Voltage stability control method under island operation of multi-energy system

Abstract

The invention relates to a voltage stability control method under island operation of a multi-energy system, and the method comprises the steps: generating a voltage prediction value sequence in a target time period; calculating a dynamic response matching index based on the voltage predicted value sequence in combination with the response time of the energy storage system and the conventional unit; in response to the dynamic response matching index exceeding a preset threshold value, triggering a compensation mechanism of the voltage; calculating a compensation capability evaluation factor; generating a power control instruction through a feed-forward compensation path and a feedback compensation path; an upper layer optimization target is embedded into a lower layer real-time control process through a mapping mode from a cost function to a reward function, a mapping relation is corrected according to execution feedback, and a closed-loop optimization operation record is formed; and determining a voltage stability control scheme of the multi-energy system under island operation based on the closed-loop optimization operation record. Therefore, the problem of voltage stability control caused by new energy fluctuation and heterogeneous energy response time difference during island operation of the multi-energy system is solved.

Description

Technical Field

[0001] This application relates to the field of power system automation control technology, and in particular to a voltage stability control method for islanded operation of a multi-energy system. Background Technology

[0002] With the rapid development of new energy power generation technologies, multi-energy systems, with their clean and efficient characteristics, are widely used in industrial parks, remote microgrids, and other scenarios. Multi-energy systems operate in two modes: grid-connected and islanded. In grid-connected mode, the system can maintain voltage and frequency stability through the power support of the main grid. In islanded mode, the system operates independently from the main grid, relying solely on internal distributed power sources, energy storage devices, and load-side regulation to maintain energy balance, significantly increasing the difficulty of voltage stability control.

[0003] In related technologies, voltage control for islanded systems typically employs a simple triggering mechanism based on a fixed threshold, or uses modules such as timing prediction, feedforward-feedback control, and hierarchical optimization. For example, a common approach is to adjust the voltage based on real-time voltage deviation using a PID controller; another approach is to pre-schedule the energy storage system by predicting short-term future voltage.

[0004] However, in related technologies, there are several issues: Response mismatch: Due to the order-of-magnitude difference in response time between energy storage systems (millisecond-level response) and thermal power units (second / minute-level response), existing static threshold methods cannot dynamically adapt to this difference, leading to compensation commands being triggered too early or too late, affecting control real-time performance. Insufficient module coordination: Prediction, compensation, and optimization modules are often designed independently, lacking effective integration, resulting in poor data flow and difficulty in forming a closed-loop control system to cope with complex fluctuations. Weak adaptability: When facing drastic fluctuations in new energy output and complex weather conditions, controllers with fixed parameters and optimization targets cannot adaptively adjust, leading to decreased control performance and difficulty in ensuring system stability, which urgently needs improvement. Summary of the Invention

[0005] This application provides a voltage stability control method for islanded operation of a multi-energy system to address issues in related technologies, such as how to achieve accurate short-term voltage prediction, how to design an adaptive dynamic response matching mechanism to accurately trigger compensation, how to quickly offset voltage deviations through feedforward and feedback collaborative compensation, and how to perform hierarchical collaborative optimization based on the real-time state of the system, ultimately ensuring the voltage stability of the system.

[0006] The first aspect of this application provides a voltage stability control method for a multi-energy system operating in islanded mode, comprising the following steps: acquiring historical voltage data, renewable energy output data, and meteorological data of the multi-energy system in islanded operation, performing time-series feature extraction and feature fusion to generate a voltage prediction value sequence for a target time period; calculating a dynamic response matching index based on the voltage prediction value sequence, combined with the response time of the energy storage system and the response time of the conventional unit, to characterize the degree of response mismatch of heterogeneous energy units; triggering a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path in response to the dynamic response matching index exceeding a preset threshold; after the compensation mechanism is triggered, calculating a compensation capability evaluation factor based on the current state of charge, available output power, and preset compensation power of the energy storage system, and selecting an execution path according to the compensation capability evaluation factor, wherein, in response to the compensation capability evaluation factor satisfying a preset independent compensation condition, the execution path is selected. An independent compensation path for energy storage is established. In response to the compensation capacity assessment factor not meeting the preset independent compensation condition, a linkage signal is generated to initiate a multi-energy unit collaborative compensation path. Under this path, a pre-scheduling instruction is generated through the feedforward compensation path, and a supplementary adjustment instruction is generated through the feedback compensation path. Power control instructions are obtained based on these instructions and the supplementary adjustment instructions. Based on the power control instructions, within a hierarchical collaborative multi-objective rolling optimization framework, a reference power instruction sequence is generated with operating cost and voltage deviation as upper-level optimization objectives. The final power instruction is output in conjunction with the real-time operating status of the equipment. The upper-level optimization objectives are embedded into the lower-level control process through a cost function-to-reward function mapping. The mapping relationship is corrected based on the execution feedback of the final power instruction, forming a closed-loop optimized operation record. Based on the closed-loop optimized operation record, the output ratio of each energy unit in the multi-energy system is adjusted to determine the voltage stability control scheme.

[0007] Optionally, in one embodiment of this application, adjusting the output ratio of each energy unit in the multi-energy system to determine a voltage stability control scheme includes: determining the current state of the multi-energy system based on the compensation capability assessment factor, and determining an initial energy output allocation action in response to an adjustment command associated with the current state and a linkage signal; acquiring voltage prediction value sequence data corresponding to the initial energy output allocation action, and locally correcting the output ratio of the multi-energy system in response to the voltage prediction value sequence data satisfying a preset instability range to obtain adjusted allocation data; determining voltage fluctuation based on the adjusted allocation data, and based on the fluctuation... The system corrects deviation data of the multi-energy system to generate corrected voltage control parameters. Based on the corrected voltage control parameters, it acquires real-time feedback signals and classifies these signals using a pre-established support vector machine model to obtain classified control priorities. It acquires resource scheduling information associated with the classified control priorities and, in response to the resource scheduling information meeting preset conditions for uneven resource allocation, reorganizes the resource scheduling information to determine an optimized execution plan. Based on the optimized execution plan, it acquires operating status data for different time periods and organizes this data to generate the voltage stability control scheme.

[0008] Optionally, in one embodiment of this application, the step of acquiring historical voltage data, renewable energy output data, and meteorological data of a multi-energy system operating in an isolated state, and performing temporal feature extraction and feature fusion to generate a voltage prediction value sequence within a target time period includes: constructing an initial dataset of the multi-energy system based on the historical voltage data, renewable energy output data, and meteorological data, and inputting the initial dataset into a preset multimodal temporal feature fusion prediction network to obtain a comprehensive data record of the multi-energy system; extracting temporal features from the historical voltage data and renewable energy output data based on the comprehensive data record to determine a local spatiotemporal feature set of the multi-energy system; performing feature fusion processing on the weighted information in the local spatiotemporal feature set and the meteorological data to obtain a fused feature vector, and outputting a voltage prediction value sequence within the target time period based on the local spatiotemporal features and the fused feature vector.

[0009] Optionally, in one embodiment of this application, after acquiring historical voltage data, new energy output data, and meteorological data of a multi-energy system operating in an isolated state, performing time-series feature extraction and feature fusion to generate a voltage prediction value sequence within a target time period, the method further includes: in response to the fluctuation range of the voltage prediction value sequence within the target time period exceeding a preset fluctuation range, re-weighting the fused feature vector to obtain an adjusted prediction sequence; based on the adjusted prediction sequence, analyzing the voltage change trend within the target time period to determine the voltage prediction deviation range; generating corresponding prediction correction parameters based on the voltage prediction deviation range and the real-time update status of the meteorological data, and generating the final voltage prediction result based on the prediction correction parameters.

