Voltage Predictive Control Method and System Based on Multimodal Disturbance Sensing and Incremental Optimization

Abstract

This invention belongs to the field of power system technology and provides a voltage predictive control method and system based on multimodal disturbance sensing and incremental optimization. The method includes: acquiring multimodal data including meteorological channel data from the renewable energy side and electrical channel data from the power grid side; performing dual-channel disturbance identification on the multimodal data to obtain disturbance probabilities; calculating voltage deviation and voltage deviation change rate based on the electrical channel data from the power grid side; determining whether adaptive regulation is triggered by combining the disturbance probability, voltage deviation, and voltage deviation change rate; calculating the disturbance severity when adaptive regulation is triggered; updating the incremental weight matrix of the model predictive control objective function according to the disturbance severity; constructing and solving a quadratic programming problem for model predictive control based on the incremental weight matrix to obtain the optimal control increment sequence; executing the control increments in the optimal control increment sequence, acquiring new multimodal data until voltage recovery, and completing voltage predictive control based on multimodal disturbance sensing and incremental optimization.

Description

Technical Field

[0001] This invention belongs to the field of power system technology, specifically relating to a voltage prediction control method and system based on multimodal disturbance sensing and incremental optimization. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As the penetration rate of new energy sources in the power system continues to increase, the power grid is gradually shifting from a form dominated by synchronous generators to one with a higher proportion of power electronic converters. Consequently, the proportion of physical inertia in the power system decreases, and the dynamic coupling between the power electronic control system and the power grid strengthens. This makes the system more susceptible to problems such as rapid voltage drops, slow voltage recovery, local oscillations, and interactions between control devices under disturbances such as fluctuations in new energy output and short-circuit faults. Therefore, it is necessary to achieve rapid and smooth voltage support while meeting input, input change rate, and voltage safety constraints.

[0004] Model predictive control (MPC) can solve for control variables in a rolling manner within a finite time domain based on a predictive model and explicitly handle constraints, making it a feasible technical route for voltage control in new power systems. However, existing MPC voltage control methods rely on measurable or predictable disturbance information that can be directly measured or accurately predicted. However, changes in renewable energy sources such as wind speed and solar irradiance are random and time-varying, and electrical faults and fluctuations in renewable energy output may occur in combination, making it difficult for the controller to obtain disturbance intensity information that can be used for online adjustment in a timely manner. Furthermore, the weights in the objective function of traditional MPC mostly use fixed values ​​after offline tuning. Fixed weights may limit the controller's rapid support capability during severe disturbances, while during disturbance dissipation and voltage recovery phases, they may lead to excessive changes in control output and an insufficiently smooth recovery process, making it difficult to balance "rapid support" and "smooth recovery." For control requirements, relying solely on single electrical quantities such as voltage and current for disturbance judgment often requires waiting until changes on the renewable energy side have been transmitted to the grid before obvious characteristics can be obtained, which can easily lead to identification lag or insufficient information. Although meteorological data cannot be used to predict the timing of electrical faults, it can reflect the precursors of wind power and photovoltaic output changes and the operating environment, thus providing supplementary information for disturbance identification under complex operating conditions. Existing disturbance identification and predictive control functions are usually relatively independent, and the output of the disturbance identification model is rarely further converted into an explicit online regularization signal for constraining voltage MPC. Therefore, there is a lack of a closed-loop interface for directly using multimodal disturbance information to adjust the transient regularization of MPC. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a voltage prediction control method and system based on multimodal disturbance sensing and incremental optimization. By constructing an integrated control link of "multimodal sensing—disturbance probability—severity assessment—weighted scheduling—constrained MPC—closed-loop feedback," it solves the technical challenges of insufficient utilization of disturbance information in high-energy-penetration power systems, difficulty in balancing rapid support and smooth recovery with fixed MPC weights, and the inability of disturbance identification results to directly participate in control parameter adjustment.

[0006] According to some embodiments, the first aspect of the present invention provides a voltage prediction control method based on multimodal disturbance sensing and incremental optimization, employing the following technical solution: A voltage predictive control method based on multimodal disturbance sensing and incremental optimization includes: Acquire multimodal data, including meteorological channel data from the new energy sector and electrical channel data from the power grid sector; Dual-channel perturbation identification is performed on the acquired multimodal data to obtain the perturbation probability; Calculate voltage deviation and voltage deviation rate based on the acquired power grid-side electrical channel data; The perturbation probability, voltage deviation, and voltage deviation change rate are combined to determine whether adaptive adjustment is triggered. When adaptive adjustment is triggered, the severity of the perturbation is calculated. The incremental weight matrix of the control objective function is updated based on the obtained disturbance severity. The optimal control increment sequence is obtained by constructing and solving the model predictive control quadratic programming problem based on the updated incremental weight matrix. The control increments in the obtained optimal control increment sequence are executed to acquire new multimodal data until the voltage recovers, forming a rolling closed-loop control and completing voltage predictive control based on multimodal disturbance sensing and incremental optimization.

