power gradient distribution curve, second-order difference operation is performed on the distribution curve, and when the power speed continuously exceeds five sampling points and is greater than three times the standard deviation of the power change rate, the segment is marked as a potential acceleration segment, and the potential acceleration segment is compared with the initial energy transition baseline and the phase shift trajectory point by point to confirm.

5. The method of claim 4, wherein, The confirmation of the potential acceleration segment includes: when the power speed continuously exceeds five sampling points and the corresponding phase mutation amplitude is greater than 5 degrees, it is determined that the segment is an effective acceleration segment of the rapid transition trend of photovoltaic power.

6. The method of claim 5, wherein, The generation steps of the enhanced energy mutation curve include: After identifying the effective acceleration segment, the core interval is taken as the effective acceleration segment, the time sequence is traced back not less than five times the length of the segment, and the time sequence is extended not less than two times the length of the segment, the phase shift trajectory and the power gradient distribution are re-sampled in logarithmic intervals in the core interval to form a cross-scale time backtracking curve; On the cross-scale time backtracking curve, the phase shift trajectory and the power gradient distribution are subjected to second-order difference and third-order difference operations respectively, and a composite enhancement signal is obtained through amplitude normalization and energy weight weighting; After obtaining the composite enhancement signal, it is coupled with the multi-band phase fingerprint sequence point by point, and the key change amplitude after difference amplification is used as the weight factor to weight the phase fingerprint vector, and finally the enhanced energy mutation curve is formed in time sequence.

7. The method of claim 1, wherein, The steps of setting a dynamic critical threshold index and demarcating a high-risk window based on the enhanced energy mutation curve include: After obtaining the enhanced energy mutation curve, the curve is subjected to sliding window statistics, the mean and standard deviation are calculated, and the dynamic critical threshold index is set to the mean plus two times the standard deviation; After establishing the dynamic critical threshold index, the threshold index is compared with the enhanced energy mutation curve point by point, and if the amplitudes of three or more consecutive sampling points are greater than the corresponding threshold, the interval of the consecutive sampling points is marked as an effective threshold segment; After obtaining the effective threshold segment, all consecutive threshold points are time-aggregated and extended by 20% of the interval length in the forward and backward directions to form a high-sensitivity time segment, and pseudo segments with insufficient time length are removed; After obtaining the set of high-sensitivity time segments, the start point, end point, maximum amplitude point and span length are taken as risk indicators, and are compared with the phase fingerprint sequence, and if they correspond to each other, the high-sensitivity time segment is demarcated as a high-risk window of rapid rise of photovoltaic output.

8. The method of claim 7, wherein, The steps of constructing an adaptive disturbance recognition channel based on the high-risk window include: After obtaining the high-risk window, the high-risk window is taken as the start point of the discrimination interval, and the mutation features of the phase shift trajectory, the power gradient distribution and the enhanced energy mutation curve are called in the interval respectively, and are aligned through unified time stamp; After completing the feature alignment, the energy transition baseline offset amplitude is taken as the basic weight, the power gradient distribution instantaneous slope is taken as the dynamic weight, and the enhanced energy mutation curve amplitude is taken as the enhancement weight, which are normalized and then linearly combined to generate an adaptive disturbance feature curve; After obtaining the adaptive disturbance characteristic curve, it is detected whether the curve continuously exceeds the threshold value in the high-risk window, the threshold value being the average amplitude of the stationary phase plus three times the standard deviation, if the duration of the adaptive disturbance characteristic curve continuously exceeding the threshold value exceeds one third of the length of the high-risk window, it is determined that the photovoltaic output is in a sharp rising state, and a warning signal containing an amplitude index, a duration index and a phase shift index is output.

9. The method of claim 8, wherein, The step of selectively coupling the output spectrum of the photovoltaic array and realizing energy layered deposition when the photovoltaic output is determined to enter the sharp rising state comprises: After it is determined that the photovoltaic output enters the sharp rising state, the direct current output signal of the photovoltaic array is frequency domain decomposed, the output spectrum is obtained through fast Fourier transform, and the frequency spectrum is divided into a short wave band corresponding to transient power and a long wave band corresponding to sustained power; After the wave band division of the photovoltaic output spectrum is obtained, the short wave band power is guided to a direct heating path to quickly consume transient energy on the surface layer, and the long wave band power is guided to a slow release heating path to gradually inject deep layers of the heat storage body; After the power layered guiding is completed, dynamic current limiting control is set for the direct heating path, progressive energy release control is set for the slow release heating path, and the temperature difference between the surface layer and the deep layer is monitored in real time; During the energy layered deposition process, the energy flow of the two energy storage paths is continuously closed loop corrected, the short wave band power proportion is reduced when the surface layer temperature rises too fast, and the long wave band power proportion is increased when the deep layer temperature rises lagged, so as to realize adaptive layered regulation.

10. A multi-energy collaborative control system with adaptive disturbance discrimination function, for implementing the multi-energy collaborative control method with adaptive disturbance discrimination function according to any one of claims 1-9, characterized in that, It comprises a phase fingerprint construction module, a dynamic gradient mapping module, a feature enhancement coupling module, a dynamic threshold identification module, an adaptive disturbance discrimination module and an energy layered deposition module; The phase fingerprint construction module constructs a multi-band phase fingerprint sequence based on the transient voltage response of the photovoltaic array, extracts microsecond level phase shift trajectories in different frequency bands to form an initial energy transition baseline; The dynamic gradient mapping module constructs a time series dynamic gradient mapping based on the initial energy transition baseline, converts the phase shift trajectory into a continuous power gradient distribution, and identifies potential acceleration segments in the power gradient distribution; The feature enhancement coupling module establishes a cross-scale time backtracking curve in the potential acceleration segment, and uses differential amplification to strengthen the key change characteristics, couples the key change characteristics with the phase fingerprint sequence, and generates an enhanced energy mutation curve; The dynamic threshold identification module sets a dynamic critical threshold index based on the enhanced energy mutation curve, screens high sensitivity time segments that continuously cross the dynamic critical threshold index in the enhanced energy mutation curve, and marks the high sensitivity time segments as a high-risk window of photovoltaic output sharp rising; The adaptive disturbance discrimination module constructs an adaptive disturbance recognition channel based on the high-risk window, fuses the features of the energy transition baseline, the power gradient distribution and the enhanced energy mutation curve, and completes the discrimination and early warning output of the photovoltaic output sharp rising state; The energy layered deposition module selectively couples the output spectrum of the photovoltaic array when the photovoltaic output is determined to enter the sharp rising state, guides different wave band powers into different energy storage paths, and realizes energy layered deposition.

Images