Day-ahead and intra-day triggering multi-time scale scheduling system and method for micro-energy network

Abstract

The invention discloses a day-ahead and intra-day triggering multi-time-scale scheduling system and method for a micro-energy grid, relates to the technical field of power system scheduling, and mainly aims to solve the problems that in the prior art, due to single-time-scale optimization or power quality single-index management, the robustness of the system to uncertainty is poor, the controllability of power quality is low, and the scheduling efficiency is high. And the multi-agent cooperation efficiency is low. The system comprises a source load data prediction module, a dynamic time-of-use electricity price module, an optimization scheduling module, an electric energy quality monitoring module, a trigger scheduling module and an instruction distribution module. And a multi-time scale framework of'prediction-monitoring-scheduling-execution 'full-chain coordination is constructed, so that the robustness of the system to uncertainty, the controllability of electric energy quality and the high efficiency of multi-agent coordination are improved.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and in particular to a microgrid day-ahead-intraday multi-timescale dispatching system and method. Background Technology

[0002] With the acceleration of the global energy transition, microgrids, as the core carrier integrating distributed energy (such as photovoltaic and wind power), energy storage systems and diverse loads (such as electric vehicle charging and conventional electricity), are becoming increasingly important in improving energy utilization efficiency and promoting the consumption of renewable energy.

[0003] Currently, domestic and international research on microgrid scheduling mainly focuses on single-timescale optimization or single-indicator power quality management, lacking a multi-timescale framework that coordinates the entire chain of "prediction-monitoring-scheduling-execution". For example, existing prediction methods do not fully integrate multi-source data feature decomposition, monitoring technologies do not achieve hierarchical triggering and dynamic monitoring mode switching, and game theory models do not consider the linkage between power quality constraints and dynamic electricity prices. Therefore, there is an urgent need to propose a multi-timescale scheduling method for microgrids that integrates source-load integrated prediction, hierarchical power quality monitoring, multi-agent game theory optimization, and intelligent command allocation, in order to improve the system's robustness to uncertainty, the controllability of power quality, and the efficiency of multi-agent collaboration. Summary of the Invention

[0004] In view of this, the present invention provides a microgrid day-ahead-day triggering multi-timescale scheduling system and method, the main purpose of which is to solve the problems of poor robustness to uncertainty, low controllability of power quality, and low efficiency of multi-entity cooperation caused by single-timescale optimization or single-item power quality management in the prior art.

[0005] According to one aspect of the present invention, a microgrid day-ahead intraday triggering multi-timescale scheduling system is provided, comprising: The source-load data prediction module uses a source-load integrated prediction method with secondary decomposition and feature selection to predict user electricity load, new energy power generation and electric vehicle load, and outputs the prediction results to the dynamic time-of-use pricing module and the optimized scheduling module. The dynamic time-of-use pricing module is used to transmit the prediction results of the source-load data prediction module and the dynamic time-of-use pricing of electric vehicles to the optimization scheduling module. The optimized scheduling module includes a day-ahead optimized scheduling unit and an intraday optimized scheduling unit. The day-ahead optimized scheduling unit performs day-ahead optimized scheduling processing based on the data transmitted by the dynamic time-of-use pricing module to obtain a day-ahead optimized scheduling strategy. The intraday optimized scheduling unit performs intraday optimized scheduling processing based on the day-ahead optimized scheduling strategy obtained by the day-ahead optimized scheduling unit to obtain an intraday optimized scheduling strategy, and transmits the intraday optimized scheduling strategy to the instruction allocation module. The power quality monitoring module analyzes historical monitoring data of power quality indicators of each station to determine the power quality anomaly trigger threshold and classify the quality anomaly trigger level; and uses an edge computing monitoring mode to determine the target quality anomaly trigger level based on the power quality anomaly trigger threshold, and transmits the target quality anomaly trigger level to the trigger scheduling module. The trigger scheduling module generates a corresponding power quality anomaly solution based on the target power quality anomaly trigger level output by the power quality monitoring module, and transmits the power quality anomaly solution to the instruction allocation module. The instruction allocation module generates scheduling instructions by combining the power quality anomaly solution and the intraday optimized scheduling strategy.

[0006] According to another aspect of the present invention, a method for triggering multi-timescale scheduling within a day-ahead period of a microgrid is provided, comprising: The source-load data prediction module in the microgrid is used to predict user electricity load, new energy power generation and charging station load using a source-load integrated prediction method based on quadratic decomposition and feature selection, and the prediction results are obtained. A dynamic time-of-use pricing model for electric vehicles is established based on the prediction results using the dynamic time-of-use pricing module in a microgrid; the dynamic time-of-use pricing model for electric vehicles dynamically adjusts the time-of-use pricing for electric vehicles based on the difference between power supply and power consumption; The power quality monitoring module in the microgrid is used to perform hierarchical monitoring and analysis of the power quality of the microgrid, and hierarchical triggering is performed when power quality is abnormal. The optimization scheduling module and the triggering scheduling module in the microgrid are used to perform daytime optimization scheduling, intraday optimization scheduling, and power quality anomaly handling, resulting in intraday optimization scheduling strategy and power quality anomaly solution. The instruction allocation module in the microgrid is used in conjunction with the power quality anomaly solution and the intraday optimized scheduling strategy to generate scheduling instructions.