[0007] As a further technical limitation, the acquired meteorological channel data and electrical channel data are time-aligned and preprocessed to construct meteorological sliding observation windows and electrical sliding observation windows for the current moment, respectively. The meteorological sliding observation windows and electrical sliding observation windows are then input into the meteorological time series feature extraction branch and the electrical time series feature extraction branch, respectively. After extracting the time series features, feature fusion is performed, and the perturbation probability corresponding to the current observation window is obtained through classification.

[0008] Furthermore, both the meteorological time-series feature extraction branch and the electrical time-series feature extraction branch employ a two-layer stacked bidirectional long short-term memory network. The backward long short-term memory network only reverses the samples inside the currently acquired sliding window and does not use future measurement values ​​after the decision time, thus ensuring the causality of the online implementation.

[0009] As a further technical limitation, the trigger value is calculated by combining the disturbance probability, voltage deviation, and voltage deviation change rate; when the trigger value is not zero, adaptive adjustment is triggered; when adaptive adjustment is triggered, the disturbance severity is calculated.

[0010] As a further technical limitation, before updating the incremental weight matrix, an adaptive scaling factor is calculated based on the obtained perturbation severity, and the incremental weight matrix of the model predicts the control objective function is updated based on the obtained adaptive scaling factor and the preset benchmark output incremental weight matrix. As the severity of the disturbance increases, the adaptive scaling factor decreases, reducing the penalty for the predicted output increment; as the voltage gradually recovers, the adaptive scaling factor approaches 1, restoring the output increment penalty to the baseline level and suppressing excessive changes and oscillations during the recovery process.

[0011] As a further technical limitation, the model predicts the objective function of the control quadratic programming problem. for ;in, Used to constrain the tracking error of the predicted voltage to the reference value. Used to limit changes in control input. To adaptively output incremental regularization terms; Indicates the predicted output sequence Y p The increment of (k), i.e. The predicted output increment sequence is represented as ; For reference output sequence; The output tracking weight matrix is ​​fixed. To control the incremental sequence; The control increment weight matrix is ​​fixed. This is the incremental weight matrix.

[0012] According to some embodiments, a second aspect of the present invention provides a voltage prediction control system based on multimodal disturbance sensing and incremental optimization, employing the following technical solution: A voltage predictive control system based on multimodal disturbance sensing and incremental optimization includes: The acquisition module is configured to acquire multimodal data, including meteorological channel data from the new energy side and electrical channel data from the power grid side. The identification module is configured to perform dual-channel perturbation identification on the acquired multimodal data to obtain the perturbation probability; The calculation module is configured to calculate voltage deviation and voltage deviation rate of change based on the acquired power grid-side electrical channel data; The triggering module is configured to determine whether to trigger adaptive adjustment by combining the disturbance probability, voltage deviation and voltage deviation change rate, and to calculate the disturbance severity when adaptive adjustment is triggered. The update module is configured to update the incremental weight matrix of the model-predicted control objective function based on the obtained perturbation severity. The solver module is configured to construct and solve the model predictive control quadratic programming problem based on the updated incremental weight matrix to obtain the optimal control increment sequence. The control module is configured to execute the control increments in the obtained optimal control increment sequence, acquire new multimodal data until the voltage recovers, form a rolling closed-loop control, and complete voltage predictive control based on multimodal disturbance sensing and incremental optimization.

[0013] According to some embodiments, a third aspect of the present invention provides a computer-readable storage medium, employing the following technical solution: A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in the first aspect of the present invention.

[0014] According to some embodiments, the fourth aspect of the present invention provides an electronic device, which adopts the following technical solution: An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in the first aspect of the present invention.

