Grid fluctuation adaptive response control method for building photovoltaics and related devices

Abstract

This application relates to the field of building energy management technology and provides a grid fluctuation adaptive response control method for building-integrated photovoltaics (BIPV). The method includes: acquiring the power waveform sequence of the target BIPV at the grid connection point; performing fluctuation type analysis on the power waveform sequence; if an abnormal fluctuation type is identified, obtaining the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence; determining the risk coefficient and the principal component of the fluctuation based on the target power waveform segment; and outputting an optimal control sequence based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, and executing control commands for the corresponding equipment of the target BIPV according to the optimal control sequence. This invention solves the problems of low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types in existing methods, making it difficult to meet the needs of BIPV for active grid support under different operating conditions.

Description

Technical Field

[0001] This application belongs to the field of building energy management technology, and in particular relates to a grid fluctuation adaptive response control method and related equipment for building photovoltaics. Background Technology

[0002] With the advancement of the global energy transition, building-integrated photovoltaics (BIPV), a technology that combines photovoltaic power generation with building envelopes, has been widely applied. However, existing BIPV fluctuation detection and response methods typically set fixed voltage or power change rate thresholds to determine whether fluctuations have occurred and trigger energy storage or inverters for adjustment. These methods suffer from low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types, and are difficult to adapt to dynamic changes under different operating conditions. Therefore, there is an urgent need for a grid fluctuation response method capable of intelligent fluctuation identification, risk quantification assessment, and adaptive adjustment. This would address the shortcomings of existing methods, such as low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types, thus failing to meet the needs of BIPV for proactive grid support under various operating conditions. Summary of the Invention

[0003] This application provides a grid fluctuation adaptive response control method for building-integrated photovoltaics (BIPV), which solves the problems of low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types in existing methods, making it difficult to meet the needs of BIPV for active grid support under different operating conditions. By analyzing the fluctuation type of the power waveform sequence of the target BIPV at the grid connection point, when an abnormal fluctuation type is identified, the corresponding target power waveform segment is obtained from the power waveform sequence. Based on the target power waveform segment, the risk coefficient and the principal component of the fluctuation are determined. Using the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, the optimal control sequence is output, and the control commands for the corresponding equipment of the target BIPV are executed according to the optimal control sequence. This solves the problems of low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types in existing methods, making it difficult to meet the needs of BIPV for active grid support under different operating conditions.

[0004] In a first aspect, embodiments of this application provide a grid fluctuation adaptive response control method for building photovoltaics, the method comprising the following steps:

[0005] Obtain the power waveform sequence of the target building photovoltaic system at the grid connection point;

[0006] The power waveform sequence is subjected to fluctuation type analysis. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence.

[0007] Based on the target power waveform segment, the risk coefficient and the main component of the fluctuation are determined;

[0008] Based on the risk coefficient and the principal component of the fluctuation, combined with the preset optimization control model, the optimal control sequence is output, and the control instructions of the target building photovoltaic corresponding equipment are executed according to the optimal control sequence.

[0009] Optionally, the step of performing fluctuation type analysis on the power waveform sequence, and if an abnormal fluctuation type is identified, then obtaining the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence, specifically includes:

[0010] The power waveform sequence is subjected to multi-scale morphological filtering to extract waveform morphological features at different scales, and the filtered waveform segments are obtained.

[0011] The filtered waveform segment is classified and identified using a preset classification model. When the classification model identifies the abnormal fluctuation type and the confidence level is higher than the preset confidence level threshold, the fluctuation start time and fluctuation type are marked, and the marked waveform data is extracted from the filtered waveform segment to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

[0012] Optionally, determining the risk coefficient and the principal component of fluctuation based on the target power waveform segment specifically includes:

[0013] The target power waveform segment is decomposed by variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment;

[0014] Based on the multiple intrinsic mode components, the ripple principal component of the target power waveform segment is determined;

[0015] Furthermore, based on the principal component of the fluctuation, the risk coefficient of the target power waveform segment is determined in the historical event database, which stores all previous fluctuations and the power grid impact consequences caused by those fluctuations.

[0016] Optionally, the step of decomposing the target power waveform segment through variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment specifically includes:

[0017] The gray wolf optimization algorithm is used to optimize the number of modes and the penalty factor of the variational mode decomposition, resulting in optimized number of modes and optimized penalty factor.

[0018] Based on the optimized number of modes and the optimized penalty factor, the target power waveform segment is decomposed to obtain multiple intrinsic mode components of the target power waveform segment.

[0019] Optionally, determining the wave principal component of the target power waveform segment based on multiple intrinsic mode components specifically includes:

[0020] Calculate the multiscale entropy vector for each of the intrinsic mode components;

[0021] And, calculate the Pearson correlation coefficient between each of the intrinsic mode components and the target power waveform segment;

[0022] The intrinsic mode component with the highest Pearson correlation coefficient and the largest multi-scale entropy fluctuation was selected as the main fluctuation component.

[0023] Optionally, before outputting the optimal control sequence based on the risk coefficient and the principal component of volatility, combined with a preset optimization control model, the method further includes:

[0024] An optimized control model is constructed, wherein the state variables of the optimized control model include: energy storage state of charge, building indoor temperature, and current grid-connected power; the control variables of the optimized control model include: energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset; and the weights of the optimized control model are adjusted according to the risk coefficient.

[0025] Based on the principal component of the fluctuation, the constraint boundary of the optimized control model is dynamically adjusted.

[0026] Optionally, after executing the control commands for the target building photovoltaic (BPV) equipment according to the optimal control sequence, the method further includes:

[0027] Calculate the volatility suppression rate and control cost;

[0028] The optimized control model is subjected to reinforcement learning based on the fluctuation suppression rate and the control cost.

[0029] Secondly, embodiments of this application provide a grid fluctuation adaptive response control device for building-integrated photovoltaics (BIPV), the grid fluctuation adaptive response control device for BIPV comprising:

[0030] The acquisition module is used to acquire the power waveform sequence of the target building photovoltaic system at the grid connection point;

[0031] The processing module is used to perform fluctuation type analysis on the power waveform sequence. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence.

[0032] The determination module is used to determine the risk coefficient and the main component of fluctuation based on the target power waveform segment;

[0033] The control module is used to output an optimal control sequence based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, and execute control commands for the target building photovoltaic equipment according to the optimal control sequence.