[0010] Optionally, in one embodiment of this application, the step of triggering a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path in response to the dynamic response matching index exceeding a preset threshold includes: determining the dynamic response matching index based on the voltage prediction value sequence; generating a trigger signal and determining the priority classification of the trigger signal in response to the dynamic response matching index being greater than the preset threshold; obtaining the activation condition of the compensation mechanism based on the priority classification of the trigger signal; and triggering the compensation mechanism in response to the compensation mechanism satisfying the activation condition.

[0011] Optionally, in one embodiment of this application, after triggering a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path in response to the dynamic response matching index exceeding a preset threshold, the method further includes: generating an activation command according to the compensation mechanism, obtaining real-time status data of the energy storage system according to the activation command, determining at least one execution parameter of the compensation mechanism based on the real-time status data and the response time of the thermal power unit; adjusting the output power of the energy storage system according to the at least one execution parameter to obtain an adjusted system matching state; and obtaining an updated value of the dynamic response matching index in response to the adjusted system matching state reaching a preset response balance condition, so as to determine the final system operation stability record based on the updated value of the dynamic response matching index.

[0012] Optionally, in one embodiment of this application, the step of generating a pre-scheduling instruction through the feedforward compensation path and generating a supplementary adjustment instruction through the feedback compensation path under the execution path, so as to obtain a power control instruction according to the pre-scheduling instruction and the supplementary adjustment instruction, includes: acquiring the current operating status data of the energy storage system; activating the dual-modal mechanism of the energy storage system in response to the current operating status data and the preset voltage range standard meeting a preset deviation; configuring the parameters of the feedforward compensation path using a pre-established support vector machine model based on the dual-modal mechanism to generate the pre-scheduling instruction, and transmitting the pre-scheduling instruction to the energy storage system to determine the initial voltage adjustment direction; acquiring the real-time response data of the energy storage system based on the initial voltage adjustment direction, and processing the real-time voltage deviation according to the real-time response data and the monitoring information of the feedback compensation path to generate the supplementary adjustment instruction.

[0013] Optionally, in one embodiment of this application, after generating a pre-scheduling instruction through the feedforward compensation path and a supplementary adjustment instruction through the feedback compensation path to obtain a power control instruction based on the pre-scheduling instruction and the supplementary adjustment instruction, the method further includes: in response to the supplementary adjustment instruction satisfying a preset voltage balance condition, acquiring the output state of the energy storage system, and adjusting the control parameters of the feedback compensation path according to the output state to determine at least one voltage correction action; based on the at least one voltage correction action, acquiring updated operating data of the energy storage system, and based on the updated operating data, the feedforward compensation path, and the feedback compensation path, in response to the voltage satisfying a preset standard condition, acquiring the operating log of the dual-mode mechanism; recording the response time and deviation correction data of the energy storage system according to the operating log, and determining the final system operating state according to the response time and the deviation correction data.

[0014] Optionally, in one embodiment of this application, the step of generating a reference power command sequence with operating cost and voltage deviation as upper-level optimization objectives under a hierarchical collaborative multi-objective rolling optimization framework, and outputting a final power command in conjunction with the real-time operating status of the equipment, includes: processing the target rolling cycle data using the hierarchical collaborative multi-objective rolling optimization framework to obtain an initial power demand distribution; constructing a power command sequence generation rule based on the initial power demand distribution, and determining a globally referenced power command sequence in response to the predicted demand corresponding to the generation rule meeting a preset demand threshold; obtaining dynamic update information of the equipment data based on the globally referenced power command sequence, and generating a feedback result of the real-time status of the equipment in response to the equipment meeting a preset available state condition corresponding to the dynamic update information; obtaining the operating constraints of the equipment data based on the feedback result of the real-time status of the equipment, and locally correcting the power command sequence in response to the equipment data not meeting the operating constraints to determine an adjusted power command sequence; and obtaining the load allocation data of the equipment within the target rolling cycle based on the adjusted power command sequence, and generating the final power command in response to the load allocation data meeting the preset allocation balancing condition.

[0015] Optionally, in one embodiment of this application, the step of embedding the upper-level optimization objective into the lower-level control process through a cost function-to-reward function mapping, and correcting the mapping relationship based on the execution feedback of the final power command to form a closed-loop optimization operation record, includes: processing cost function data using a cost function conversion interface to map the cost function data into the input form of the lower-level reward function, and obtaining the reward function definition corresponding to the input form to determine the initial collaborative mapping result; based on the initial collaborative mapping result, obtaining the real-time state execution data of the energy storage system, and in response to the threshold of the real-time state execution data being less than a preset execution threshold, performing a local adjustment on the lower-level reward function to obtain the adjusted reward function data; based on the adjusted reward function data, obtaining the bias of the energy storage system... The system identifies and processes deviation information. If the deviation information exceeds a preset range, a pre-established support vector machine model is used to classify the deviation information to determine the priority sequence for strategy updates. Based on the priority sequence, a parameter set for the upper-level optimization objective is obtained. If the parameter set does not meet a preset matching condition with the current execution state, the parameter set is locally corrected to obtain corrected parameter configuration data. Based on the corrected parameter configuration data, resource scheduling information of the energy storage system is obtained and reallocated to determine an optimized resource scheduling scheme. Based on the optimized resource scheduling scheme, target data for the upper-level and lower-level optimization objectives are obtained and processed to generate the final closed-loop optimization operation record.

[0016] The embodiments of this application have the following beneficial effects: (1) Improve the accuracy and timing of compensation triggering: By introducing a dynamic response matching index based on prediction information, compensation triggering can detect future voltage risks in advance and consider the differences in response of heterogeneous energy sources, thus avoiding the problem of compensation triggering being too early or too late in the traditional static threshold method.

[0017] (2) Reduce system control complexity and operating costs: By using the compensation capacity assessment factor, the dynamic selection between independent energy storage compensation and multi-resource collaborative compensation is realized, reducing unnecessary multi-energy linkage control times and reducing the operating costs and control burden caused by frequent system scheduling.

[0018] (3) Enhance system stability under islanded operation conditions: Through the closed-loop control mechanism of prediction-matching-compensation-evaluation-optimization, the voltage deviation can be quickly suppressed and the system can maintain stable operation under the conditions of new energy output fluctuation and load change.

[0019] (4) Achieving synergistic unity between global optimization objectives and local control objectives: Through the mapping and feedback correction mechanism of upper-level optimization objectives to lower-level control rewards, the system can ensure rapid voltage stabilization while also taking into account operational economy, thereby improving the overall operating efficiency of the multi-energy system.