[0015] According to some embodiments, the fifth aspect of the present invention provides a computer program product, which adopts the following technical solution: A computer program product includes software code, wherein the program in the software code performs the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in the first aspect of the present invention.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention transforms the current observation window into a continuous disturbance probability signal through joint timing identification of meteorological and electrical channels. It enhances voltage support during the disturbance phase using multimodal disturbance information and improves control smoothness during the recovery phase by combining adaptive output incremental weights. These two mechanisms synergistically improve overall dynamic performance, allowing the controller to implement more aggressive corrective actions within constraints. When the voltage approaches the reference value, the output incremental weights recover to a stronger regularization level, suppressing unnecessary rapid output changes. The adaptive mechanism only updates the weight matrix online, without increasing the number of decision variables in the MPC. The disturbance probability participates in controller triggering and directly alters the transient regularization of the MPC through disturbance severity, rather than treating the classifier merely as an independent alarm module. This allows the perceived information to have a direct and interpretable effect on control response speed and recovery smoothness. Attached Figure Description

[0017] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0018] Figure 1 This is a flowchart of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization according to Embodiment 1 of the present invention; Figure 2 This is an architecture diagram of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization according to Embodiment 1 of the present invention. Figure 3 This is a detailed schematic diagram illustrating the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization in Embodiment 1 of the present invention. Figure 4 This is a schematic diagram comparing the voltage signals before and after Kalman filtering in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the CSEE-VS receiving-end network topology in Embodiment 1 of the present invention; Figure 6 This is a performance comparison diagram in Embodiment 1 of the present invention, wherein, Figure 6 (a) in the figure is a comparison chart of independent test performance of the threshold method. Figure 6 (b) in the figure is a comparison chart of the independent test performance of LSTM for electrical performance only. Figure 6 (c) in the figure is a comparison of the independent test performance of the dual-channel bidirectional LSTM. Figure 6 (d) in the figure is a performance comparison chart of independent test sets; Figure 7 This is the dynamic response diagram of the controller under a single-phase equivalent disturbance in Embodiment 1 of the present invention, wherein... Figure 7 (a) in the figure is the dynamic response diagram of the minimum voltage monitoring. Figure 7 (b) in the figure is the dynamic response diagram of the power factor at the point of common coupling. Figure 7 (c) in the diagram represents the dynamic response of the total reactive power support. Figure 8 This is a diagram showing the dynamic response of the controller under an equivalent disturbance of a two-phase fault in Embodiment 1 of the present invention. Figure 8 (a) in the figure is the dynamic response diagram of the minimum voltage monitoring. Figure 8 (b) in the figure is the dynamic response diagram of the power factor at the point of common coupling. Figure 8 (c) in the diagram represents the dynamic response of the total reactive power support. Figure 9 This is the controller dynamic response diagram under the equivalent disturbance of a severe three-phase fault in Embodiment 1 of the present invention, wherein... Figure 9 (a) in the figure is the dynamic response diagram of the minimum voltage monitoring. Figure 9 (b) in the figure is the dynamic response diagram of the power factor at the point of common coupling. Figure 9 (c) in the diagram represents the dynamic response of the total reactive power support. Figure 10 This is a structural block diagram of a voltage prediction control system based on multimodal disturbance sensing and incremental optimization, as shown in Embodiment 2 of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0022] In this invention, terms such as "upper," "lower," "left," "right," "front," "back," "vertical," "horizontal," "side," and "bottom" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used only to facilitate the description of the structural relationships of the various components or elements of this invention and do not specifically refer to any component or element in this invention. They should not be construed as limiting the invention.

[0023] In this invention, terms such as "fixed connection," "connected," and "linked" should be interpreted broadly, indicating a fixed connection, an integral connection, or a detachable connection; a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can determine the specific meaning of these terms in this invention based on the specific circumstances, and they should not be construed as limitations on the invention.

[0024] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0025] Example 1 Embodiment 1 of this invention introduces a voltage prediction control method based on multimodal disturbance sensing and incremental optimization.

[0026] like Figure 1 The voltage predictive control method shown includes: (The method is described in the original text, but the provided text is incomplete and cannot be translated accurately.) Acquire multimodal data, including meteorological channel data from the new energy sector and electrical channel data from the power grid sector; Dual-channel perturbation identification is performed on the acquired multimodal data to obtain the perturbation probability; Calculate voltage deviation and voltage deviation rate based on the acquired power grid-side electrical channel data; The perturbation probability, voltage deviation, and voltage deviation change rate are combined to determine whether adaptive adjustment is triggered. When adaptive adjustment is triggered, the severity of the perturbation is calculated. The incremental weight matrix of the control objective function is updated based on the obtained disturbance severity. The optimal control increment sequence is obtained by constructing and solving the model predictive control quadratic programming problem based on the updated incremental weight matrix. The control increments in the obtained optimal control increment sequence are executed to acquire new multimodal data until the voltage recovers, forming a rolling closed-loop control and completing voltage predictive control based on multimodal disturbance sensing and incremental optimization.