[0034] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the grid fluctuation adaptive response control method for building photovoltaics provided in embodiments of the present invention.

[0035] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the grid fluctuation adaptive response control method for building photovoltaics provided in the embodiments of the present invention.

[0036] The above-mentioned solution of this application has the following beneficial effects: It obtains the power waveform sequence of the target building-integrated photovoltaic (BIPV) system at the grid connection point; it performs fluctuation type analysis on the power waveform sequence, and if an abnormal fluctuation type is identified, it obtains the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence; it determines the risk coefficient and the principal component of the fluctuation based on the target power waveform segment; based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, it outputs the optimal control sequence, and executes the control command of the corresponding equipment of the target BIPV system according to the optimal control sequence. This invention solves the problems of low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types in existing methods, making it difficult to meet the needs of BIPV for active grid support under different operating conditions.

[0037] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 A flowchart illustrating a grid fluctuation adaptive response control method for building photovoltaics provided in one embodiment of this application;

[0040] Figure 2 A schematic diagram of a grid fluctuation adaptive response control device for building photovoltaics provided in an embodiment of this application;

[0041] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0042] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0043] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0044] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0045] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0046] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0047] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0048] like Figure 1 As shown, Figure 1 This is a flowchart of a grid fluctuation adaptive response control method for building-integrated photovoltaics (BIPV) provided by an embodiment of the present invention. The grid fluctuation adaptive response control method for BIPV includes the following steps:

[0049] 101. Obtain the power waveform sequence of the target building photovoltaic system at the grid connection point.

[0050] In this embodiment of the invention, the above-described adaptive response control method for grid fluctuations in building-integrated photovoltaics (BIPV) can be applied to a BIPV platform. The BIPV platform can be built on a server-based or distributed architecture. The BIPV platform includes a data interface (for sensors or users to upload data), a knowledge database, and a knowledge database construction program. The data interface can be used to acquire the power waveform sequence of the target BIPV at the grid connection point. The knowledge database construction program can be used to construct the knowledge database, which is specifically used to provide additional correlation information for identified data entities, thereby improving the depth of the data recognition system's understanding of the content.

[0051] The aforementioned target building-integrated photovoltaics (BIPV) can be BIPV that requires adaptive response control to grid fluctuations.

[0052] The aforementioned grid connection point can be the connection point between the target building's photovoltaic system and the public power grid.

[0053] The aforementioned power waveform sequence can be a sequence of active power values ​​collected or calculated in continuous time sequence at the grid connection point of the target building photovoltaic system. For example, it can be a discrete sequence of active power obtained by collecting the instantaneous values ​​of the three-phase voltage and current of the target building photovoltaic system at the grid connection point at a sampling rate of 10kHz, calculating the instantaneous power, and arranging them in time sequence.

[0054] It is understandable that the instantaneous values ​​of the three-phase voltage and current at each sampling moment can be synchronously acquired by voltage and current sensors installed at the grid connection point at a sampling frequency of 10kHz or higher. After instantaneous power calculation, the power waveform sequence is arranged in chronological order to obtain a power waveform sequence, which retains the transient change characteristics of the power at the grid connection point.

[0055] 102. Perform fluctuation type analysis on the power waveform sequence. If an abnormal fluctuation type is identified, obtain the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence.

[0056] In this embodiment of the invention, the above-mentioned fluctuation type analysis processing can be a process of identifying and classifying the fluctuations of power waveform sequences in different dimensions.

[0057] The aforementioned fluctuation types include normal waveform types and abnormal fluctuation types. Normal waveform types refer to power waveforms where the power variation of the target building-integrated photovoltaic system at the grid connection point is within an acceptable range and does not threaten the safe operation of the power grid. Abnormal fluctuation types refer to abnormal power fluctuations that affect the safe operation of the power grid or power quality.

[0058] When the fluctuation type of the power waveform sequence is identified as an abnormal fluctuation type, the target power waveform segment corresponding to the abnormal fluctuation type can be obtained from the power waveform sequence.

[0059] The aforementioned target power waveform segment can be a data segment extracted from the power waveform sequence based on the type of abnormal fluctuation, containing the complete process of the abnormal fluctuation.

[0060] It should be noted that fluctuation type analysis can be performed on the power waveform sequence. When the fluctuation type of the power waveform sequence is identified as an abnormal fluctuation type, the target power waveform segment corresponding to the abnormal fluctuation type can be obtained from the power waveform sequence.

[0061] 103. Based on the target power waveform segment, determine the risk coefficient and the principal component of the fluctuation.

[0062] In this embodiment of the invention, the risk coefficient and the main component of fluctuation can be determined based on the target power waveform segment.

[0063] The aforementioned risk coefficient can be a quantifiable value, such as a number between 0 and 1. This risk coefficient measures the degree of harm that a target power waveform segment poses to grid stability and equipment safety.

[0064] In one possible implementation, the risk coefficient can be labeled as R. When R > 0.7, it is marked as high risk and requires emergency suppression; when 0.3 ≤ R ≤ 0.7, it is marked as medium risk and regular suppression can be initiated; when R < 0.3, it is marked as low risk and only observations can be recorded.

[0065] The aforementioned main component of fluctuation can be the dominant component extracted from the target power waveform segment that best represents the core physical characteristics of the current abnormal fluctuation, i.e., the random fluctuation component that needs to be suppressed.

[0066] 104. Based on the risk coefficient and the principal component of the fluctuation, combined with the preset optimization control model, the optimal control sequence is output, and the control instructions of the target building photovoltaic corresponding equipment are executed according to the optimal control sequence.

[0067] In this embodiment of the invention, the aforementioned preset optimization control model can be a pre-set optimization control model. This optimization control model can be an optimization control model built based on deep learning or machine learning, such as MPC, LQR, etc. The core idea of ​​the aforementioned NPC (Model Predictive Control) is "model-based, predicting the future, and rolling optimization." NPC does not rely on a fixed control law, but instead solves a finite-time domain optimal control problem online at each sampling time, executing only the first control action in the optimization sequence, and then repeating the process at the next time step. The aforementioned LQR (Linear Quadratic Regulator) achieves the optimal trade-off between system state deviation and control energy consumption by minimizing a quadratic performance index.