[0020] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0021] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a voltage stability control method for a multi-energy system operating in islanded mode, according to an embodiment of this application. Figure 2 This is a detailed flowchart of a voltage stability control method for islanded operation of a multi-energy system according to an embodiment of this application; Figure 3 This is a detailed flowchart of a voltage stability control method for islanded operation of a multi-energy system according to another embodiment of this application. Detailed Implementation

[0022] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0023] The following description, with reference to the accompanying drawings, illustrates a voltage stability control method for a multi-energy system operating in islanded mode according to an embodiment of this application. Addressing the issues raised in the background art regarding how to achieve accurate short-term voltage prediction, how to design an adaptive dynamic response matching mechanism for precise compensation triggering, how to quickly offset voltage deviations through feedforward and feedback collaborative compensation, and how to perform hierarchical collaborative optimization based on the real-time system status to ultimately ensure system voltage stability, this application provides a voltage stability control method for a multi-energy system operating in islanded mode. This method utilizes a closed-loop control chain of "prediction-matching-compensation-evaluation-optimization," integrating multimodal data prediction, adaptive dynamic response matching, dual-modal compensation, and hierarchical collaborative optimization to solve the voltage stability control problem caused by differences in the response time of heterogeneous energy sources. It enables high-precision prediction of short-term (e.g., within one hour) voltage fluctuations. When voltage fluctuations occur, it can trigger and execute compensation commands within hundreds of milliseconds, stabilizing the voltage deviation within ±2% of the rated value, and achieving multi-objective optimization of system operating costs and stability. This solves the problems in related technologies, such as how to achieve accurate short-term voltage prediction, how to design an adaptive dynamic response matching mechanism to accurately trigger compensation, how to quickly offset voltage deviations through feedforward and feedback collaborative compensation, and how to perform hierarchical collaborative optimization based on the real-time status of the system to ultimately ensure the voltage stability of the system.

[0024] Before introducing a voltage stability control method for islanded operation of a multi-energy system, let's first introduce the improvements of this application: (i) Prediction-driven dynamic response matching triggering mechanism This application addresses the problem that in the isolated operation scenario of multi-energy systems, the response time of energy storage devices differs by milliseconds and seconds from that of conventional units, making it difficult for existing voltage control methods based on static thresholds to accurately trigger compensation. To address this issue, a predictive-driven dynamic response matching triggering mechanism is proposed.

[0025] Specifically, this application does not directly use voltage prediction results to generate control commands. Instead, based on the predicted future voltage deviation and combined with the actual response times of the energy storage system and conventional generating units, a dynamic response matching index is constructed to quantify the degree of response mismatch between different energy units under future voltage fluctuation conditions. The compensation mechanism is only triggered when the dynamic response matching index exceeds a preset threshold.

[0026] In this way, the present application maps "future voltage risk" to "energy response mismatch degree", which transforms the compensation trigger from the traditional static threshold judgment to a dynamic judgment based on predictive information and heterogeneous energy characteristics, thereby improving the accuracy and timing rationality of the compensation trigger from the control logic level.

[0027] II. Execution path bifurcation control mechanism based on compensation capacity assessment After the compensation mechanism is triggered, this application does not adopt a unified and fixed compensation strategy, but further proposes an execution path forking control mechanism based on compensation capability assessment.

[0028] Specifically, this application constructs a compensation capacity assessment factor to comprehensively evaluate the current state of charge, available power, and power demand to be compensated of the energy storage system, and dynamically selects between the following two execution paths based on the assessment results: (i) When the energy storage system has sufficient compensation capacity, the independent energy storage compensation path shall be executed; (ii) When the energy storage system’s compensation capacity is insufficient, a linkage signal is generated to initiate a multi-resource collaborative optimization compensation path.

[0029] By introducing the aforementioned "compensation capability judgment - execution path bifurcation" decision layer into the control flow, this application avoids the problem of frequently calling multi-energy linkage control in all voltage fluctuation scenarios, reduces system control complexity and operating costs, and at the same time reduces unnecessary system disturbances and improves the overall stability in islanded operation.

[0030] III. A closed-loop mapping mechanism between upper-level optimization goals and lower-level control rewards. To address the differences in time scale and optimization direction between the global operational economic objective and the lower-level real-time voltage control objective in multi-energy systems, this application further proposes a closed-loop mapping mechanism that maps the upper-level optimization objective to the lower-level control reward.

[0031] This application uses a cost function conversion interface to dynamically map cost indicators such as operating costs and voltage deviations in the upper-level multi-objective optimization process into reward function inputs in the lower-level real-time control process. This allows the global optimization objective to be embedded into the lower-level control decision-making process as a reward constraint. Simultaneously, based on the real-time feedback results during the lower-level control execution process, the mapping relationship is continuously corrected, thereby forming a closed-loop collaborative optimization mechanism across time scales.

[0032] Through the aforementioned target mapping and feedback correction mechanism, this application achieves coordination and unity between global economic objectives and local rapid and stable control objectives, avoiding the problem of disconnection between upper and lower level objectives in traditional hierarchical control.

[0033] In summary, this application is not a simple superposition of existing prediction, compensation or optimization methods, but rather focuses on the specific technical contradiction of heterogeneous energy response time mismatch in the isolated operation of multi-energy systems. It constructs a voltage stability control mechanism that is driven by prediction, executed by capability assessment, and achieves cross-layer closed-loop optimization through target mapping. It has essential differences in both control flow structure and trigger logic level.

[0034] Specifically, Figure 1 This is a schematic flowchart illustrating a voltage stability control method for a multi-energy system operating in islanded mode, as provided in an embodiment of this application.

[0035] like Figure 1 As shown, the voltage stability control method for a multi-energy system operating in islanded mode includes the following steps: In step S101, historical voltage data, new energy output data, and meteorological data of the multi-energy system in islanded operation are acquired, and time-series feature extraction and feature fusion are performed to generate a voltage prediction value sequence within the target time period.

[0036] In practical implementation, this application embodiment can collect historical voltage data, new energy output data, wind and solar power data, meteorological data, and equipment status data (such as energy storage status of charge and thermal power unit response time) in real time from the multi-energy system operating environment. The data is then classified, organized, and preprocessed to obtain preliminary fluctuation records. A pre-trained voltage analysis model (based on a support vector machine algorithm) is used to extract key fluctuation parameters, such as fluctuation frequency and amplitude, and to identify the core factors of voltage fluctuations (such as load mutations and wind and solar power fluctuations). If the support vector machine classification results show an abnormal fluctuation trend, control demand points are generated as triggering conditions for subsequent steps.

[0037] The embodiments of this application are accurate and forward-looking: by multimodal fusion prediction, the voltage fluctuation trend can be perceived in advance, and multimodal time series feature fusion prediction can be performed. The multimodal time series feature fusion prediction network receives historical voltage data, wind and solar power data and meteorological data, extracts local spatiotemporal features and fuses weighted meteorological features to obtain a sequence of future short-term voltage prediction values, thereby providing a reference for subsequent predictions.

[0038] Optionally, in one embodiment of this application, historical voltage data, renewable energy output data, and meteorological data of a multi-energy system operating in an isolated state are acquired, and time-series feature extraction and feature fusion are performed to generate a voltage prediction value sequence within a target time period. This includes: constructing an initial dataset of the multi-energy system based on historical voltage data, renewable energy output data, and meteorological data, and inputting the initial dataset into a preset multimodal time-series feature fusion prediction network to obtain a comprehensive data record of the multi-energy system; extracting time-series features from historical voltage data and renewable energy output data based on the comprehensive data record to determine the local spatiotemporal feature set of the multi-energy system; performing feature fusion processing on the weighted information in the local spatiotemporal feature set and meteorological data to obtain a fused feature vector, and outputting a voltage prediction value sequence within the target time period based on the local spatiotemporal features and the fused feature vector.

[0039] It is understood that the local spatiotemporal feature extraction in this application adopts a combination of convolutional neural networks (CNN) and long short-term memory networks (LSTM). The CNN+LSTM can be replaced by a Transformer time series model, a GRU network, etc., and the meteorological feature weighting is based on historical correlation analysis (such as the Pearson correlation coefficient). The fused feature vector is input into a fully connected layer, outputting a sequence of future short-term voltage prediction values ​​(e.g., voltage prediction values ​​every 5 minutes within the next hour). This application can achieve high-precision prediction of future short-term voltage fluctuations. When voltage fluctuations occur, it can trigger and execute compensation instructions within hundreds of milliseconds, stabilizing the voltage deviation within ±2% of the rated value, and achieving multi-objective optimization of system operating costs and stability.