[0027] The core technical problem solved in this embodiment is how to integrate meteorological information from the new energy side with electrical information from the power grid side to form online-usable disturbance probability and severity information, and to introduce this information into the output increment weight adjustment process of the constraint MPC objective function in real time, thereby improving the voltage support capability during the disturbance stage and reducing the control fluctuation during the recovery stage without increasing the number of MPC optimization variables and maintaining the convexity of quadratic programming.

[0028] This embodiment employs a voltage prediction control method based on multimodal disturbance sensing and incremental optimization, as follows: Figure 2The architecture shown can be divided into perturbation detection, prediction optimization, and closed-loop execution; specifically: the perception and recognition layer completes multimodal data preprocessing and perturbation recognition and outputs the perturbation probability p. d (k); The decision and optimization layer combines voltage feedback to calculate the severity of the disturbance, update S(k), and solve the constraint MPC; The execution and feedback layer completes the execution of wind power, photovoltaic and energy storage support instructions and returns new measurements; The three layers form a closed loop through "disturbance probability - severity - weight scheduling - MPC - feedback", so that the identification results directly participate in the update of control parameters.

[0029] This embodiment executes the discrete control cycle k in a loop. The sequential relationship, rolling update relationship, and exit judgment between each step are as follows: Figure 3 As shown.

[0030] As one or more implementation methods, this embodiment first acquires multimodal data, namely, collecting meteorological channel data from the new energy side and electrical channel data from the power grid side; in a preferred embodiment of the present invention, the meteorological channel includes solar irradiance G, wind speed v, and ambient temperature T, and the electrical channel includes voltage deviation and fault current I. f .

[0031] As one or more implementation methods, this embodiment performs time alignment and preprocessing on multimodal data; performs time alignment on measurement data from meteorological and electrical channels, and uses the following discrete state-space model to describe the state evolution and observation process of the data between adjacent sampling times: ; ; Where, x k A represents the system state vector at time k; A represents the state transition matrix of the system state over time; B represents the control input matrix, used to characterize the influence of external control variables on the system state; u k w represents the control input vector at time k. k Indicates process noise; z k V represents the observation vector at time k; H represents the observation matrix that maps the system state to the observation space; v k This indicates observation noise.

[0032] After time alignment, Kalman filtering is applied to the measurement data to attenuate high-frequency measurement noise and retain key transient characteristics. The input features are standardized using the mean and standard deviation calculated solely from the training set, and these training set statistics are used for the validation set, test set, and online input data. Subsequently, the processed meteorological and electrical channel data are organized into sliding observation windows. At each decision time, the oldest sample in the current window is deleted, and the most recently acquired sample is added to form the multimodal input sequence corresponding to the current time. A comparison of voltage signals before and after Kalman filtering is provided. Figure 4 As shown.

[0033] As one or more implementation methods, this embodiment uses meteorological window X met (k) and electrical window X ele (k) Input the two temporal feature extraction branches respectively, obtain the temporal representation of each branch, perform feature fusion, and output the normal class probability P through the classification head. N (k) and the probability of disturbance / fault class P d (k), satisfying P N (k)+P d (k)=1.

[0034] Both branches employ a two-layer Bi-LSTM network; this network only reverses the samples already obtained within the sliding window and does not use future measurements after decision time k.

[0035] In this embodiment, the meteorological and electrical channels each employ a two-layer stacked Bi-LSTM. The first layer contains 64 hidden units / directions and outputs the complete timing sequence, while the second layer contains 32 hidden units / directions and outputs the final bidirectional representation. Each branch ultimately forms a 64-dimensional representation, and the two branches are concatenated to form a 128-dimensional fused representation. The fused representation sequentially passes through a 32-unit fully connected layer, a ReLU layer, a Dropout layer, a Dropout probability of 0.20, a 2-unit fully connected layer, and a Softmax layer, outputting the probabilities of the normal class and the disturbance / fault class.

[0036] In this embodiment, each input window contains 20 consecutive samples with a sampling interval of 0.002 s, corresponding to an observation window of 0.040 s. The meteorological input dimension is 20×3, and the electrical input dimension is 20×2. The classification model is trained using the cross-entropy loss function and the Adam optimizer, with an initial learning rate set to 1×10⁻⁶. -3 The mini-batch size is 64, and the maximum number of training epochs is 30. The validation set loss is used as the basis for early stopping. Training is stopped when the validation set loss does not improve for 6 consecutive epochs, and the model parameters are restored to the optimal level corresponding to the validation set loss.