[0007] By employing the above-described technical solutions, the technical solutions provided by the embodiments of the present invention have at least the following advantages: This invention provides a microgrid day-ahead and day-intraday triggering multi-timescale scheduling system and method. Compared with existing technologies, this invention constructs a multi-timescale scheduling system that includes a source-load data prediction module, a dynamic time-of-use pricing module, an optimized scheduling module, a power quality monitoring module, a triggering scheduling module, and an instruction allocation module. It integrates technologies such as source-load integrated prediction, power quality graded monitoring, multi-stakeholder game optimization, and intelligent instruction allocation, constructing a multi-timescale framework for full-chain collaboration of "prediction-monitoring-scheduling-execution." This enhances the robustness of microgrids to renewable energy consumption, reduces wind and solar curtailment rates, accelerates the response speed to power quality anomalies, prevents fault escalation, and reduces equipment wear and power outage risks. It also improves the collaborative scheduling efficiency of multiple stakeholders (central energy management platform, DES, EVA, LAS), balancing the interests of all parties. Furthermore, it promotes the engineering application of microgrid multi-timescale intelligent scheduling technology in scenarios involving diverse distributed energy sources and electric vehicle clusters.

[0008] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0009] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This diagram illustrates the structure of a microgrid day-ahead-day-triggered multi-timescale scheduling system according to an embodiment of the present invention. Figure 2 This diagram illustrates the structure of the source load data prediction module in the system provided by an embodiment of the present invention. Figure 3 This diagram illustrates the process of feature selection and nested cross-validation units using a dual data isolation method in the system provided by an embodiment of the present invention. Figure 4 This diagram illustrates the process of optimization solution in the day-ahead optimization scheduling unit of the system provided in this embodiment of the invention. Figure 5 This illustration shows a schematic diagram of the power quality anomaly triggering threshold and triggering level classification process in the system provided by an embodiment of the present invention; Figure 6 A schematic diagram of the instruction allocation optimization solution for an energy storage system provided in an embodiment of the present invention is shown. Detailed Implementation

[0010] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0011] This invention provides a microgrid day-ahead intraday triggering multi-timescale scheduling system, such as... Figure 1 As shown, the system includes: The source-load data prediction module 11 uses a source-load integrated prediction method with secondary decomposition and feature selection to predict user electricity load, new energy power generation and electric vehicle load, and outputs the prediction results to the dynamic time-of-use pricing module 12 and the optimization scheduling module. The dynamic time-of-use pricing module 12 is used to transmit the prediction results of the source-load data prediction module and the dynamic time-of-use pricing of electric vehicles to the optimization scheduling module. The optimized scheduling module includes a day-ahead optimized scheduling unit 13-1 and an intraday optimized scheduling unit 13-2. The day-ahead optimized scheduling unit 13-1 performs day-ahead optimized scheduling processing based on the data transmitted by the dynamic time-of-use pricing module to obtain a day-ahead optimized scheduling strategy. The intraday optimized scheduling unit 13-2 performs intraday optimized scheduling processing based on the day-ahead optimized scheduling strategy obtained by the day-ahead optimized scheduling unit 13-1 to obtain an intraday optimized scheduling strategy, and transmits the intraday optimized scheduling strategy to the instruction allocation module 16. The power quality monitoring module 14 analyzes historical monitoring data of power quality indicators of each station to determine the power quality anomaly trigger threshold and classify the quality anomaly trigger level; and uses an edge computing monitoring mode to determine the target quality anomaly trigger level based on the power quality anomaly trigger threshold, and transmits the target quality anomaly trigger level to the trigger scheduling module 15. The trigger scheduling module 15 generates a corresponding power quality anomaly solution based on the target power quality anomaly trigger level output by the power quality monitoring module, and transmits the power quality anomaly solution to the instruction allocation module 16. The instruction allocation module 16 generates scheduling instructions by combining the power quality anomaly solution and the intraday optimized scheduling strategy.