[0040] In this embodiment, historical voltage, wind and solar power, and meteorological data can be obtained from a multi-energy system to construct an initial dataset and obtain a comprehensive data record. Based on the comprehensive data record, a pre-established support vector machine algorithm is used to extract time-series features of historical voltage and wind and solar power to determine a local spatiotemporal feature set. For the local spatiotemporal feature set, feature fusion processing is performed in combination with weighted information in meteorological data to obtain a fused feature vector. Through the fused feature vector, a short-term voltage prediction value sequence is constructed.

[0041] Optionally, in one embodiment of this application, after acquiring historical voltage data, new energy output data, and meteorological data of the multi-energy system in islanded operation, performing time-series feature extraction and feature fusion to generate a voltage prediction value sequence within a target time period, the method further includes: in response to the fluctuation range of the voltage prediction value sequence within the target time period exceeding a preset fluctuation range, re-weighting the fused feature vector to obtain an adjusted prediction sequence; analyzing the voltage change trend within the target time period based on the adjusted prediction sequence to determine the voltage prediction deviation range; generating corresponding prediction correction parameters based on the voltage prediction deviation range and the real-time update of meteorological data, and generating the final voltage prediction result based on the prediction correction parameters.

[0042] In actual implementation, this application embodiment can determine whether the voltage value prediction sequence conforms to the preset fluctuation range. If it exceeds the preset range, the fused feature vector is re-weighted to obtain an adjusted prediction sequence. Based on the adjusted prediction sequence, the short-term voltage change trend is analyzed to obtain dynamic correlation information related to wind and solar power, and the deviation range of voltage prediction is determined. For the deviation range of voltage prediction, combined with the real-time update of meteorological data, corresponding prediction correction parameters are generated to obtain the final voltage prediction result. Based on the final voltage prediction result, the operation status record of the multi-energy system is generated, the input dataset of the prediction model is updated, and the reference basis for subsequent predictions is determined.

[0043] In step S102, based on the application voltage prediction value sequence and combined with the response time of the energy storage system and the response time of the conventional unit, a dynamic response matching index is calculated to characterize the degree of response mismatch of heterogeneous energy units.

[0044] In actual implementation, the embodiments of this application can perform dynamic response matching index calculation and compensation triggering. The dynamic response matching index is calculated based on the voltage prediction value sequence, and the dynamic response matching index characterizes the degree of response mismatch of heterogeneous energy units in the multi-energy system within the target time period. The dynamic response matching index is based on the ratio of the response time of the energy storage system and the thermal power unit. If the dynamic response matching index exceeds a preset threshold, a trigger signal is generated to activate the compensation mechanism.

[0045] The formula for calculating the dynamic response matching index is:

[0046] in, and These are the current response times (in seconds) for thermal power units and energy storage systems, respectively. and The adaptive weighting coefficients are updated online based on historical fluctuation data. This represents the deviation between the predicted voltage value and the rated voltage value.

[0047] It should be noted that the preset threshold is dynamically adjusted according to the system's operating status (for example, the initial threshold value is 3.5, which fluctuates between 2.5 and 4.5 depending on the load). If the dynamic response matching index exceeds the preset threshold, a trigger signal is generated to activate the compensation mechanism; otherwise, monitoring continues.

[0048] This application features adaptability and speed: the dynamic response matching index and adaptive threshold solve the problem of heterogeneous energy response mismatch, and the compensation trigger is more timely and accurate.

[0049] In step S103, in response to the application dynamic response matching index exceeding a preset threshold, a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path is triggered.

[0050] It is understood that the compensation mechanism in this application embodiment is a bimodal adaptive compensation mechanism, including a feedforward compensation path and a feedback compensation path. This application triggers a voltage compensation mechanism in response to the dynamic response matching index exceeding a preset threshold, exhibiting robustness and stability. The bimodal compensation combines the predictive power of feedforward with the error-correcting capability of feedback, effectively offsetting various voltage deviations.

[0051] Furthermore, embodiments of this application can initiate a dual-mode adaptive compensation mechanism. Based on the voltage prediction sequence and trigger signal, the dual-mode adaptive compensation mechanism is initiated. A pre-scheduling command is generated through a feedforward compensation path and sent to the energy storage system to offset the expected voltage deviation. A feedback compensation path processes the real-time voltage deviation and outputs supplementary commands.

[0052] Feedforward compensation path: A predictive model such as ARIMA or a neural network is used, where the ARIMA predictive model can be replaced by other time series predictive models, such as exponential smoothing. Pre-schedule instructions are generated based on the voltage prediction value sequence and sent to the energy storage system to offset the expected voltage deviation. The pre-schedule instructions include the direction and magnitude of power adjustment (e.g., increasing the energy storage output power by 10%).

[0053] Feedback compensation path: Real-time voltage deviation is processed by a PID controller (which can be replaced by a fuzzy controller, sliding mode controller, etc.) or model predictive control (MPC can be replaced by adaptive control, optimal control, etc.), and a supplementary command is output. The outputs of the feedforward and feedback compensation paths are superimposed to form the final compensation command.

[0054] Optionally, in one embodiment of this application, in response to the application dynamic response matching index exceeding a preset threshold, a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path is triggered, including: determining the dynamic response matching index based on the voltage prediction value sequence; in response to the dynamic response matching index being greater than the preset threshold, generating a trigger signal and determining the priority classification of the trigger signal; obtaining the activation condition of the compensation mechanism based on the priority classification of the trigger signal; and triggering the compensation mechanism in response to the compensation mechanism meeting the activation condition.

[0055] Specifically, in this embodiment, the dynamic response index can be calculated using a pre-established support vector machine model based on the voltage prediction value sequence. The index is based on the ratio of the response time of the energy storage system to that of the thermal power unit to obtain a preliminary index result. If the dynamic response index exceeds a preset threshold, a trigger signal is generated for the preliminary index result. The priority classification of the signal is determined. Based on the priority classification of the signal, the activation conditions of the compensation mechanism are obtained. It is determined whether the activation requirements are met. If the conditions are met, the compensation mechanism is triggered.

[0056] Optionally, in one embodiment of this application, after triggering a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path in response to the application's dynamic response matching index exceeding a preset threshold, the method further includes: generating an activation command according to the compensation mechanism, obtaining real-time status data of the energy storage system according to the activation command, determining at least one execution parameter of the compensation mechanism based on the real-time status data and the response time of the thermal power unit; adjusting the output power of the energy storage system according to the at least one execution parameter to obtain an adjusted system matching state; and obtaining an updated value of the dynamic response matching index in response to the adjusted system matching state reaching a preset response balance condition, so as to determine the final system operation stability record based on the updated value of the dynamic response matching index.

[0057] In actual implementation, this embodiment can generate an activation command based on the compensation mechanism. Through the activation command, real-time status data of the energy storage system is obtained. Combined with the response time of the thermal power unit, the specific execution parameters of the compensation mechanism are determined. Based on the execution parameters, the output power of the energy storage system is adjusted, and the adjusted system matching state is obtained. It is determined whether the expected response balance has been achieved. For the adjusted system matching state, the updated value of the dynamic response index is obtained, and the final system operation stability record is determined. In step S104, after the application compensation mechanism is triggered, a compensation capacity assessment factor is calculated based on the current state of charge, available output power, and preset compensation power of the application energy storage system. An execution path is selected based on the application compensation capacity assessment factor. Specifically, if the application compensation capacity assessment factor meets the preset independent compensation conditions, the independent energy storage compensation path is executed. If the application compensation capacity assessment factor does not meet the preset independent compensation conditions, a linkage signal is generated to initiate the multi-energy unit collaborative compensation path.