[0037] The disturbance detection module determines the disturbance based on its impact on the voltage and activates the controller after the disturbance determination is completed; the disturbance probability P d (k) Enter the predictive control process through trigger logic and disturbance severity index.

[0038] It should be noted that the output of the disturbance identification in this embodiment is not only used for a binary judgment of "whether a disturbance has occurred", but rather the disturbance probability P is used to determine the disturbance probability P. d (k) serves as a continuous control signal for subsequent controller triggering and online weight adjustment.

[0039] As one or more implementation methods, this embodiment sets the reference voltage to V. ref If the current measured voltage is V(k), then the voltage deviation e v (k) and voltage deviation rate of change r v (k) can be represented as: ; ; Among them, T s This is the controller sampling period. When dealing with multiple monitoring nodes, the maximum deviation, weighted deviation, or vector norm can be selected as the voltage deviation characterization for severity calculation, depending on the control objective.

[0040] This embodiment constructs the trigger quantity χ(k) based on the disturbance probability, voltage deviation, and voltage deviation change rate; that is... ; where p th e th and r th These are the disturbance probability threshold, voltage deviation threshold, and deviation change rate threshold, respectively.

[0041] In this embodiment, the normalized perturbation severity ρ(k) is: ; where α p α v α r The weights are non-negative and α p +α v +α r =1; e ref and r ref For normalized reference values; sat [0,1] This means that the result is limited to the range of 0 to 1.

[0042] In this embodiment, a baseline output increment weight matrix S0 is pre-set. This matrix is ​​preferably a diagonal matrix, where zero elements can be used to preserve the structural characteristics of the output delay or inverse response related channels, and non-zero elements are scaled online in a unified or grouped manner. The adaptive scaling factor η(k) is... ; .

[0043] As the severity of the disturbance increases, η(k) decreases, thereby reducing the penalty on the predicted output increment and allowing the controller to take faster corrective actions. As the voltage gradually recovers, η(k) approaches 1, restoring the output increment penalty to the baseline level to suppress excessive changes and oscillations during the recovery process. S(k) is calculated before solving the MPC in each control cycle and remains unchanged in the current prediction time domain.

[0044] It should be noted that in this embodiment, the severity is obtained by "disturbance probability + voltage deviation + deviation change rate", and the severity is used to directly adjust the incremental regularization term of the MPC output, thereby realizing the closed-loop coupling between the disturbance perception result and the predictive control optimization problem.

[0045] Let the predicted output sequence be Y. p The reference output sequence is W(k), the control increment sequence is ΔU(k), and the predicted output increment is ΔY(k). p Given (k), a fixed output tracking weight matrix Q and a fixed control increment weight matrix R, the objective function is: ;in, Used to constrain the tracking error of the predicted voltage to the reference value. Used to limit changes in control input. To adaptively output incremental regularization terms; Indicates the predicted output sequence Y p The increment of (k), i.e. The predicted output increment sequence is represented as For the i-th predicted output, its output increment in the prediction time domain is expressed as: Where P represents the prediction time domain length, y i (k) represents the actual output value of the i-th output at the current sampling time, y p，i (k+j) represents the i-th predicted output at the current time k to the j-th future sampling time; T2 and T3 are matrices used to construct the difference of the predicted output; ; ;and Therefore, the prediction output increment regularization term is used to penalize rapid changes in the output between adjacent prediction times.

[0046] The prediction model can use a discrete state-space model to construct a lifting prediction equation, so that Y p (k) and ΔY p (k) represents the affine function of the current state and the future control increment sequence.

[0047] The following constraints are applied simultaneously during the optimization process: ; ; ; Since Q is a positive semi-definite matrix, R is a positive definite matrix, and S(k) is a positive semi-definite matrix, the adaptive update will not change the number of decision variables; when the feasible region is non-empty, the optimization problem can remain a strictly convex quadratic programming problem with a unique global optimum.

[0048] This embodiment obtains the optimal control increment sequence ΔU. After (k), only the first control increment is executed, and the control command is sent to the wind power, photovoltaic reactive power support link and the energy storage active / reactive power support link; then new measurements are acquired, the observation window is updated by sliding and the next control cycle is entered until the voltage is restored and the exit condition is met.