[0012] In embodiments of the present invention, such as Figure 1As shown, the source-load data prediction module 11 is directly connected to the dynamic time-of-use pricing module 12 and the day-ahead optimization scheduling unit 13-1, outputting prediction results to the aforementioned modules and units; one end of the dynamic time-of-use pricing module 12 receives the prediction results from the source-load data prediction module 11, and the other end outputs the dynamic time-of-use pricing for electric vehicles to the day-ahead optimization scheduling unit 13-1; the power quality monitoring module 14 is connected to the trigger scheduling module 15, outputting the power quality anomaly trigger level to the trigger scheduling module 15; the day-ahead optimization scheduling unit 13-1 receives the prediction results from the source-load data prediction module 11 and the dynamic pricing from the dynamic time-of-use pricing module 12, and outputs the optimal... The day-ahead optimized scheduling strategy is sent to the intraday optimized scheduling unit 13-2; the intraday optimized scheduling unit 13-2 receives the day-ahead optimized scheduling strategy and outputs the optimal intraday optimized scheduling strategy to the instruction allocation module 16; the trigger scheduling module 15 receives the power quality anomaly trigger level from the power quality monitoring module 14 and outputs the power quality anomaly solution to the instruction allocation module 16; the instruction allocation module 16 receives the intraday optimized scheduling strategy and the power quality anomaly solution, and decomposes the scheduling instructions to the individual devices of the distributed energy system (DES), electric vehicle cluster (EVA), conventional load aggregator (LAS), and energy storage system (ESS).

[0013] It should be noted that the connection relationship in the embodiments of the present invention can be a physical connection relationship, or a wireless connection relationship or a wired connection relationship for signal transmission, etc. The embodiments of the present invention do not make specific limitations.

[0014] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to quantify nonlinear complexity, extract multi-scale features, and improve the prediction accuracy of user load, new energy power generation, and charging station load, another microgrid day-ahead intraday triggering multi-timescale scheduling system is provided, such as... Figure 2 As shown, the source load data prediction module includes a secondary decomposition unit, a feature selection unit, a nested cross-validation unit, and an ensemble learning prediction unit; The secondary decomposition unit is used to decompose the time-series load data of user electricity load, new energy power generation and electric vehicle load into multiple intrinsic mode components using CEEMDAN, respectively. The nonlinear complexity is quantified by calculating the permutation entropy of each component, and high-frequency non-stationary components are selected as key modes. Furthermore, the wavelet packet decomposition method is used to perform secondary decomposition on the remaining residual terms to obtain the secondary decomposition results. In embodiments of the present invention, such as Figure 2As shown, the current execution end treats the time-series load data of user load, new energy power generation, and electric vehicle load as nonlinear and non-stationary signals. It uses CEEMDAN (Fully Adaptive Empirical Mode Decomposition of Noise Ensemble) to decompose these signals into multiple intrinsic mode components. The nonlinear complexity is quantified by calculating the permutation entropy of each component, and high-frequency non-stationary components are selected as key modes. Next, the current execution end uses wavelet packet decomposition to perform a secondary decomposition on the remaining residual terms.

[0015] The feature selection unit is used to extract low-frequency features and sudden fluctuation features from the secondary decomposition results, detect energy mutation points based on HHT transform, generate time-varying feature vectors, calculate the HSIC value of each influencing factor, and sort the corresponding influencing factors according to the HSIC value from largest to smallest to obtain a candidate feature set. In embodiments of the present invention, such as Figure 2 As shown, the current execution end extracts low-frequency features and sudden fluctuation features from the secondary decomposition results, and detects energy mutation points based on HHT transformation to generate time-varying feature vectors; the current execution end calculates the HSIC (Hilbert-Schmidt independence criterion) values ​​of influencing factors such as temperature, wind force, ultraviolet intensity, electric vehicle battery power, and charging station electricity price, and arranges the correlation from large to small based on the HSIC values ​​as candidate feature sets.

[0016] The nested cross-validation unit has an outer layer used to divide the 5-fold test set in chronological order, and an inner layer that uses a Bayesian optimizer to synchronously search for parameters on the training set, iteratively eliminates redundant features, and finally outputs a stable feature set and the optimal parameter combination. In embodiments of the present invention, such as Figure 2 As shown, the current execution end is designed with nested cross-validation units: the outer layer divides the test set into 5 folds in chronological order, and the inner layer searches parameters such as HSIC threshold, number of CNN convolution kernels, and learning rate on the training set through a Bayesian optimizer, iteratively eliminates redundant features, and finally outputs a stable feature set and the optimal parameter combination.

[0017] It should be noted that in this embodiment, the feature selection unit and the nested cross-validation unit can also adopt a dual data isolation method to ensure the reliability and generalization of the feature selection results; at the same time, Bayesian optimization is used to achieve automatic parameter tuning, ultimately outputting a stable and reliable optimal feature set and model. Figure 3 As shown, the main steps are as follows: (1) Calculate the HSIC (Hilbert-Schmidt Independence Criterion) values ​​of influencing factors such as temperature, wind force, ultraviolet intensity, electric vehicle battery capacity, and charging station electricity price, and arrange the correlations from largest to smallest as candidate feature sets; (2) Data partitioning: First, the candidate feature set is divided into multiple non-overlapping segments in chronological order. The outer loop uses the last 20% of the data as the test set and the first 80% as the training set, strictly isolating future data to prevent leakage. (3) Inner layer optimization: The training set is split again in chronological order, and a Bayesian optimizer is used to automatically search within the hyperparameter space, including the HSIC relevance threshold, the number of CNN convolutional kernels, and the learning rate. First, dynamic feature selection is performed: each iteration retains strongly correlated features from the candidate feature set based on the current HSIC threshold; second, joint parameter optimization is performed: the CNN parameters and learning rate are adjusted synchronously, and the combined performance is evaluated through the validation set RMSE; finally, probabilistic directional search is performed: based on historical evaluation results, the TPE algorithm predicts the parameter range with higher potential. The optimal parameter combination and the corresponding feature subset are determined through continuous iteration.