[0058] In actual implementation, the embodiments of this application can perform compensation capacity assessment and collaborative optimization allocation. The compensation capacity assessment factor is calculated based on the current state of charge of the energy storage, available power, and the power deficit requiring compensation. a) When the independent compensation conditions are met, the independent compensation path for energy storage is executed; b) when the conditions are not met, a linkage signal is generated to initiate a multi-energy unit collaborative compensation path. The calculation formula for the compensation capacity assessment factor is:

[0059] The set level is determined based on the system capacity (e.g., 1.2). If the compensation capability assessment factor is greater than the set level, independent energy storage compensation is performed; otherwise, a linkage signal is generated to initiate collaborative optimization allocation. Collaborative optimization allocation uses a support vector machine model to assign tasks to multiple energy storage units. The objects of collaborative optimization can be expanded from multiple energy storage units to various flexible resources such as energy storage, interruptible loads, and demand-side response resources. To improve the voltage stability of multi-energy islanded systems, more diverse controllable resources are introduced, enhancing the system's regulation capability and reliability. The goal is to minimize the total power deficit and balance the load of each unit. The allocation scheme is dynamically adjusted based on available power distribution and state of charge.

[0060] By introducing the aforementioned compensation capability discrimination and execution path branching decision layer into the control flow, this application avoids the problem of frequently calling multi-energy linkage control in all voltage fluctuation scenarios, reduces system control complexity and operating costs, and at the same time reduces unnecessary system disturbances and improves the overall stability in islanded operation.

[0061] In step S105, under the application execution path, a pre-scheduling instruction is generated through the application feedforward compensation path, and a supplementary adjustment instruction is generated through the application feedback compensation path, so as to obtain a power control instruction based on the application pre-scheduling instruction and the application supplementary adjustment instruction.

[0062] This application combines pre-scheduling commands with supplementary adjustment commands, enabling precise, layered, and time-segmented control of the output of multi-energy systems. First, pre-scheduling commands are generated based on the voltage prediction sequence, allowing for the allocation of controllable resources such as distributed power sources and energy storage devices before the start of the scheduling cycle. This proactively eliminates foreseeable voltage deviations, improving system stability and economy. Second, supplementary adjustment commands are generated based on real-time voltage deviations, enabling rapid correction of prediction errors and voltage fluctuations caused by unforeseen conditions, ensuring the voltage remains within acceptable limits. The combination of pre-scheduling and real-time supplementary adjustment gives the system both long-term planning and rapid real-time adjustment capabilities, effectively improving the voltage control accuracy and robustness of multi-energy systems.

[0063] Optionally, in one embodiment of this application, under the application execution path, a pre-schedule instruction is generated through the application feedforward compensation path, and a supplementary adjustment instruction is generated through the application feedback compensation path, so as to obtain a power control instruction based on the application pre-schedule instruction and the application supplementary adjustment instruction. This includes: acquiring the current operating status data of the energy storage system; activating the dual-mode mechanism of the energy storage system in response to the current operating status data and the preset voltage range standard meeting a preset deviation; configuring the parameters of the feedforward compensation path using a pre-established support vector machine model based on the dual-mode mechanism to generate the pre-schedule instruction, and transmitting the pre-schedule instruction to the energy storage system to determine the initial voltage adjustment direction; acquiring the real-time response data of the energy storage system based on the initial voltage adjustment direction, and processing the real-time voltage deviation according to the real-time response data and the monitoring information of the feedback compensation path to generate the supplementary adjustment instruction.

[0064] In actual implementation, the embodiments of this application can obtain the current operating status data of the energy storage system through the voltage prediction value sequence, and determine whether there is an expected deviation by combining it with the preset voltage range standard. If the expected deviation is detected, a trigger signal is generated to activate the dual-mode mechanism to obtain the preliminary activation basis. Based on the preliminary activation basis, the parameters of the feedforward compensation path are configured using a pre-established support vector machine model according to the priority of the trigger signal, a pre-scheduled instruction is generated and transmitted to the energy storage system to determine the initial voltage adjustment direction. Through the initial voltage adjustment direction, the real-time response data of the energy storage system is obtained. Combined with the monitoring information of the feedback compensation path, the real-time deviation is processed and a supplementary adjustment instruction is generated.

[0065] Optionally, in one embodiment of this application, after generating a pre-scheduling instruction through the application feedforward compensation path and a supplementary adjustment instruction through the application feedback compensation path to obtain a power control instruction based on the application pre-scheduling instruction and the application supplementary adjustment instruction, the method further includes: in response to the supplementary adjustment instruction satisfying a preset voltage balance condition, acquiring the output state of the energy storage system, and adjusting the control parameters of the feedback compensation path according to the output state to determine at least one voltage correction action; based on at least one voltage correction action, acquiring updated operating data of the energy storage system, and based on the updated operating data, the feedforward compensation path, and the feedback compensation path, in response to the voltage satisfying a preset standard condition, acquiring the operating log of the dual-mode mechanism; recording the response time and deviation correction data of the energy storage system according to the operating log, and determining the final system operating state according to the response time and deviation correction data.

[0066] In actual implementation, this application embodiment can determine whether the voltage balance condition is met based on supplementary instructions. If the voltage balance condition is met, the output state of the energy storage system is obtained according to the execution result of the supplementary instructions. For the residual part of the real-time deviation, the control parameters of the feedback compensation path are adjusted to determine a further voltage correction scheme. Through the voltage correction scheme, the updated operating data of the energy storage system is obtained. Combining the synergistic effect of the feedforward compensation path and the feedback compensation path, it is determined whether the voltage adjustment has reached the preset standard. Based on the voltage adjustment determination result, the operation log of the dual-mode mechanism is obtained. For the execution process of adaptive compensation, the response time and deviation correction data of the energy storage system are recorded to determine the final system operating state. In step S106, based on the power control instructions of the application, under the hierarchical collaborative multi-objective rolling optimization framework, a reference power instruction sequence is generated with operating cost and voltage deviation as the upper-level optimization objectives, and the final power instruction is output in combination with the real-time operating status of the equipment.

[0067] It is understood that the target rolling cycle data in this application embodiment can be future rolling cycle data, such as the next 24 hours, with each cycle lasting 15 minutes. Rolling optimization can be performed by re-executing optimization based on the latest data in each control cycle and applying the optimization results to the next cycle, thus continuously advancing in this way.

[0068] In actual implementation, the embodiments of this application can perform hierarchical collaborative multi-objective rolling optimization. The hierarchical collaborative multi-objective rolling optimization framework processes future rolling cycle data. The upper-level global optimization layer solves for the reference power command sequence, while the lower-level real-time control layer receives the reference power command sequence and outputs the final power command based on the real-time status of the equipment.

[0069] Upper-level global optimization layer: This layer aims to minimize operating costs and voltage deviation by solving for the reference power command sequence. The optimization model includes constraints (such as equipment power limits and energy output ratios).

[0070] The lower real-time control layer receives the reference power command sequence, combines it with the real-time status of the equipment (such as the energy storage charge status and the availability of thermal power units), and outputs the final power command through model predictive control. Model predictive control (MPC) is an advanced process control method that predicts the future behavior of the system through a dynamic model and calculates the optimal control command based on an optimization algorithm to obtain the final power command.