[0049] Case Analysis Taking the receiving-end power system with a high proportion of renewable energy as the control object, a three-region, six-state control-oriented equivalent model is established with reference to the receiving-end network topology and power supply configuration of the CSEE-VS voltage stability benchmark system. The three voltage outputs represent the renewable energy generation area, load center, and weak receiving-end area, respectively; the four coordinated control inputs correspond to wind power reactive power support, photovoltaic reactive power support, and battery energy storage active and reactive power support, respectively. All controllers are compared under the same prediction model, initial state, disturbance input, actuator constraints, and simulation settings. The adopted receiving-end network topology is as follows: Figure 5 As shown.

[0050] This embodiment is used for numerical verification at the controller level. It does not explicitly analyze the phase domain switching process, negative sequence / zero sequence components, or hardware implementation factors such as communication delay, asynchronous sampling, processor jitter, converter current limiting, and switching nonlinearity.

[0051] 5000 labeled sequences were generated through 5000 independent simulation runs, including 2500 normal samples and 2500 perturbation / fault samples. These were then stratified into 3500 training sequences, 750 validation sequences, and 750 independent test sequences. Each independent simulation run belonged to only one subset of the data, and standardized statistics were calculated only from the training set and fixed for use in the validation and test sets.

[0052] On the independent test set, the confusion matrix of the dual-channel Bi-LSTM is TN=374, FP=1, FN=18, TP=357, corresponding to an accuracy of 97.47%, a false alarm rate of 0.27%, a precision of 99.72%, a recall of 95.20%, and an F1 score of 97.41%. On the same test set, the accuracy of the threshold method and the electrical quantity-only LSTM is 78.40% and 94.00%, respectively. Since the network structures of the electrical quantity-only LSTM and the dual-channel Bi-LSTM are not completely identical, the above results are used to illustrate the differences in overall model performance and are not used as quantitative causal proof of the meteorological channel contribution. The performance comparison of the three disturbance detection methods on the same independent test set is as follows: Figure 6 As shown.

[0053] The controller sampling period is set to 0.010 s, the prediction time domain Np=30, and the control time domain Nc=2; the stage output tracking weight is Q. y =42I3, the stage control increment weight is RΔu=1.10I4. Standard MPC does not use Bi-LSTM perturbation information and sets the output increment weight to 0; this embodiment introduces a time-varying output increment weight while keeping other parameters the same, i.e. It belongs to the same type of monotonic weighted scheduling idea based on the severity of disturbance as the aforementioned S(k)=η(k)S0: the more severe the disturbance, the weaker the output increment penalty; when the voltage recovers, the output increment penalty gradually increases.

[0054] Three verification conditions were set up, not used for parameter tuning. Controller parameters were tuned using only six development disturbance conditions; the three verification conditions described later were not involved in parameter selection. The final parameters were fixed before verification, and each verification condition was not individually retuned: the equivalent disturbance for a single-phase fault was applied at 1.50–2.00 s; the equivalent disturbance for a two-phase fault was applied at 1.50–2.50 s; and the equivalent disturbance for a severe three-phase fault was applied at 1.70–1.82 s. Each verification condition was simultaneously superimposed with non-stationary wind and solar power output changes caused by time-varying wind speed and solar irradiance trajectories. Therefore, the verification conditions were a composite disturbance of "fault equivalent component + renewable energy output change". Meteorological variables were used to characterize changes on the renewable energy side and were not used to predict the onset time of the electrical fault component.

[0055] (1) Equivalent disturbance condition of single-phase fault.

[0056] The equivalent disturbance of a single-phase fault is applied for 1.50–2.00 s. The dynamic response of each controller under this condition is as follows: Figure 7As shown. Under the same constraints and disturbances, this embodiment can further reduce the voltage deficit during faults compared to the optimized standard MPC. Under this condition, the minimum voltage of the standard MPC is 0.9886 pu, while the complete method of this invention improves it to 0.9972 pu, resulting in a relative reduction of 81.9% in the voltage deficit integral.

[0057] (2) Equivalent disturbance condition of two-phase fault.

[0058] The equivalent disturbance of a two-phase fault is applied for 1.50–2.50 s. The dynamic response of each controller under this condition is as follows: Figure 8 As shown. Under a prolonged two-phase fault equivalent disturbance, this embodiment maintains voltage support within constraints. Under this condition, the minimum voltage of the standard MPC is 0.9902 pu, while the complete method of this invention improves it to 0.9932 pu, resulting in a relative reduction of 34.4% in the voltage deficit integral.