[0068] The aforementioned optimized control model, within a control cycle, solves for a set of optimal control variables based on the current state and future predictions of the target building-integrated photovoltaic (BIPV) system, minimizing the objective function while satisfying various constraints. The control cycle can be as short as 1 minute. The current state includes energy storage state of charge, building indoor temperature, and current grid-connected power. The control variables include energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset. The future prediction can be a power prediction interval within a future time period, such as 5 minutes or 10 minutes. Specifically, this can be achieved by reading the power prediction interval within a future time period output by an ultra-short-term photovoltaic (USTP) power prediction model (such as CNN-LSTM-Attention). The power prediction interval can be the predicted value at each moment and its 95% confidence interval upper and lower bounds. The aforementioned USTP power prediction model can predict photovoltaic power generation within the next 0-4 hours (typically with a 15-minute time resolution, i.e., outputting a prediction value every 15 minutes). The aforementioned CNN-LSTM-Attention is a hybrid neural network model that integrates three deep learning architectures, specifically designed to process sequence data with spatiotemporal correlations. The core idea of ​​CNN-LSTM-Attention is to extract and enhance key features in stages, thereby improving prediction accuracy and model interpretability.

[0069] Furthermore, the risk coefficient and the principal component of volatility can be used as inputs to a preset optimization control model for calculation, and the optimal control sequence can be output.

[0070] Specifically, the objective function for optimizing the control model is:

[0071]

[0072] in, Represents the objective function value. Representing the future time domain, Indicates time Grid-connected power; Indicates the grid-connected power reference value; Indicates time The energy storage state of charge; Indicates the reference value for the state of charge of energy storage; Indicates time The air conditioner set temperature offset; and Satisfying 1.

[0073] The constraints for optimizing the control model are:

[0074] Power balance: ,in, Indicates time The photovoltaic power generation capacity is represented in interval form within the constraints, meaning that the solution must ensure that within... and The constraints can be satisfied for any value (robust equivalence transformation); Indicates time The energy storage charging and discharging power; Indicates time Grid-connected power; Indicates time The basic load, and the power of other electrical equipment in the building excluding air conditioning; Indicates time The air conditioning load.

[0075] Energy storage constraints: , .in, This represents the minimum value of the energy storage state of charge. This represents the maximum value of the energy storage state of charge. Indicates time The energy storage state of charge; Indicates time The energy storage charging and discharging power; This indicates the maximum charging and discharging power of the energy storage.

[0076] Thermal comfort constraints: .in, Indicates the lower limit of the temperature setting; This indicates the upper limit of the temperature setting; Indicates indoor temperature.

[0077] Grid-connected power constraints: .in, Indicates time Grid-connected power; This indicates the maximum allowable grid-connected power.

[0078] Power factor constraints can be restrictions imposed on the reactive power output of inverters to ensure power quality and equipment safety at the grid connection point.

[0079] Furthermore, in each control cycle, the current risk coefficient and the principal component of fluctuation can be input into the optimization control model for optimization, obtaining the optimal control sequence for the future time period, and then executing the control commands for the target building photovoltaic equipment according to the optimal control sequence. It should be noted that only the control commands for the current time period (…) are executed. Control commands:

[0080] The control command can be to send the optimal charging and discharging power of the energy storage to the power storage converter (PCS) at the current moment. The control commands will optimize the charging and discharging power of the energy storage. Control commands are sent to the energy storage converter (PCS) to control the charging and discharging of the battery.

[0081] Control commands can be used to send reactive power to the grid-connected inverter. The control command to optimize the inverter's reactive power. The control commands are sent to the grid-connected inverter to adjust the reactive power output to support the voltage.

[0082] The control command can be to send the current air conditioning set temperature adjustment amount to the building automation system. Control commands. Utilizing the virtual energy storage characteristics of building thermal inertia, the air conditioning power can be adjusted in a short time without affecting human comfort, thus smoothing out net load fluctuations.

[0083] It should be noted that at the next moment, the status can be updated based on the latest measured data. By combining the latest obtained risk coefficient and the principal component of fluctuation with the preset optimization control model, the optimal control sequence can be output. The control instructions of the target building photovoltaic equipment can be executed according to the optimal control sequence to achieve rolling optimization control.

[0084] The aforementioned optimal control sequence can be a sequence of optimal values ​​predicted based on the risk coefficient and the principal component of fluctuation, combined with a preset optimization control model. The optimal control sequence can be a sequence where the preset optimization control model, based on the input risk coefficient and principal component of fluctuation, considers the current state of the target building-integrated photovoltaic (BIPV) system and adjustable modes, seeking the best balance among multiple objectives, such as suppressing fluctuations while minimizing battery life loss and minimizing impact on indoor comfort. The current state of the target BIPV system can include remaining energy storage capacity, indoor temperature, and current grid-connected power, while the adjustable modes can include energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature.

[0085] The aforementioned control commands can be execution commands sent to the corresponding photovoltaic devices in the target building.

[0086] In one possible embodiment, when the fluctuation type is cloud-shielded, the optimal control sequence is obtained by combining the risk factor and the main component of the fluctuation with the optimized control model. This sequence prioritizes the use of energy storage, with a response time of <200ms and a prediction time domain shortened to 5 minutes by the optimized control model. The corresponding control command for the device is then executed as a fast power compensation command. When the fluctuation type is load-switching, the optimal control sequence is obtained by combining the risk factor and the main component of the fluctuation with the optimized control model. This sequence prioritizes the inactive inverter, dynamically adjusting the QV droop curve. The corresponding control command for the device is then executed as a voltage support command. When the fluctuation type is resonant, the optimal control sequence is obtained by combining the risk factor and the main component of the fluctuation with the optimized control model. This sequence switches the inverter to an active damping control algorithm, actively injecting harmonic cancellation current. The corresponding control command for the device is then executed as an active damping command.

[0087] It should be noted that this invention optimizes the control model, takes into account the uncertainty of photovoltaic output, avoids control failure caused by prediction errors, and reduces the amplitude of grid power fluctuations.

[0088] In this embodiment of the invention, the power waveform sequence of the target building-integrated photovoltaic (BIPV) system at the grid connection point is obtained; fluctuation type analysis is performed on the power waveform sequence, and if an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence; based on the target power waveform segment, the risk coefficient and the principal component of the fluctuation are determined; based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimization control model, the optimal control sequence is output, and the control command of the corresponding equipment of the target BIPV system is executed according to the optimal control sequence. This invention solves the problems of low detection sensitivity, susceptibility to noise interference, and inability to identify fluctuation types in existing methods, making it difficult to meet the needs of BIPV for active grid support under different operating conditions.