[0071] Optionally, in one embodiment of this application, based on the power control command, under a hierarchical collaborative multi-objective rolling optimization framework, a reference power command sequence is generated with operating cost and voltage deviation as the upper-level optimization objectives, and the final power command is output in conjunction with the real-time operating status of the equipment. This includes: processing target rolling cycle data using the hierarchical collaborative multi-objective rolling optimization framework to obtain an initial power demand distribution; constructing a power command sequence generation rule based on the initial power demand distribution, and determining a globally referenced power command sequence in response to the predicted demand corresponding to the generation rule meeting a preset demand threshold; obtaining dynamic update information of the equipment data based on the globally referenced power command sequence, and generating a feedback result of the equipment's real-time status in response to the equipment meeting a preset available state condition corresponding to the dynamic update information; obtaining the operating constraints of the equipment data based on the feedback result of the equipment's real-time status, and locally correcting the power command sequence in response to the equipment data not meeting the operating constraints to determine an adjusted power command sequence; and obtaining the load allocation data of the equipment within the target rolling cycle based on the adjusted power command sequence, and generating a final power command in response to the load allocation data meeting a preset allocation balancing condition.

[0072] As one possible implementation, embodiments of this application can use a hierarchical collaborative framework to process future data within a rolling cycle. A pre-established support vector machine model is used to classify and predict the data, obtaining a preliminary power demand distribution. Based on this preliminary power demand distribution, a power command sequence generation rule is constructed for the global reference target of the upper optimization layer. If the predicted demand exceeds a preset threshold, high-priority device resources are allocated first, determining the global reference power command sequence. Using this global reference power command sequence, the real-time status monitoring of the lower control layer is used to obtain dynamic updates of device data, determine whether the device is in an available state, and obtain real-time status feedback results. Based on the real-time status feedback results, adjustments are made to the final output power command. The system acquires operational constraints on device data. If these constraints are not met, the instruction sequence is locally modified to determine the adjusted power instruction sequence. Using this adjusted power instruction sequence, and considering the balance requirements of multi-objective optimization, the load distribution data of each device within the rolling cycle is acquired to determine if the expected distribution balance has been achieved, resulting in the final power instruction output. Based on the final power instruction output and continuous updates to future data, the predicted input information for the next rolling cycle is acquired to determine whether a new round of optimization is triggered, thus determining the subsequent optimization starting point data. Using this subsequent optimization starting point data, and considering the continuous operation of hierarchical collaboration, the overall resource scheduling information of the system is acquired to determine if the comprehensive requirements of multi-objective optimization are met, resulting in a dynamic adjustment scheme for system operation. In step S107, the upper-level optimization objective is embedded into the lower-level control process through a cost function-to-reward function mapping, and the mapping relationship is corrected based on the execution feedback of the final power instruction, forming a closed-loop optimization operation record.

[0073] In actual implementation, the embodiments of this application can perform cost function transformation and closed-loop optimization. The upper-level optimization objective is converted into a lower-level reward function through a cost function transformation interface, achieving synergy between the two objectives. A feedback loop collects updates to the lower-level control effect strategy and refreshes the upper-level parameters, forming a closed-loop optimization.

[0074] The transformation rules are based on the weighted summation method, for example:

[0075] Among them, weight and Adjust dynamically based on real-time status.

[0076] By collecting the lower-level control effects (such as actual voltage deviation) through feedback loop, updating the strategy and refreshing the upper-level parameters, a closed-loop optimization is formed, and the final closed-loop optimization operation record is generated.

[0077] This application features synergy and economy: the hierarchical optimization framework and cost function conversion interface achieve a balance between global economy and local stability, optimize resource allocation efficiency, and complete a systematic closed loop: the entire scheme, from data acquisition to closed-loop optimization, forms an adaptive intelligent control system, significantly improving the operational reliability of isolated systems.

[0078] Through the aforementioned target mapping and feedback correction mechanism, this invention achieves coordination and unity between global economic objectives and local rapid and stable control objectives, avoiding the problem of disconnection between upper and lower level objectives in traditional hierarchical control.

[0079] Optionally, in one embodiment of this application, the upper-level optimization objective is embedded into the lower-level control process through a cost function-to-reward function mapping, and the mapping relationship is corrected according to the execution feedback of the final power command to form a closed-loop optimization operation record. This includes: processing cost function data using a cost function conversion interface to map the cost function data into the input form of the lower-level reward function, and obtaining the reward function definition corresponding to the input form to determine the initial collaborative mapping result; based on the initial collaborative mapping result, obtaining the real-time state execution data of the energy storage system, and in response to the threshold of the real-time state execution data being less than a preset execution threshold, locally adjusting the lower-level reward function to obtain the adjusted reward function data; and based on the adjusted reward function data, obtaining the energy storage system... The system collects deviation information and, in response to deviation information exceeding a preset range, classifies the deviation information using a pre-established support vector machine model to determine the priority sequence for policy updates. Based on the priority sequence, it obtains the parameter set of the upper-level optimization objective and, in response to the parameter set not meeting a preset matching condition with the current execution state, performs local correction on the parameter set to obtain corrected parameter configuration data. Based on the corrected parameter configuration data, it obtains the resource scheduling information of the energy storage system and performs reallocation processing on the resource scheduling information to determine the optimized resource scheduling scheme. Based on the optimized resource scheduling scheme, it obtains the target data of the upper-level and lower-level optimization objectives and processes the target data to generate the final closed-loop optimization operation record.

[0080] In actual execution, this embodiment can process cost function data through a conversion interface, mapping it to the input form of the lower-level reward function. For the alignment of the upper-level optimization objective and the lower-level execution logic, the corresponding reward function definition is obtained, and the initial collaborative mapping result is determined. Based on the initial collaborative mapping result, real-time status data during execution is obtained for the actual application of the lower-level reward function. This status data is collected through a feedback loop mechanism to determine whether a preset execution threshold has been reached. If the preset threshold has not been reached, the reward function is locally adjusted to obtain the adjusted reward function data. Through the adjusted reward function data, deviation information during execution is obtained for dynamic monitoring of the control effect. If the deviation information exceeds a preset range, an update strategy process is triggered. The deviation information is classified and processed using a pre-established support vector machine model to determine the priority sequence of strategy updates.

[0081] Furthermore, based on the priority sequence of strategy updates and the specific requirements for parameter refresh, the relevant parameter set of the upper-level optimization target is obtained. If the parameter set does not match the current execution state, the parameter set is locally corrected to obtain corrected parameter configuration data. Using the corrected parameter configuration data, for the continuous operation of closed-loop optimization, resource scheduling information of the system under the hierarchical structure is obtained. If the resource scheduling information shows uneven allocation, the scheduling information is reallocated to determine the optimized resource scheduling scheme. Based on the optimized resource scheduling scheme, for the dynamic balance of target collaboration, consistency data of the two-layer targets during execution is obtained. The consistency data is stored and organized through a data collection mechanism to obtain the final closed-loop optimization operation record. In step S108, based on the requested closed-loop optimization operation record, the output ratio of each energy unit in the requested multi-energy system is adjusted to determine the voltage stability control scheme.

[0082] In this embodiment, a voltage stability control scheme can be generated based on the final closed-loop optimized operation record. The energy output ratio is adjusted according to the compensation capability assessment factor and linkage signal, such as increasing the output of thermal power units and reducing the output of energy storage units. By integrating the voltage prediction value sequence and real-time deviation data, a voltage stability control scheme for islanded operation of a multi-energy system is determined.