[0018] (4) Outer layer validation: Finally, the entire training set is retrained with the parameters, and the generalization performance is evaluated on the test set reserved in the outer layer. After the data segment is split and validated in a loop, the parameter combination with the smallest average RMSE in the 5-fold test is selected. The frequency of the selected features in each fold is counted, stable features are retained, low-frequency features are removed, and the final high robust feature set is generated.

[0019] The integrated learning prediction unit is used to decompose the obtained stable feature set into high-frequency components, mid-frequency components, and low-frequency components, which are modeled by two-layer LSTM / TCN, single-layer LSTM / TCN, and BPNN base learners, respectively, to capture multi-scale features; and integrates the prediction results of each component through the fully connected layer of the meta-learner, and obtains the final prediction result after optimizing the weights.

[0020] In embodiments of the present invention, such as Figure 2 As shown, the current execution end decomposes the obtained stable feature set into high-frequency components, mid-frequency components, and low-frequency components, which are modeled by two-layer LSTM / TCN, single-layer LSTM / TCN, and BPNN base learners, respectively, to capture multi-scale features. Then, the prediction results of each component are integrated through the fully connected layer of the meta-learner, the weights are optimized, and the final prediction result is output. The prediction effect is evaluated using two evaluation metrics, RMSE and MAE.

[0021] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to balance the benefits of multiple stakeholders during the day-ahead scheduling and forecasting phase, another microgrid day-ahead intraday triggered multi-timescale scheduling system is provided. The day-ahead optimized scheduling unit in the optimized scheduling module is also used for: The central energy management platform is taken as the leader, and a day-ahead optimization scheduling model for the leader is established with the goal of maximizing the daily operating revenue of the leader. Taking the distributed energy system as the first follower, and establishing a day-ahead optimization scheduling model for the first follower with the goal of maximizing the daily operating revenue of the first follower; The electric vehicle cluster is regarded as the second follower, and the day-ahead optimization scheduling model of the second follower is established with the optimization objective of maximizing the sum of the consumer surplus of the second follower and the revenue of electric vehicle users. A day-ahead optimization scheduling model for the third follower is established by taking the regular load aggregator as the third follower and maximizing the total consumer surplus of the third follower as the optimization objective. By combining the leader's day-ahead optimization scheduling model, the first follower's day-ahead optimization scheduling model, the second follower's day-ahead optimization scheduling model, and the third follower's day-ahead optimization scheduling model, a master-slave Stackelberg game model is established, so that the day-ahead optimization scheduling strategy after game equilibrium is generated based on the master-slave Stackelberg game model.

[0022] In this embodiment of the invention, the day-ahead optimization scheduling unit constructs a leader-follower model based on Stackelberg game theory. The central energy management platform plays the role of the leader. Distributed Energy Systems (DES), Electric Vehicle Clusters (EVA), and Conventional Load Aggregators (LAS) are the followers. Through this game process, each participant gradually finds its optimal strategy, leading the system to an equilibrium state. The objective functions of the various stakeholders in the microgrid are as follows: (1) The leader establishes a day-ahead optimization scheduling model with the goal of maximizing his own daily operating revenue:

[0023] in, , These are the revenues from the central energy management platform selling electricity and heat to LAS and EVA, respectively. It is the energy supply shortage compensation paid by DES when it fails to provide sufficient energy to the central government on time; , , The cost of purchasing energy from DES, EVA, and the Ess energy storage system by the central energy management platform; V i,t , V rated These represent the voltage and rated voltage of node i, respectively; , Costs include voltage deviation and network loss. The cost of penalizing the abandonment of wind and solar power; , for tThe power and unit cost coefficient corresponding to the revenue of each part of the central period.

[0024] (2) The first follower distributed energy system (DES) establishes a day-ahead optimization scheduling model with the goal of maximizing daily operating revenue:

[0025]

[0026] in, , , , They are respectively t Fuel costs, operation and maintenance costs, average daily investment costs, and cost of penalties for wind and solar curtailment of DES equipment during the period; a , b , c , d The cost coefficient for GT; The pollution cost coefficient per unit of electricity generated is GT.