[0089] It is understood that in the specific embodiments of this application, data related to photovoltaic data, power data, waveform data, control data, etc. are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required. Furthermore, the collection, use and processing of related data, as well as the training, deployment and invocation of algorithm models, must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0090] Optionally, in the step of analyzing the fluctuation type of the power waveform sequence and obtaining the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence if an abnormal fluctuation type is identified, the power waveform sequence can be subjected to multi-scale morphological filtering to extract waveform morphological features at different scales and obtain filtered waveform segments. The filtered waveform segments are then classified and identified using a preset classification model. When the classification model identifies the abnormal fluctuation type and the confidence level is higher than the preset confidence level threshold, the fluctuation start time and fluctuation type are marked, and the marked waveform data is extracted from the filtered waveform segments to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

[0091] In this embodiment of the invention, the power waveform sequence can be a power waveform sequence composed of active power values ​​collected or calculated in a continuous time sequence at the grid connection point.

[0092] The aforementioned multi-scale morphological filtering process can be achieved by using structuring elements of different scales to perform opening-closing and closing-opening operations on the power waveform sequence. Specifically, a flat structuring element can be used, setting n different scales to perform opening-closing and closing-opening operations on the power waveform sequence respectively, extracting waveform morphological features at different scales, and obtaining filtered waveform segments. The aforementioned flat structure is an organizational structure model in management, which can be a management model formed by reducing vertical levels, expanding the management span, and delegating power. The aforementioned opening-closing operation can involve first performing an opening operation on the power waveform sequence, and then performing a closing operation on the result of the opening operation. The opening-closing operation can effectively remove white noise while preserving the target contour. The aforementioned closing-opening operation can involve first performing a closing operation on the power waveform sequence, and then performing an opening operation on the result of the closing operation. The closing-opening operation can effectively remove black noise and fill small holes. Understandably, in actual power grid fluctuations, noise can be bidirectional (with both upward spikes and downward dips). Performing on-off and off-on operations on the power waveform sequence separately ensures that bidirectional noise is effectively filtered out, achieving more comprehensive noise suppression and structure preservation. The aforementioned n can be an integer number such as 2, 3, or 4, and the aforementioned scale can be 5, 10, or 20 sampling points, etc. By performing multi-scale morphological filtering on the power waveform sequence, high-frequency noise can be effectively filtered out while preserving the transient characteristics of the fluctuations.

[0093] The aforementioned waveform morphology features can be quantitative indicators that reflect the geometric shape attributes of the power waveform in the time domain, extracted after multi-scale morphological filtering of the power waveform sequence.

[0094] The filtered waveform segments described above can be obtained by performing multi-scale morphological filtering on the power waveform sequence to extract waveform morphological features at different scales. The filtered waveform segments can be obtained by dividing the waveform morphological features at different scales according to a preset time window length, resulting in several data units of equal length. The preset time window length can be a pre-set time window length, such as 0.1 seconds.

[0095] The aforementioned preset classification model can be a pre-defined classification model, which can be a classification model built based on deep learning or machine learning, such as 1D-CNN. The 1D-CNN (One-Dimensional Convolutional Neural Network) is a deep learning model used to process one-dimensional sequential data. The core idea of ​​1D-CNN is to extract local features in a single temporal or spatial dimension through sliding convolutional kernels. The model structure of the 1D-CNN is as follows: ① Input layer: 1000×3 (three-phase current / voltage); ② Convolutional layer 1: 32 convolutional kernels, size 3×3, stride 1, ReLU activation function; ③ Pooling layer 1: Max pooling, size 2×2; ④ Convolutional layer 2: 64 convolutional kernels, size 3×3; ⑤ Pooling layer 2: Global average pooling; ⑥ Fully connected layer: 128 neurons, Dropout=0.5; ⑦ Output layer: Softmax, 4 fluctuation types (normal, cloud cover fluctuation, load switching fluctuation, resonance fluctuation). 1D-CNN automatically learns local patterns of waveforms at different scales and channels through convolutional layers, reduces dimensionality and enhances robustness through pooling layers, and finally outputs classification results through fully connected layers and output layers.

[0096] The above classification and recognition process can be a process of classifying and recognizing filtered waveform segments using a preset classification model.

[0097] The aforementioned fluctuation types include normal fluctuation types and abnormal fluctuation types. Normal waveform types refer to normal power waveforms where the power variation of the target building-integrated photovoltaic system at the grid connection point is within an acceptable range and will not threaten the safe operation of the power grid. Abnormal fluctuation types refer to abnormal power fluctuation types that affect the safe operation of the power grid or power quality. Abnormal fluctuation types can include fluctuations caused by cloud cover, load switching, resonance, etc.

[0098] The aforementioned preset confidence threshold can be a pre-set confidence threshold. The confidence threshold is a probability threshold used to determine whether the classification result output by the preset classification model is sufficiently reliable, such as 85%.

[0099] Furthermore, when the classification model identifies the abnormal fluctuation type and the confidence level is higher than the preset confidence threshold, the fluctuation start time and fluctuation type are marked, and the marked waveform data is extracted from the filtered waveform segment to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

[0100] The aforementioned starting time of fluctuation can be the starting point of the fluctuation event on the time axis in the abnormal fluctuation type.

[0101] The aforementioned target power waveform segment can be a data segment extracted from the filtered waveform segment that contains the complete process of abnormal fluctuations.

[0102] It should be noted that multi-scale morphological filtering can be applied to the power waveform sequence to extract waveform morphological features at different scales, resulting in filtered waveform segments. These segments are then classified and identified using a pre-set classification model. When the classification model identifies the abnormal fluctuation type and the confidence level is higher than the pre-set confidence threshold, the fluctuation start time and fluctuation type are marked. The marked waveform data is then extracted from the filtered waveform segments to obtain the target power waveform segment corresponding to the abnormal fluctuation type. By combining multi-scale morphological filtering with a classification model, transient fluctuations can be identified and classified, improving the identification accuracy and providing high-quality input for subsequent analysis.

[0103] Optionally, in the step of determining the risk coefficient and the principal component of fluctuation based on the target power waveform segment, the target power waveform segment can be decomposed by variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment; the principal component of fluctuation of the target power waveform segment can be determined based on the multiple intrinsic mode components; and the risk coefficient of the target power waveform segment can be determined in the historical event database based on the principal component of fluctuation.