[0083] This application includes a power output ratio adjustment table, a compensation command sequence, and optimization parameters, and records the operating status through a data storage mechanism for subsequent optimization.

[0084] Optionally, in one embodiment of this application, based on the closed-loop optimization operation record of the application, the output ratio of each energy unit in the multi-energy system is adjusted to determine a voltage stability control scheme, including: determining the current state of the multi-energy system according to the compensation capability assessment factor, and in response to the adjustment command associated with the current state and the linkage signal, determining the initial energy output allocation action; acquiring the voltage prediction value sequence data corresponding to the initial energy output allocation action, and in response to the voltage prediction value sequence data meeting the preset instability range, locally correcting the output ratio of the multi-energy system to obtain the adjusted allocation data; determining the voltage fluctuation based on the adjusted allocation data, and based on the fluctuation... The system dynamically corrects deviation data of the multi-energy system to generate corrected voltage control parameters. Based on the corrected voltage control parameters, it acquires real-time feedback signals and uses a pre-established support vector machine model to classify the feedback signals to obtain classified control priorities. It acquires resource scheduling information associated with the classified control priorities and, in response to resource scheduling information meeting preset conditions for uneven resource allocation, reorganizes the resource scheduling information to determine an optimized execution plan. Based on the optimized execution plan, it acquires and organizes operating status data for different time periods to generate a voltage stability control scheme for islanded operation of the multi-energy system.

[0085] In actual implementation, this embodiment can analyze the current state of the energy system through compensation capability and evaluation factors, obtain adjustment commands related to linkage signals, determine the initial energy output allocation scheme, and, based on the initial energy output allocation scheme, acquire voltage prediction value sequence data for the dynamic changes in the output ratio. It then determines whether the prediction value sequence conforms to a preset stability range. If it does not conform to the preset range, the output ratio is locally corrected to obtain adjusted allocation data. By combining the adjusted allocation data with real-time deviation data, the voltage stability fluctuation is analyzed. If the real-time deviation exceeds a preset threshold, a deviation correction process is triggered to determine the corrected voltage control parameters. Based on the corrected voltage control parameters and to meet the requirements of stable control, real-time feedback signals during system operation are acquired. A pre-established support vector machine model is used to classify these feedback signals, resulting in classified control priorities. Based on these priorities and the execution order of the control scheme, resource scheduling information related to voltage stability is obtained. If the resource scheduling information shows uneven allocation, it is reorganized to determine an optimized execution plan. Based on this optimized plan, continuous monitoring of voltage stability is performed to acquire system operating status data at different times. This data is then processed through a data storage mechanism to obtain the final control scheme record. Specifically, this can be combined with... Figure 2As shown, a specific embodiment is used to illustrate in detail the working principle of a voltage stability control method for islanded operation of a multi-energy system in this application.

[0086] like Figure 2 As shown, embodiments of this application may include the following steps: Step S201: Convert the interface to process the cost function data.

[0087] Step S202: Map to the input of the lower-level reward function.

[0088] Step S203: Obtain the reward function definition.

[0089] Step S204: Determine the initialization co-mapping result.

[0090] Step S205: Obtain real-time status data.

[0091] Step S206: Determine whether the execution threshold has been reached.

[0092] Step S207: If the threshold is not reached, adjust the reward function.

[0093] Step S208: Obtain the adjusted reward function data.

[0094] Furthermore, such as Figure 3 As shown, embodiments of this application may include the following steps: Step S301: Obtain the current status of the energy storage system.

[0095] Step S302: Determine if there is any expected deviation.

[0096] Step S303: Generate a trigger signal to activate the mechanism.

[0097] Step S304: Configure feedforward compensation path parameters.

[0098] Step S305: Generate pre-scheduled instructions.

[0099] Step S306: Obtain real-time response data.

[0100] Step S307: Process the deviation and generate supplementary instructions.

[0101] Step S308: Determine whether the equilibrium condition is met.

[0102] According to the embodiments of this application, a voltage stability control method for a multi-energy system operating in islanded mode is proposed. This method utilizes a closed-loop control chain of "prediction-matching-compensation-evaluation-optimization," integrating multimodal data prediction, adaptive dynamic response matching, bimodal compensation, and hierarchical collaborative optimization to solve the voltage stability control problem caused by the difference in response time of heterogeneous energy sources. It can achieve high-precision prediction of short-term voltage fluctuations (e.g., within one hour). When voltage fluctuations occur, it can trigger and execute compensation commands within hundreds of milliseconds, stabilizing the voltage deviation within ±2% of the rated value, and achieving multi-objective optimization of system operating cost and stability. Therefore, it solves the problems in related technologies, such as how to achieve accurate short-term voltage prediction, how to design an adaptive dynamic response matching mechanism to accurately trigger compensation, how to quickly offset voltage deviations through feedforward and feedback collaborative compensation, and how to perform hierarchical collaborative optimization based on the real-time system status, ultimately ensuring the voltage stability of the system.

[0103] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0104] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0105] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

Claims

1. A voltage stability control method for islanded operation of a multi-energy system, characterized in that, Includes the following steps: Historical voltage data, renewable energy output data, and meteorological data of multi-energy systems operating in an isolated state are acquired, and time-series feature extraction and feature fusion are performed to generate a voltage prediction value sequence for the target time period. Based on the voltage prediction sequence, and combined with the response time of the energy storage system and the response time of the conventional unit, a dynamic response matching index is calculated to characterize the degree of response mismatch of heterogeneous energy units. In response to the dynamic response matching index exceeding a preset threshold, a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path is triggered. After the compensation mechanism is triggered, a compensation capacity evaluation factor is calculated based on the current state of charge, available output power and preset compensation power of the energy storage system, and an execution path is selected according to the compensation capacity evaluation factor. In response to the compensation capacity evaluation factor meeting the preset independent compensation condition, an independent compensation path for energy storage is executed. In response to the compensation capacity evaluation factor not meeting the preset independent compensation condition, a linkage signal is generated to start a multi-energy unit collaborative compensation path. Under the execution path, a pre-scheduling instruction is generated through the feedforward compensation path, and a supplementary adjustment instruction is generated through the feedback compensation path, so as to obtain a power control instruction based on the pre-scheduling instruction and the supplementary adjustment instruction; Based on the power control command, under the hierarchical collaborative multi-objective rolling optimization framework, a reference power command sequence is generated with operating cost and voltage deviation as the upper-level optimization objectives, and the final power command is output in combination with the real-time operating status of the equipment. The upper-level optimization objective is embedded into the lower-level control process through a cost function to a reward function mapping, and the mapping relationship is corrected according to the execution feedback of the final power command to form a closed-loop optimization operation record; Based on the closed-loop optimized operation record, the output ratio of each energy unit in the multi-energy system is adjusted to determine the voltage stability control scheme.

2. The method according to claim 1, characterized in that, The adjustment of the output ratio of each energy unit in the multi-energy system to determine the voltage stability control scheme includes: The current state of the multi-energy system is determined based on the compensation capacity assessment factor, and the initial energy output allocation action is determined in response to the adjustment command associated with the current state and the linkage signal. Obtain the voltage prediction value sequence data corresponding to the initial energy output allocation action, and in response to the voltage prediction value sequence data meeting the preset instability range, locally correct the output ratio of the multi-energy system to obtain the adjusted allocation data; The voltage fluctuation is determined based on the adjusted allocation data, and the deviation data of the multi-energy system is corrected based on the fluctuation to generate corrected voltage control parameters. Based on the corrected voltage control parameters, real-time feedback signals are obtained, and the feedback signals are classified using a pre-established support vector machine model to obtain the classified control priority. Obtain resource scheduling information associated with the classified control priority, and in response to the resource scheduling information meeting the preset conditions for uneven distribution, reorganize the resource scheduling information to determine the optimized execution plan; Based on the optimized execution plan, the operating status data for different time periods are obtained and organized to generate the voltage stability control scheme.