[0027] (3) The second follower electric vehicle cluster EVA establishes a day-ahead optimization scheduling model with the optimization objective of maximizing the sum of consumer surplus and EV user revenue. Consumer surplus is defined as the difference between the utility function and the payment cost. The utility function represents the total satisfaction that consumers obtain when consuming a good or service, while the payment cost is the total amount that consumers actually pay to obtain the good or service.

[0028]

[0029]

[0030] in, Let be the utility function of EVA; Cost of EV battery degradation; w ev , v ev This represents the preference coefficient for electric vehicle users' electricity consumption. L et 、E s 、C bat These are the battery charge / discharge cycle life, battery capacity, and battery purchase cost.

[0031] (4) The third follower, the regular load aggregator LAS, establishes a day-ahead optimization scheduling model with the goal of maximizing the total consumer surplus of users.

[0032]

[0033] The utility function of LAS is:

[0034] In this multi-stakeholder game model, the central energy management platform sets energy prices based on supply, demand, and market information. DES, EVA, and LAS optimize the prices offered by the central entity. Their optimization results, in turn, influence the central entity's pricing decisions. The decisions of these four stakeholders are not only continuous but also mutually influential. This energy trading process conforms to the dynamic game characteristics of a hierarchical master-slave structure.

[0035] With the central energy management platform as the leader and DES, EVA, and LAS as followers, representing different stakeholders, a master-follower Stackelberg game model is established, as shown below.

[0036]

[0037] in, P , S , F These are the three elements of a game theory model: participants, strategies, and payoffs. In this embodiment, the participants... P This includes the Central Energy Management Platform, DES, EVA, and LAS. The strategies for the Central Energy Management Platform, DES, EVA, and LAS are represented as follows:

[0038] Benefits of the Central Energy Management Platform, DES, EVA, and LAS Let be the objective function. When the central player formulates a strategy, DES, EVA, and LAS respond optimally to the central player's strategy. Once the central player accepts these responses, the game reaches equilibrium. In the Stackelberg equilibrium, no player can unilaterally change their strategy to gain a greater payoff.

[0039] It should be noted that the central energy management platform, acting as the leader in the optimized dispatch unit, proactively regulates the charging and discharging behavior of electric vehicles based on the real-time supply-demand difference through a dynamic time-of-use pricing module: when wind power (WT) and photovoltaic (PV) output are sufficient, a dynamic time-of-use pricing is set based on the system power supply-demand difference to guide electric vehicles to charge; when renewable energy generation is insufficient, the same mechanism is used to guide electric vehicles to discharge. The specific implementation method is as follows: Step 1: Power Shortage Judgment Criteria

[0040] in, , , , , , These represent the predicted power deficit, base load power, EV disordered charging load, wind power (WT) output, photovoltaic (PV) output, and gas turbine (GT) output for time period t, respectively.

[0041] Step 2: Dynamic Electricity Price Trigger Mechanism

[0042] in, This represents the difference between the power deficit and the maximum power reserve of electric vehicles during time period t. This represents the maximum dispatchable power of the electric vehicle; s is the electric vehicle reserve power coefficient, which is the proportion of the maximum charging capacity of electric vehicles that can be used as reserve capacity to the total charging capacity during this period. At times, the power supply is less than the power consumption, indicating insufficient power supply to the system.

[0043] when At this time, the power supply exceeds the power consumption, and the system does not experience power shortage.

[0044] in, , This indicates the dynamic charging and discharging price of the EV during time period t; , This indicates the EV's regular charging and discharging price during time period t.

[0045] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to combine the global optimization capability of the differential evolution algorithm with the efficient search mechanism of quantum computing to solve the day-ahead optimization model, another microgrid day-ahead intraday triggered multi-timescale scheduling system is provided, such as... Figure 4 As shown, the day-ahead optimization scheduling strategy generated based on the master-slave Stackelberg game model after generating the game equilibrium in the day-ahead optimization scheduling unit includes: The central energy management platform generates an electricity price strategy set S based on the energy supply and demand forecast of the entire network; and encodes multiple pricing schemes through quantum superposition states. The distributed energy system initializes the corresponding strategy set. S DES Based on the central government's energy purchase price, the output of each unit is adjusted through the quantum rotating door to optimize energy production costs and energy sales revenue. The electric vehicle cluster initializes the corresponding strategy set SEVA; and dynamically adjusts the charging power and discharging power of electric vehicles according to the central charging and discharging price, combining quantum cross-balancing user demand and grid constraints. The strategy set corresponding to the initialization of the regular load aggregator. S LAS Furthermore, the quantum Grover algorithm is used to search for the optimal load combination and minimize the energy cost under the central electricity price. The central energy management platform receives the optimization results from the distributed energy system, the electric vehicle cluster, and the conventional load aggregator, generates a hyperlocation strategy pool through quantum entanglement fusion, and uses quantum measurement to screen for Pareto optimal solutions to obtain the day-ahead optimized scheduling strategy.