[0104] In this embodiment of the invention, Variational Mode Decomposition (VMD) is a signal decomposition method based on variational optimization. By solving a constrained variational problem, it adaptively decomposes the original signal into several intrinsic mode functions (IMFs) with specific center frequencies and finite bandwidths. The core objective of VMD is to minimize the bandwidth of each IMF while ensuring that the sum of all IMFs equals the original signal.

[0105] The aforementioned target power waveform segment is a data segment extracted from the filtered waveform segment that contains the complete process of abnormal fluctuations.

[0106] The above decomposition process can be a process of decomposing the target power waveform segment into multiple intrinsic mode components through variational mode decomposition.

[0107] The aforementioned intrinsic mode components can be the basic signal components extracted from the target power waveform segment through variational mode decomposition, used to characterize the local oscillation modes at different time scales in the target power waveform segment.

[0108] The aforementioned main component of fluctuation can be the dominant component that best represents the core physical characteristics of abnormal fluctuations, extracted from the target power waveform segment based on multiple intrinsic mode components, i.e., the random fluctuation component that needs to be suppressed.

[0109] The aforementioned historical event database stores the risk coefficients of each fluctuation and the resulting power grid impact, such as the duration of voltage over-limit and the integral value of frequency deviation.

[0110] The aforementioned risk coefficient can measure the degree of harm that a target power waveform segment poses to grid stability and equipment safety; for example, it can be a number between 0 and 1.

[0111] It should be noted that the target power waveform segment can be decomposed by variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment. Based on the multiple intrinsic mode components, the main fluctuation component of the target power waveform segment can be determined. Furthermore, based on the main fluctuation component, the risk coefficient of the target power waveform segment can be determined in the historical event database, thereby improving the accuracy of abnormal fluctuation identification.

[0112] Optionally, in the step of decomposing the target power waveform segment through variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment, the gray wolf optimization algorithm can be used to optimize the number of modes and the penalty factor of variational mode decomposition to obtain the optimized number of modes and the optimized penalty factor; based on the optimized number of modes and the optimized penalty factor, the target power waveform segment is decomposed to obtain multiple intrinsic mode components of the target power waveform segment.

[0113] In this embodiment of the invention, the Grey Wolf Optimizer (GWO) is a swarm intelligence optimization algorithm that simulates the social hierarchy and hunting behavior of grey wolves.

[0114] The aforementioned mode number and penalty factor are parameters for variational mode decomposition (VMD). The mode number refers to the number of modal components obtained after decomposing the original signal using VMD. Each modal component represents the oscillation mode of a specific frequency band within the target power waveform segment. A larger mode number results in a finer decomposition, distinguishing more different frequency components; conversely, a smaller mode number leads to a coarser decomposition, where multiple frequency components may overlap within the same intrinsic mode component. The penalty factor controls the bandwidth of each modal component. It determines the concentration of each intrinsic mode component in the frequency domain. A larger penalty factor results in a narrower mode bandwidth, more concentrated frequency components, and higher requirements for frequency resolution; conversely, a smaller penalty factor results in a wider mode bandwidth, a larger allowable frequency range, and the inclusion of more noise or adjacent frequency components.

[0115] The above optimization can be achieved by using the Grey Wolf optimization algorithm to optimize the number of modes and the penalty factor in variational mode decomposition.

[0116] The optimized number of modes can be obtained by optimizing the number of modes in variational mode decomposition using the Grey Wolf optimization algorithm.

[0117] The optimized penalty factor mentioned above can be obtained by optimizing the penalty factor of variational mode decomposition using the Grey Wolf optimization algorithm.

[0118] The above decomposition process can be a process of decomposing the target power waveform segment according to the optimized mode number and the optimized penalty factor to obtain multiple modal components of the target power waveform segment.

[0119] The aforementioned intrinsic mode components can be basic signal components extracted from the target power waveform segment based on the optimized mode number and the optimized penalty factor, and are used to characterize the local oscillation modes at different time scales in the target power waveform segment.

[0120] Understandably, the objective function of optimization is to minimize the envelope entropy of each intrinsic mode component after decomposition. The optimized number of modes can range from 3 to 10, the optimized penalty factor can range from 500 to 4000, the number of optimization iterations of the gray wolf optimization algorithm can be set to 30, and the population size can be 20, etc.

[0121] It should be noted that the Gray Wolf optimization algorithm can be used to optimize the number of modes and the penalty factor in variational mode decomposition, resulting in optimized mode numbers and penalty factors. Based on the optimized mode numbers and penalty factors, the target power waveform segment is decomposed to obtain multiple intrinsic mode components of the target power waveform segment. The Gray Wolf optimization algorithm achieves the decoupling of the wave origin.

[0122] Optionally, in the step of determining the wave principal component of the target power waveform segment based on multiple intrinsic mode components, the multi-scale entropy vector of each intrinsic mode component can be calculated; and the Pearson correlation coefficient between each intrinsic mode component and the target power waveform segment can be calculated; the intrinsic mode component with the highest Pearson correlation coefficient and the largest multi-scale entropy value fluctuation can be selected as the wave principal component.

[0123] In this embodiment of the invention, the aforementioned multi-scale entropy vector can be a vector composed of the calculated sample entropies for each intrinsic mode component at multiple different scale factors. The multi-scale entropy vector is used as a quantitative indicator to measure the complexity of intrinsic mode components at different time scales. For example, different scale factors τ = 1, 2, 3, 5, 8, etc., can be set to calculate the sample entropy values ​​at each scale, forming a multi-scale entropy vector.

[0124] The Pearson correlation coefficient is a statistical indicator used to measure the degree of linear correlation between each intrinsic mode component and the target power waveform segment. The Pearson correlation coefficient ranges from [-1, 1]. The closer the Pearson correlation coefficient is to 1, the stronger the linear correlation between the intrinsic mode component and the target power waveform segment, that is, the higher the proportion of energy of the intrinsic mode component in the target power waveform segment.

[0125] Furthermore, the intrinsic mode component with the highest Pearson correlation coefficient and the largest multi-scale entropy fluctuation can be selected as the principal component of the fluctuation.

[0126] The aforementioned principal components of fluctuation can be the intrinsic mode components with the highest Pearson correlation coefficient and the largest multi-scale entropy fluctuation, the components that best represent the core characteristics of abnormal fluctuation events, i.e., the random fluctuation components that need to be suppressed.