3. The method according to claim 1, characterized in that, The process of acquiring historical voltage data, new energy output data, and meteorological data of the multi-energy system in islanded operation, performing time-series feature extraction and feature fusion to generate a voltage prediction value sequence for the target time period includes: Based on the historical voltage data, the new energy output data, and the meteorological data, an initial dataset for the multi-energy system is constructed, and the initial dataset is input into a preset multimodal time-series feature fusion prediction network to obtain a comprehensive data record of the multi-energy system. Based on the comprehensive data records, time-series features are extracted from the historical voltage data and the new energy output data to determine the local spatiotemporal feature set of the multi-energy system. The local spatiotemporal feature set and the weighted information in the meteorological data are subjected to feature fusion processing to obtain a fused feature vector. Based on the local spatiotemporal features and the fused feature vector, the voltage prediction value sequence within the target time period is output.

4. The method according to claim 3, characterized in that, After acquiring historical voltage data, renewable energy output data, and meteorological data of the multi-energy system in isolated operation, and performing time-series feature extraction and feature fusion to generate a voltage prediction sequence for the target time period, the process also includes: In response to the fluctuation range of the voltage prediction value sequence within the target time period exceeding the preset fluctuation range, the fused feature vector is reweighted to obtain an adjusted prediction sequence. Based on the adjusted prediction sequence, the voltage change trend within the target time period is analyzed to determine the deviation range of the voltage prediction. Based on the deviation range of the voltage prediction and the real-time update of the meteorological data, corresponding prediction correction parameters are generated, and the final voltage prediction result is generated according to the prediction correction parameters.

5. The method according to claim 1, characterized in that, The response that the dynamic response matching index exceeds a preset threshold triggers a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path, including: The dynamic response matching index is determined based on the voltage prediction value sequence; In response to the dynamic response matching index being greater than the preset threshold, a trigger signal is generated, and the priority classification of the trigger signal is determined; Based on the priority classification of the trigger signals, the activation conditions of the compensation mechanism are obtained, and the compensation mechanism is triggered in response to the fulfillment of the activation conditions.

6. The method according to claim 5, characterized in that, After triggering a voltage compensation mechanism including a feedforward compensation path and a feedback compensation path in response to the dynamic response matching index exceeding a preset threshold, the mechanism further includes: The activation command is generated according to the compensation mechanism, and the real-time status data of the energy storage system is obtained according to the activation command. Based on the real-time status data and the response time of the thermal power unit, at least one execution parameter of the compensation mechanism is determined. The output power of the energy storage system is adjusted according to the at least one execution parameter to obtain the adjusted system matching state; In response to the adjusted system matching state reaching the preset response balance condition, the updated value of the dynamic response matching index is obtained, and the final system operation stability record is determined based on the updated value of the dynamic response matching index.

7. The method according to claim 1, characterized in that, In the execution path, a pre-scheduling instruction is generated through the feedforward compensation path, and a supplementary adjustment instruction is generated through the feedback compensation path, so as to obtain a power control instruction based on the pre-scheduling instruction and the supplementary adjustment instruction, including: The current operating status data of the energy storage system is acquired, and the dual-mode mechanism of the energy storage system is activated in response to the current operating status data and the preset voltage range standard meeting a preset deviation. Based on the dual-modal mechanism, a pre-established support vector machine model is used to configure the parameters of the feedforward compensation path to generate the pre-scheduling command, and the pre-scheduling command is transmitted to the energy storage system to determine the initial voltage adjustment direction; Based on the initial voltage adjustment direction, the real-time response data of the energy storage system is acquired, and the real-time voltage deviation is processed according to the real-time response data and the monitoring information of the feedback compensation path to generate the supplementary adjustment command.

8. The method according to claim 7, characterized in that, After generating a pre-scheduling command through the feedforward compensation path and a supplementary adjustment command through the feedback compensation path to obtain a power control command based on the pre-scheduling command and the supplementary adjustment command, the method further includes: In response to the supplementary adjustment command satisfying the preset voltage balance condition, the output state of the energy storage system is obtained, and the control parameters of the feedback compensation path are adjusted according to the output state to determine at least one voltage correction action. Based on the at least one voltage correction action, the updated operating data of the energy storage system is obtained, and based on the updated operating data, the feedforward compensation path, and the feedback compensation path, in response to the voltage meeting the preset standard conditions, the operating log of the dual-mode mechanism is obtained. The response time and deviation correction data of the energy storage system are recorded in the operation log, and the final system operation status is determined based on the response time and the deviation correction data.

9. The method according to claim 1, characterized in that, The process, within the hierarchical collaborative multi-objective rolling optimization framework, involves generating a reference power command sequence with operating cost and voltage deviation as the upper-level optimization objectives, and outputting the final power command in conjunction with the real-time operating status of the equipment. This includes: The target rolling cycle data is processed using a hierarchical collaborative multi-objective rolling optimization framework to obtain the initial power demand distribution; Based on the initial power demand distribution, a power command sequence generation rule is constructed, and in response to the predicted demand corresponding to the generation rule satisfying a preset demand threshold, a global reference power command sequence is determined. Based on the power command sequence of the global reference, obtain dynamic update information of device data, and generate feedback results of the real-time status of the device in response to the device corresponding to the dynamic update information meeting the preset available state conditions. Based on the feedback results of the real-time status of the device, the operating constraints of the device data are obtained, and in response to the situation where the device data does not meet the operating constraints, the power command sequence is locally corrected to determine the adjusted power command sequence. Based on the adjusted power command sequence, the load allocation data of the device within the target rolling cycle is obtained, and in response to the load allocation data satisfying the preset allocation balancing condition, the final power command is generated.

10. The method according to claim 1, characterized in that, The step of embedding the upper-level optimization objective into the lower-level control process through a cost function-to-reward function mapping, and correcting the mapping relationship based on the execution feedback of the final power command to form a closed-loop optimization operation record, includes: The cost function data is processed using a cost function transformation interface to map the cost function data into the input form of the lower-level reward function, and the reward function definition corresponding to the input form is obtained to determine the initial collaborative mapping result. Based on the initial collaborative mapping result, the real-time status execution data of the energy storage system is obtained, and in response to the threshold of the real-time status execution data being less than the preset execution threshold, the lower-level reward function is locally adjusted to obtain the adjusted reward function data. Based on the adjusted reward function data, the deviation information of the energy storage system is obtained, and in response to the deviation information exceeding the preset range, the deviation information is classified and processed by a pre-established support vector machine model to determine the priority sequence of policy updates. Based on the priority sequence of the strategy update, the parameter set of the upper-level optimization target is obtained, and in response to the situation where the parameter set does not meet the preset matching conditions with the current execution state, the parameter set is locally modified to obtain the modified parameter configuration data. Based on the corrected parameter configuration data, the resource scheduling information of the energy storage system is obtained, and the resource scheduling information is reallocated to determine the optimized resource scheduling scheme. Based on the optimized resource scheduling scheme, the target data of the upper-level optimization target and the lower-level optimization target are obtained, and the target data is processed to generate the final closed-loop optimization operation record.

Images