[0059] (3) Equivalent disturbance condition of severe three-phase fault.

[0060] The equivalent disturbance of a severe three-phase fault is applied in the range of 1.70–1.82 s. The dynamic response of each controller under this condition is as follows: Figure 9 As shown. Under the equivalent disturbance of a severe three-phase fault, this embodiment maintains a higher minimum voltage and achieves smooth recovery after the fault is cleared. Under this condition, the minimum voltage of the standard MPC is 0.9884 pu, which is increased to 0.9974 pu in this embodiment, and the voltage deficit integral is reduced by 65.8%.

[0061] Under the same operating conditions, the lowest voltages of the optimized standard MPC in the three operating conditions were 0.9886, 0.9902, and 0.9884 pu, respectively. In this embodiment, the corresponding lowest voltages are increased to 0.9972, 0.9932, and 0.9974 pu, respectively. Compared with the standard MPC, the voltage deficit integral in this embodiment is reduced by 81.9%, 34.4%, and 65.8%, respectively. No input constraint or input rate of change constraint violations occurred in any of the verification operating conditions, and no quadratic programming solution failures occurred. The maximum 95th percentile solution time in each MPC variant was 2.2 ms, with a control sampling period of less than 10 ms.

[0062] The ablation results further show that after adding disturbance information, the average voltage deficit integral of the three verification disturbances decreased from 0.0069 pu·s to 0.0032 pu·s, indicating that multimodal disturbance information is the main source of voltage support improvement. When only adaptive output incremental weights are used, the voltage deficit integral is not significantly reduced, but the reactive power peak and total control variation can be reduced. Under the combination of "disturbance information + adaptive weights", compared with using only disturbance information, the total control variation is further reduced by about 1.4% to 5.1%, while maintaining approximately the same minimum voltage.

[0063] Example 2 Embodiment 2 of the present invention introduces a voltage prediction control system based on multimodal disturbance sensing and incremental optimization.

[0064] like Figure 10 The voltage predictive control system shown includes: (The system is based on multimodal disturbance sensing and incremental optimization.) The acquisition module is configured to acquire multimodal data, including meteorological channel data from the new energy side and electrical channel data from the power grid side. The identification module is configured to perform dual-channel perturbation identification on the acquired multimodal data to obtain the perturbation probability; The calculation module is configured to calculate voltage deviation and voltage deviation rate of change based on the acquired power grid-side electrical channel data; The triggering module is configured to determine whether to trigger adaptive adjustment by combining the disturbance probability, voltage deviation and voltage deviation change rate, and to calculate the disturbance severity when adaptive adjustment is triggered. The update module is configured to update the incremental weight matrix of the model-predicted control objective function based on the obtained perturbation severity. The solver module is configured to construct and solve the model predictive control quadratic programming problem based on the updated incremental weight matrix to obtain the optimal control increment sequence. The control module is configured to execute the control increments in the obtained optimal control increment sequence, acquire new multimodal data until the voltage recovers, form a rolling closed-loop control, and complete voltage predictive control based on multimodal disturbance sensing and incremental optimization.

[0065] The detailed steps are the same as those of the voltage prediction control method based on multimodal disturbance sensing and incremental optimization provided in Example 1, and will not be repeated here.

[0066] Example 3 Embodiment 3 of the present invention provides a computer-readable storage medium.

[0067] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in Embodiment 1 of the present invention.

[0068] The detailed steps are the same as those of the voltage prediction control method based on multimodal disturbance sensing and incremental optimization provided in Example 1, and will not be repeated here.

[0069] Example 4 Embodiment 4 of the present invention provides an electronic device.

[0070] An electronic device includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements the steps in a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in Embodiment 1 of the present invention.

[0071] The detailed steps are the same as those of the voltage prediction control method based on multimodal disturbance sensing and incremental optimization provided in Example 1, and will not be repeated here.

[0072] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0073] A computer program product includes software code, wherein the program in the software code performs the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in Embodiment 1 of the present invention.

[0074] The detailed steps are the same as those of the voltage prediction control method based on multimodal disturbance sensing and incremental optimization provided in Example 1, and will not be repeated here.