[0127] Understandably, the highest Pearson correlation coefficient indicates a stronger linear correlation between the intrinsic modal components and the target power waveform segment, meaning a higher proportion of energy in the target power waveform segment is occupied by the intrinsic modal components. The largest fluctuation in multi-scale entropy values ​​indicates that the sample entropy values ​​of the intrinsic modal components exhibit significant changes or fluctuations under different scale factors.

[0128] It should be noted that the multi-scale entropy vector of each intrinsic mode component can be calculated, as well as the Pearson correlation coefficient between each intrinsic mode component and the target power waveform segment. The intrinsic mode component with the highest Pearson correlation coefficient and the largest fluctuation in multi-scale entropy value is selected as the main fluctuation component, i.e. the random fluctuation component that needs to be suppressed, and the rest are regarded as trend components or noise components. This can accurately extract abnormal fluctuation features and improve the detection accuracy.

[0129] Optionally, before the step of outputting the optimal control sequence based on the risk coefficient and the principal component of fluctuation, combined with the preset optimal control model, an optimal control model can be constructed; and the constraint boundary of the optimal control model can be dynamically adjusted based on the principal component of fluctuation.

[0130] In this embodiment of the invention, the aforementioned optimized control model can be an optimized control model built based on deep learning or machine learning, such as MPC, LQR, etc. The core idea of ​​the aforementioned NPC (Model Predictive Control) is "model-based, predicting the future, and rolling optimization". NPC does not rely on a fixed control law, but solves an optimal control problem in a finite time domain online at each sampling time, executes only the first control action in the optimization sequence, and then repeats the process at the next time. The aforementioned LQR (Linear Quadratic Regulator) achieves the optimal trade-off between system state deviation and control energy consumption by minimizing a quadratic performance index.

[0131] The state variables of the above-mentioned optimized control model include: energy storage state of charge, building indoor temperature, and current grid-connected power; the control variables of the above-mentioned optimized control model include: energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset. The weights of the above-mentioned optimized control model are adjusted according to the risk coefficient.

[0132] Specifically, the objective function for optimizing the control model is:

[0133]

[0134] in, Represents the objective function value. Representing the future time domain, Indicates time Grid-connected power; Indicates the grid-connected power reference value; Indicates time The energy storage state of charge; Indicates the reference value for the state of charge of energy storage; Indicates time The air conditioner set temperature offset; and It equals 1.

[0135] The improved objective function is:

[0136] in, This represents the weighting adjustment factor for the grid-connected power smoothing term; Indicates time Grid-connected power; Indicates the grid-connected power reference value; The weighting adjustment factor represents the state of charge of the energy storage. Indicates time The energy storage state of charge; This indicates the reference value for the state of charge of energy storage; This indicates the weighting adjustment factor for the air conditioner set temperature offset; Indicates time The air conditioner set temperature offset; and It equals 1.

[0137] The calculation rule for the weight adjustment factor α is as follows:

[0138] When the risk coefficient R > 0.7 (high risk): =1.5, significantly increasing the weight of grid-connected power smoothing; =0.5, reducing the weight of energy storage SOC recovery and allowing energy storage to participate more deeply in the smoothing process; =0.8, allowing for a certain degree of sacrifice in comfort.

[0139] When 0.3 ≤ risk coefficient R ≤ 0.7 (medium risk): = 1.0 (regular weight); =1.0; =1.0

[0140] When the risk coefficient R < 0.3 (low risk): =0.7, reduce the smoothing weight, and prioritize economy; =1.3, prioritize restoring SOC to the optimal state; =1.2, prioritizing comfort.

[0141] The aforementioned main component of fluctuation can be the dominant component extracted from the target power waveform segment that best represents the core physical characteristics of the current abnormal fluctuation, i.e., the random fluctuation component that needs to be suppressed.

[0142] The aforementioned constraint boundaries can be the upper and lower bounds of the range of values ​​for state variables and control variables in an optimization control model.

[0143] The aforementioned dynamic adjustment can be a process of dynamically adjusting the constraint boundary of the optimized control model based on the principal component of the fluctuation.

[0144] Specifically, the frequency and amplitude of the principal components of the fluctuation can be identified, and the constraint boundaries of the optimized control model can be dynamically adjusted based on the frequency and amplitude of the principal components of the fluctuation.

[0145] Furthermore, the constraints for optimizing the control model must satisfy:

[0146]

[0147] in, Indicates time The minimum value of photovoltaic power generation; Indicates time The photovoltaic power generation capacity; Indicates time The maximum value of photovoltaic power generation.

[0148] Introducing a constraint boundary contraction coefficient β, the adjusted constraint conditions are as follows:

[0149]

[0150] in, Indicates time The minimum value of photovoltaic power generation; Indicates the coefficient of shrinkage; Indicates safety margin; Indicates time The photovoltaic power generation capacity; Indicates time The maximum value of photovoltaic power generation.

[0151] The shrinkage coefficient β is calculated as follows:

[0152]

[0153] in, Indicates the frequency of the principal component of the fluctuation; Indicates the amplitude of the principal component of the fluctuation; Indicates the critical frequency; This represents the rated amplitude. It's understandable that as the fluctuation frequency and amplitude increase, β increases, the constraint boundary contracts more inward, and the optimized control model can adopt a more conservative control strategy, reserving more adjustment margin to cope with uncertainties.

[0154] Optionally, after executing the control instructions for the target building photovoltaic equipment according to the optimal control sequence, the fluctuation suppression rate and control cost can be calculated; based on the fluctuation suppression rate and control cost, reinforcement learning can be performed on the optimized control model.

[0155] In this embodiment of the invention, the aforementioned fluctuation suppression rate can be the percentage decrease in the severity of the network power fluctuation relative to the level before control, obtained by evaluating the suppression effect of the current control after executing the control command. The fluctuation suppression rate is a core indicator for measuring the effectiveness of the current control, and can be between 0 and 1, with a rate closer to 1 being better.

[0156] The volatility suppression rate can be expressed as:

[0157]

[0158] in, Indicates volatility suppression rate; This represents the standard deviation of the grid-connected power before control. This represents the standard deviation of the power output at the grid connection point after control. The standard deviation mentioned above is a statistical indicator that measures the dispersion of a set of data. The larger the standard deviation, the greater the deviation of the power value from the average value, i.e., the more drastic the fluctuation; the smaller the standard deviation, the more concentrated the power is around the average value, i.e., the more gradual the fluctuation.