[0046] In embodiments of the present invention, such as Figure 4 As shown, the specific process of optimization in the current optimization scheduling unit is as follows: (1) Population initialization and parameter setting. The central energy management platform generates a set of electricity pricing strategies based on the energy supply and demand forecast of the entire network. S Multiple pricing schemes are encoded through quantum superposition states; simultaneously, DES, EVA, and LAS initialize their respective policy sets. S DES , S EVA , S LAS .

[0047] (2) Follower adaptive mutation and strategy correction. DES: Adjustment is made through the quantum rotating gate based on the central energy purchase price. Photovoltaic power output Wind power output, Gas turbine output optimizes energy production costs and energy sales revenue; EVA: dynamically adjusted based on central charging and discharging prices. Combining quantum cross-balancing user needs with grid constraints; LAS: using the quantum Grover algorithm to search for the optimal load combination. Minimize central electricity sales price The energy cost is reduced.

[0048] (3) Quantum superposition and global policy optimization. The central energy management platform receives the optimization results from the three parties and generates a hyperposition policy pool through quantum entanglement fusion. The Pareto optimal solution is screened using quantum measurement.

[0049] (4) Differential evolution operation. Crossover: The DES wind power, photovoltaic power output and EVA charging power are quantum entangled and crossover to improve the renewable energy absorption rate; Mutation: Quantum noise perturbation is applied to the load parameters of LAS to avoid local optima; Selection: Based on the multi-objective Pareto front, the objective functions of the central energy management platform, DES, EVA and LAS are balanced.

[0050] (5) Terminate the iteration when the maximum number of iterations is reached or the policy fitness converges, and output the optimal policy.

[0051] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to meet the needs of day-ahead global optimization and intraday real-time adjustment and improve the effectiveness of instruction allocation, another microgrid day-ahead intraday triggered multi-timescale scheduling system is provided. The intraday optimization scheduling unit in the optimization scheduling module is also used for: The load adjustment cost is determined based on the adjustment amount of the power of each unit in the day-ahead optimized scheduling strategy. The heat load adjustment cost is determined based on the adjustment amount of the heat power of each unit in the day-ahead optimized scheduling strategy. The emergency operation costs of energy storage and electric vehicles are determined based on the emergency charging and discharging power of energy storage devices and the emergency charging and discharging power of electric vehicle clusters, respectively. The cost of reserve capacity is determined by a hybrid model that combines deterministic basic reservation with dynamic adjustments based on historical events. The minimum value of the sum of the electrical load adjustment cost, the thermal load adjustment cost, the energy storage emergency operation cost, the electric vehicle emergency operation cost, and the reserve capacity reservation cost is determined as the objective function for intraday optimal scheduling.

[0052] In this embodiment of the invention, the step of determining the reserve capacity reservation cost through a hybrid model combining deterministic basic reservation with dynamic correction based on historical events includes: A deterministic basic reserved model is constructed based on the shortest duration of independent load support by energy storage devices and electric vehicle clusters, and the peak load power of microgrids; Based on the number of extreme events occurring within the statistical period and the actual energy used by the charging station within the statistical period, the deterministic basic reserved model is dynamically corrected to obtain the historical event dynamic correction model. The reserve cost of the standby capacity is determined based on the difference between the output of the historical event dynamic correction model and the minimum state of charge of the energy storage under normal operating conditions.

[0053] In this embodiment of the invention, during the intraday phase, a rolling time-domain method is adopted, with the objective function being to minimize the adjustment amount relative to the previous day's scheduling instructions. Simultaneously, to avoid triggering emergency scheduling operations that accelerate equipment aging and incur high costs, capacity is reserved for the triggering phase during the intraday phase. The model is as follows:

[0054] in, This indicates the cost of adjusting the electrical load. This indicates the cost of adjusting the heat load. This indicates the cost of emergency operations for energy storage. This indicates the emergency operation cost of EVA. This indicates the cost reserved for spare capacity.

[0055]

[0056] in, ΔP e,DES ΔP e,EVA ΔP e,LAS ΔP e,Ess They represent DES, EVA The adjustment amount of the electric power of LAS and Ess; , , , This indicates the power of each unit at the current day stage.

[0057]

[0058] in, ΔP h,DES ΔP h,EVA ΔP h,LAS ΔP h,Ess They represent DES, EVA Adjustments to the thermal power of LAS and Ess.

[0059]

[0060] in, This indicates the purchase cost of energy storage equipment; , Indicates the standard cycle number of energy storage devices and t Cyclic energy over a period of time; This indicates an aging-accelerating factor, typically measured at 1.5 to 2.0. express t Time period EssEmergency charging and discharging power; This represents the emergency loss coefficient.

[0061]

[0062] in, , These represent the standard number of cycles for the EVA device and t Cyclic energy over a period of time; Indicates factors that accelerate aging; α Indicates the power influence coefficient; β It represents the harmonic influence coefficient, reflecting the amplification effect of harmonic distortion (THD) on losses; THD represents the total harmonic distortion. This represents the emergency charging and discharging power of the EVA during time period t; This represents the emergency loss coefficient.