[0075] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0076] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0077] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0078] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0079] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0080] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0081] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A voltage predictive control method based on multimodal disturbance sensing and incremental optimization, characterized in that, include: Acquire multimodal data, including meteorological channel data from the new energy sector and electrical channel data from the power grid sector; Dual-channel perturbation identification is performed on the acquired multimodal data to obtain the perturbation probability; Calculate voltage deviation and voltage deviation rate based on the acquired power grid-side electrical channel data; The perturbation probability, voltage deviation, and voltage deviation change rate are combined to determine whether adaptive adjustment is triggered. When adaptive adjustment is triggered, the severity of the perturbation is calculated. The incremental weight matrix of the control objective function is updated based on the obtained disturbance severity. The optimal control increment sequence is obtained by constructing and solving the model predictive control quadratic programming problem based on the updated incremental weight matrix. The control increments in the obtained optimal control increment sequence are executed to acquire new multimodal data until the voltage recovers, forming a rolling closed-loop control and completing voltage predictive control based on multimodal disturbance sensing and incremental optimization.

2. The voltage predictive control method based on multimodal disturbance sensing and incremental optimization as described in claim 1, characterized in that, The acquired meteorological and electrical channel data are time-aligned and preprocessed to construct meteorological and electrical sliding observation windows for the current moment, respectively. The meteorological sliding observation window and the electrical sliding observation window are respectively input into the meteorological time series feature extraction branch and the electrical time series feature extraction branch. After extracting the time series features, feature fusion is performed, and the perturbation probability corresponding to the current observation window is obtained through classification.

3. The voltage prediction control method based on multi-modal disturbance sensing and incremental optimization as described in claim 2, characterized in that, Both the meteorological time-series feature extraction branch and the electrical time-series feature extraction branch adopt a two-layer stacked bidirectional long short-term memory network. The backward long short-term memory network only reverses the samples inside the currently acquired sliding window and does not use future measurement values ​​after the decision time, thus ensuring the causality of the online implementation.

4. The voltage predictive control method based on multimodal disturbance sensing and incremental optimization as described in claim 1, characterized in that, The trigger value is calculated by combining the disturbance probability, voltage deviation, and voltage deviation change rate; when the trigger value is not zero, adaptive adjustment is triggered; when adaptive adjustment is triggered, the disturbance severity is calculated.

5. The voltage predictive control method based on multimodal disturbance sensing and incremental optimization as described in claim 1, characterized in that, Before updating the incremental weight matrix, the adaptive scaling factor is calculated based on the obtained perturbation severity. The incremental weight matrix of the model predicts the control objective function is then updated based on the obtained adaptive scaling factor and the preset benchmark output incremental weight matrix. As the severity of the disturbance increases, the adaptive scaling factor decreases, reducing the penalty for the predicted output increment; as the voltage gradually recovers, the adaptive scaling factor approaches 1, restoring the output increment penalty to the baseline level and suppressing excessive changes and oscillations during the recovery process.

6. The voltage prediction control method based on multi-modal disturbance sensing and incremental optimization as described in claim 1, characterized in that, The objective function of the model predicts and controls the quadratic programming problem. for ;in, Used to constrain the tracking error of the predicted voltage to the reference value. Used to limit changes in control input. To adaptively output incremental regularization terms; Indicates the predicted output sequence Y p The increment of (k), i.e. The predicted output increment sequence is represented as ; For reference output sequence; The output tracking weight matrix is ​​fixed. To control the incremental sequence; The control increment weight matrix is ​​fixed. This is the incremental weight matrix.

7. A voltage predictive control system based on multimodal disturbance sensing and incremental optimization, characterized in that, include: The acquisition module is configured to acquire multimodal data, including meteorological channel data from the new energy side and electrical channel data from the power grid side. The identification module is configured to perform dual-channel perturbation identification on the acquired multimodal data to obtain the perturbation probability; The calculation module is configured to calculate voltage deviation and voltage deviation rate of change based on the acquired power grid-side electrical channel data; The triggering module is configured to determine whether to trigger adaptive adjustment by combining the disturbance probability, voltage deviation and voltage deviation change rate, and to calculate the disturbance severity when adaptive adjustment is triggered. The update module is configured to update the incremental weight matrix of the model-predicted control objective function based on the obtained perturbation severity. The solver module is configured to construct and solve the model predictive control quadratic programming problem based on the updated incremental weight matrix to obtain the optimal control increment sequence. The control module is configured to execute the control increments in the obtained optimal control increment sequence, acquire new multimodal data until the voltage recovers, form a rolling closed-loop control, and complete voltage predictive control based on multimodal disturbance sensing and incremental optimization.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of a voltage predictive control method based on multimodal disturbance sensing and incremental optimization as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in any one of claims 1-6.

10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps of a voltage prediction control method based on multimodal disturbance sensing and incremental optimization as described in any one of claims 1-6.

Images