[0159] The aforementioned control cost can be quantified as the total cost incurred in the current control operation.

[0160] The cost of control can be expressed as:

[0161] in, Indicates the cost of control; Indicates the energy storage charging and discharging power; This indicates the offset of the air conditioner's set temperature. This indicates the cumulative cost. It's understandable that the more energy storage is used and the higher the air conditioning temperature is set, the greater the control costs become.

[0162] Reinforcement Learning (RL) is a machine learning method that learns optimal behavioral policies through the interaction between an agent and its environment. The core objective of reinforcement learning is to maximize cumulative reward. This reinforcement learning can be Q-learning, a classic model-free, value-based algorithm used to learn what action an agent should take in a given state to maximize long-term cumulative reward. The goal of Q-learning is to learn an action-value function. , indicating the state Next action Afterwards, the expected cumulative reward (including future discount rewards) obtained by the intelligent agent.

[0163] It should be noted that the volatility suppression rate and control cost can be used to apply reinforcement learning to the optimized control model. The risk coefficient, volatility characteristics, control parameters, volatility suppression rate, and control cost of the current volatility event can be stored in a historical experience database. Periodically (e.g., every 24 hours), reinforcement learning algorithms can be used to fine-tune the parameters of the optimized control model, thereby continuously optimizing the control strategy and achieving continuous evolution of the system.

[0164] In this embodiment of the invention, by introducing a reinforcement learning mechanism, the coupling parameters can be continuously optimized based on historical parameters, thereby achieving adaptive evolution.

[0165] like Figure 2 As shown, this embodiment of the invention provides a grid fluctuation adaptive response control device for building-integrated photovoltaics (BIPV), which includes:

[0166] The acquisition module 201 is used to acquire the power waveform sequence of the target building photovoltaic system at the grid connection point;

[0167] The processing module 202 is used to perform fluctuation type analysis processing on the power waveform sequence. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence.

[0168] The determination module 203 is used to determine the risk coefficient and the main component of fluctuation based on the target power waveform segment;

[0169] The control module 204 is used to output an optimal control sequence based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, and to execute control commands for the target building photovoltaic equipment according to the optimal control sequence.

[0170] Optionally, the processing module 202 is further configured to perform multi-scale morphological filtering on the power waveform sequence, extract waveform morphological features at different scales, and obtain filtered waveform segments; classify and identify the filtered waveform segments using a preset classification model; when the identification result of the classification model is an abnormal fluctuation type and the confidence level is higher than a preset confidence threshold, mark the fluctuation start time and fluctuation type, and extract the marked waveform data from the filtered waveform segments to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

[0171] Optionally, the determining module 203 is further configured to decompose the target power waveform segment through variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment; determine the fluctuation principal component of the target power waveform segment based on the multiple intrinsic mode components; and determine the risk coefficient of the target power waveform segment in a historical event database according to the fluctuation principal component, wherein the historical event database stores each fluctuation and the power grid impact consequences caused by each fluctuation.

[0172] Optionally, the determining module 203 is further configured to optimize the number of modes and the penalty factor of the variational mode decomposition using the Grey Wolf optimization algorithm to obtain the optimized number of modes and the optimized penalty factor; and based on the optimized number of modes and the optimized penalty factor, to decompose the target power waveform segment to obtain multiple intrinsic mode components of the target power waveform segment.

[0173] Optionally, the determining module 203 is further configured to calculate the multi-scale entropy vector of each intrinsic mode component; and to calculate the Pearson correlation coefficient between each intrinsic mode component and the target power waveform segment; and to select the intrinsic mode component with the highest Pearson correlation coefficient and the largest fluctuation in multi-scale entropy value as the main fluctuation component.

[0174] Optionally, the device is further used to construct an optimized control model, wherein the state variables of the optimized control model include: energy storage state of charge, building indoor temperature, and current grid-connected power; the control variables of the optimized control model include: energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset; the weights of the optimized control model are adjusted according to the risk coefficient; and the constraint boundaries of the optimized control model are dynamically adjusted based on the principal components of the fluctuation.

[0175] Optionally, the device is further configured to calculate the fluctuation suppression rate and the control cost; and to perform reinforcement learning on the optimized control model based on the fluctuation suppression rate and the control cost.

[0176] like Figure 3 As shown, this embodiment of the invention also provides an electronic device, including a processor, which can execute any of the above-described grid fluctuation adaptive response control methods for building photovoltaics.

[0177] Specifically, it includes a processor 301 and a memory 302, as well as a computer program stored in the memory 302 and capable of running on the processor 301, which executes a grid fluctuation adaptive response control method for building-integrated photovoltaics, wherein:

[0178] The processor 301 executes the calculator program for the grid fluctuation adaptive response control method for building-integrated photovoltaics stored in the memory 302, and performs the following steps:

[0179] Obtain the power waveform sequence of the target building photovoltaic system at the grid connection point;

[0180] The power waveform sequence is subjected to fluctuation type analysis. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence.

[0181] Based on the target power waveform segment, the risk coefficient and the main component of the fluctuation are determined;

[0182] Based on the risk coefficient and the principal component of the fluctuation, combined with the preset optimization control model, the optimal control sequence is output, and the control instructions of the target building photovoltaic corresponding equipment are executed according to the optimal control sequence.

[0183] Optionally, the processor 301 performs fluctuation type analysis on the power waveform sequence. If an abnormal fluctuation type is identified, the processor obtains the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence. Specifically, this includes:

[0184] The power waveform sequence is subjected to multi-scale morphological filtering to extract waveform morphological features at different scales, and the filtered waveform segments are obtained.

[0185] The filtered waveform segment is classified and identified using a preset classification model. When the classification model identifies the abnormal fluctuation type and the confidence level is higher than the preset confidence level threshold, the fluctuation start time and fluctuation type are marked, and the marked waveform data is extracted from the filtered waveform segment to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

[0186] Optionally, the step of processor 301 determining the risk coefficient and the principal component of fluctuation based on the target power waveform segment specifically includes:

[0187] The target power waveform segment is decomposed by variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment;

[0188] Based on the multiple intrinsic mode components, the ripple principal component of the target power waveform segment is determined;

[0189] Furthermore, based on the principal component of the fluctuation, the risk coefficient of the target power waveform segment is determined in the historical event database, which stores all previous fluctuations and the power grid impact consequences caused by those fluctuations.