[0063] This invention employs a hybrid model that combines deterministic basic reservation with dynamic adjustments based on historical events to determine the reserve capacity reservation cost. .

[0064] A. Deterministic Foundation Reserved Model:

[0065] in, Indicates the peak load power of the microgrid; Indicates the island's running time ( h This refers to the minimum duration for which Ess and EVA need to independently support the load, set according to local extreme weather or fault repair records; , This indicates the overall efficiency and total capacity of the energy storage system; , This represents the overall efficiency and total schedulable capacity of EVA.

[0066] B. Dynamic correction model based on historical events

[0067] in, This represents the adjustment factor for reserve capacity in the event of a single extreme event. This indicates the number of extreme events that occurred within the statistical period; β This represents the correction factor for the charging station's capacity utilization rate; This indicates the actual energy used by the charging station within the statistical period.

[0068] Reserve costs for spare capacity Represented as:

[0069] in, This represents the unit cost coefficient for reserve capacity; This indicates the minimum state of charge of energy storage under normal operating conditions, which is usually the minimum limit for safe operation of the equipment.

[0070] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to perform hierarchical power quality monitoring and abnormal triggering for different triggering objects, another microgrid day-ahead intraday triggering multi-timescale scheduling system is provided, such as... Figure 5 As shown, the system's power quality monitoring module analyzes historical monitoring data of power quality indicators from each station to determine the power quality anomaly trigger threshold and classify the anomaly trigger levels, including: Collect historical monitoring data of power quality indicators for each power station and integrate the historical monitoring data of each indicator into an independent indicator dataset; The K-means clustering method based on the nearest neighbor density matrix is ​​used to cluster the index datasets corresponding to each power quality index, and k-class index clustering results corresponding to each power quality index are obtained. The clustering results of the k-type indicators are determined as k-type trigger objects, and the k-type trigger objects are divided into different quality anomaly trigger levels according to the degree of anomaly; Based on the clustering results of the k-type indicators, the average value of the indicators of the same type of triggering objects is calculated to obtain the trigger threshold of the k-type triggering objects.

[0071] In this embodiment of the invention, the current execution end uses the K-means clustering method based on the nearest neighbor density matrix to cluster historical monitoring data of power quality indicators, determine the power quality anomaly trigger threshold and the division of trigger levels, such as... Figure 5 As shown, the main steps include: Step 1: Data Preprocessing. Collect four steady-state power quality indicators (V... THD V D F D V UN The historical monitoring data, each indicator constitutes an independent one-dimensional dataset, denoted as the total harmonic distortion rate of voltage as an example:

[0072] Data standardization: To eliminate differences in the order of magnitude and range of values ​​between different features, the original numbers need to be normalized.

[0073] in, Indicates the first i The first sample j Item indicator value.

[0074] Step 2: Perform the following process independently for each power quality indicator using the K-means clustering method based on the nearest neighbor density matrix: (1) Construction KD Tree: Because the data is one-dimensional, KD The tree degenerates into a balanced binary tree. The dataset is recursively split by the median, and a tree structure is constructed to accelerate nearest neighbor search.

[0075] (2) Calculate the nearest neighbor density matrix: set the number of nearest neighbors ( n (for data volume), based on KD Tree for fast lookup of each data point k Nearest neighbors; calculate the number of shared nearest neighbors m ij (data points) x i and x j (The number of common nearest neighbors) is used to generate an n×n nearest neighbor similarity matrix. M； For matrix M Sum the rows to obtain the nearest neighbor density for each data point. t i This forms the density matrix T.

[0076] Step 3: Determine the power quality anomaly trigger threshold. Assume that after clustering... k Cluster C 1, C 2,..., C k , No. k Set the number of triggering objects of the class to N k ,in, Then the trigger threshold u of the k-th class in the j-th indicator. jk for:

[0077] in, x ij This represents the value of the j-th metric of the i-th triggering object.

[0078] Step 4: Classifying Quality Anomaly Trigger Levels. The setting of trigger thresholds largely depends on the triggering object; for different triggering objects, the trigger thresholds for each level will vary. M They vary. According to... M The relative relationship with the national standard limit is used to divide the trigger threshold into different levels for different trigger objects, and the trigger levels are classified into Level 1 green trigger (the best case), Level 2 orange trigger, Level 3 yellow trigger, and Level 4 red trigger.

[0079] In this embodiment of the invention, the power quality monitoring technology constructs a three-layer structure of "microgrid side - substation side - user side," and designs three edge computing monitoring modes: light, normal, and intensive (based on the RK3588 edge terminal for local analysis). It uses "nearest neighbor density matrix K-means clustering" to classify four trigger levels (Level 1 green - monitoring only, Level 2 orange - light regulation, Level 3 yellow - moderate regulation, Level 4 red - emergency regulation). Specifically, lightweight monitoring is implemented for substations with good operating conditions and excellent power quality, reducing monitoring indicators, lowering information frequency, and increasing data analysis intervals; centralized monitoring is implemented for substations with poor operating conditions and substandard power quality. Full indicator monitoring is performed on all data items, increasing information frequency and shortening data analysis intervals. Based on the edge computing framework, the indicator calculation interval, storage frequency, and number of monitoring indicators are configured via configuration files to achieve light, normal, and intensive monitoring of microgrid power quality.