[0190] Optionally, the processor 301 performs the decomposition process of the target power waveform segment through variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment, specifically including:

[0191] The gray wolf optimization algorithm is used to optimize the number of modes and the penalty factor of the variational mode decomposition, resulting in optimized number of modes and optimized penalty factor.

[0192] Based on the optimized number of modes and the optimized penalty factor, the target power waveform segment is decomposed to obtain multiple intrinsic mode components of the target power waveform segment.

[0193] Optionally, the process executed by processor 301 to determine the ripple principal component of the target power waveform segment based on multiple intrinsic mode components specifically includes:

[0194] Calculate the multiscale entropy vector for each of the intrinsic mode components;

[0195] And, calculate the Pearson correlation coefficient between each of the intrinsic mode components and the target power waveform segment;

[0196] The intrinsic mode component with the highest Pearson correlation coefficient and the largest multi-scale entropy fluctuation was selected as the main fluctuation component.

[0197] Optionally, before outputting the optimal control sequence based on the risk coefficient and the principal component of volatility, combined with a preset optimization control model, the method executed by the processor 301 further includes:

[0198] An optimized control model is constructed, wherein the state variables of the optimized control model include: energy storage state of charge, building indoor temperature, and current grid-connected power; the control variables of the optimized control model include: energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset; and the weights of the optimized control model are adjusted according to the risk coefficient.

[0199] Based on the principal component of the fluctuation, the constraint boundary of the optimized control model is dynamically adjusted.

[0200] Optionally, after executing the control instructions for the target building photovoltaic (BPV) equipment according to the optimal control sequence, the method executed by the processor 301 further includes:

[0201] Calculate the volatility suppression rate and control cost;

[0202] The optimized control model is subjected to reinforcement learning based on the fluctuation suppression rate and the control cost.

[0203] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the grid fluctuation adaptive response control method for building photovoltaics provided in this invention, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0204] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles described in this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A grid fluctuation adaptive response control method for building-integrated photovoltaics, characterized in that, Includes the following steps: Obtain the power waveform sequence of the target building photovoltaic system at the grid connection point; The power waveform sequence is subjected to fluctuation type analysis. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence. Based on the target power waveform segment, the risk coefficient and the main component of the fluctuation are determined; Based on the risk coefficient and the principal component of the fluctuation, combined with the preset optimization control model, the optimal control sequence is output, and the control instructions of the target building photovoltaic corresponding equipment are executed according to the optimal control sequence.

2. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to claim 1, characterized in that, The step of performing fluctuation type analysis on the power waveform sequence, and if an abnormal fluctuation type is identified, then obtaining the target power waveform segment corresponding to the abnormal fluctuation type based on the power waveform sequence, specifically includes: The power waveform sequence is subjected to multi-scale morphological filtering to extract waveform morphological features at different scales, and the filtered waveform segments are obtained. The filtered waveform segment is classified and identified using a preset classification model. When the classification model identifies the abnormal fluctuation type and the confidence level is higher than the preset confidence level threshold, the fluctuation start time and fluctuation type are marked, and the marked waveform data is extracted from the filtered waveform segment to obtain the target power waveform segment corresponding to the abnormal fluctuation type.

3. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to claim 1, characterized in that, The step of determining the risk coefficient and the principal component of fluctuation based on the target power waveform segment specifically includes: The target power waveform segment is decomposed by variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment; Based on the multiple intrinsic mode components, the ripple principal component of the target power waveform segment is determined; Furthermore, based on the principal component of the fluctuation, the risk coefficient of the target power waveform segment is determined in the historical event database, which stores all previous fluctuations and the power grid impact consequences caused by those fluctuations.

4. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to claim 3, characterized in that, The step of decomposing the target power waveform segment through variational mode decomposition to obtain multiple intrinsic mode components of the target power waveform segment specifically includes: The gray wolf optimization algorithm is used to optimize the number of modes and the penalty factor of the variational mode decomposition, resulting in optimized number of modes and optimized penalty factor. Based on the optimized number of modes and the optimized penalty factor, the target power waveform segment is decomposed to obtain multiple intrinsic mode components of the target power waveform segment.

5. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to claim 3, characterized in that, The determination of the wave principal component of the target power waveform segment based on multiple intrinsic mode components specifically includes: Calculate the multiscale entropy vector for each of the intrinsic mode components; And, calculate the Pearson correlation coefficient between each of the intrinsic mode components and the target power waveform segment; The intrinsic mode component with the highest Pearson correlation coefficient and the largest multi-scale entropy fluctuation was selected as the main fluctuation component.

6. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to claim 3, characterized in that, Before outputting the optimal control sequence based on the risk coefficient and the principal component of volatility, combined with a preset optimization control model, the method further includes: An optimized control model is constructed, wherein the state variables of the optimized control model include: energy storage state of charge, building indoor temperature, and current grid-connected power; the control variables of the optimized control model include: energy storage charging and discharging power, inverter reactive power, and air conditioning set temperature offset; and the weights of the optimized control model are adjusted according to the risk coefficient. Based on the principal component of the fluctuation, the constraint boundary of the optimized control model is dynamically adjusted.

7. The grid fluctuation adaptive response control method for building-integrated photovoltaics according to any one of claims 1 to 6, characterized in that, After executing the control commands for the target building photovoltaic (BPV) equipment according to the optimal control sequence, the method further includes: Calculate the volatility suppression rate and control costs; The optimized control model is subjected to reinforcement learning based on the fluctuation suppression rate and the control cost.

8. A grid fluctuation adaptive response control device for building-integrated photovoltaics, characterized in that, The grid fluctuation adaptive response control device for building photovoltaics includes: The acquisition module is used to acquire the power waveform sequence of the target building photovoltaic system at the grid connection point; The processing module is used to perform fluctuation type analysis on the power waveform sequence. If an abnormal fluctuation type is identified, the target power waveform segment corresponding to the abnormal fluctuation type is obtained based on the power waveform sequence. The determination module is used to determine the risk coefficient and the main component of fluctuation based on the target power waveform segment; The control module is used to output an optimal control sequence based on the risk coefficient and the principal component of the fluctuation, combined with a preset optimized control model, and execute control commands for the target building photovoltaic equipment according to the optimal control sequence.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the grid fluctuation adaptive response control method for building photovoltaics as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the grid fluctuation adaptive response control method for building photovoltaics as described in any one of claims 1 to 7.

Images