[0080] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to quickly restore system safety when power quality anomalies occur, while minimizing emergency control costs, another microgrid day-ahead-day triggering multi-timescale scheduling system is provided. In this system, the triggering scheduling module is also used for: Obtain the duration of each power quality indicator exceeding the limit, as well as the penalty coefficient for different trigger levels, and determine the cost of power quality limit exceeding penalty. Obtain the emergency operation cost of each energy storage device and the emergency operation cost of each electric vehicle in the electric vehicle cluster to determine the equipment adjustment cost; When the local resources of a microgrid cannot quickly restore power quality, in order to avoid purchasing electricity from the grid at a high price due to power shortage, the emergency power purchase cost is calculated. When the output of new energy sources increases suddenly, in order to prevent equipment overload, it is necessary to reduce the output of wind and solar power and calculate the cost of curtailment penalties. Based on the power quality violation penalty cost, the equipment adjustment cost, the emergency power purchase cost, and the power abandonment penalty cost, a trigger scheduling objective function is constructed; An improved Black-winged Kite optimization algorithm is used to solve the triggering scheduling objective function to obtain the power quality anomaly solution.

[0081] In this embodiment of the invention, the core objective of triggered scheduling is to quickly restore system safety when power quality anomalies occur, while minimizing emergency control costs. At this time, Ess and EVA schedule energy storage units and EVs that have a faster response and are controllable. The specific triggered scheduling objective function is as follows:

[0082] in, Penalty coefficients for different trigger levels; For the first k The duration of exceeding the limit for a power quality indicator; , For equipment i The adjustment costs and adjustment amounts; , This includes the costs of emergency power purchases and penalties for power curtailment.

[0083] (1) Penalty cost for exceeding power quality limits:

[0084] (2) Equipment adjustment costs:

[0085] (3) Emergency power purchase cost: When local resources such as Ess and EVA cannot quickly restore power quality, electricity must be purchased from the grid at a high price to avoid voltage collapse or frequency overruns due to power shortages.

[0086]

[0087] in, For real-time electricity prices; Power to be purchased in an emergency.

[0088] (4) Cost of power curtailment penalties: When a sudden increase in renewable energy output leads to excessive harmonics or line overload, in order to prevent equipment overload, wind and solar power output needs to be reduced, but penalties must be paid or green electricity revenue must be lost.

[0089]

[0090] Different limits are set for the rate of change of energy storage charging and discharging power, the response speed of electric vehicles, and the power adjustment amount of both, depending on the trigger level. The red trigger, which is the most critical, sets the fastest rate of change of energy storage charging and discharging power and the fastest response speed of electric vehicles, and the largest power adjustment amount for both; conversely, the green trigger, which is the first level, only monitors and does not require scheduling.

[0091] It should be noted that in this embodiment, the command allocation model of the energy storage system is solved using an improved Blackwing Kite optimization algorithm, such as... Figure 6 As shown, the steps are as follows: Step 1: Data Acquisition and Parameter Initialization. Real-time acquisition of power grid dispatch commands and energy storage unit operating parameters (such as SOC and power limits) is performed, and the scale of energy storage units requiring regulation is dynamically calculated.

[0092] Step 2: Initial Solution Set Construction. A diverse candidate solution set is generated using a reverse knowledge fusion strategy, covering the multi-dimensional solution space. The reverse knowledge fusion strategy generates opposing inverse solutions from the current solution, then compares the fitness values ​​of the two solutions to select the optimal solution for iteration. This enhances the diversity of the population during the search process and further prevents the algorithm from getting trapped in local optima. The steps are as follows: A) Initial solution set construction: Randomly generate candidate solution sets within the solution space boundary. ,in , i For individual serial numbers, j For dimensional indexing.

[0093] B) Generation of mirror solutions: Generating mirror solution sets through geometric symmetric mapping. X By utilizing the symmetry of the solution space, we can explore potential optimal regions and improve global search capabilities. Its location is calculated as follows: .

[0094] C) Filtering the mixed solution set and merging the original solution sets. X With mirror solution set X ′, forming a scale of 2 N Mixed solution set Fitness is calculated based on the objective function before screening. N The optimal solution is used as the initial population.

[0095] D) Adaptive weight adjustment, introducing dynamic weight factors. According to the iteration progress t Adjust the influence of mirror solutions, focusing on global exploration in the early stages and local development in the later stages.

[0096]

[0097] Step 3: Performance Evaluation and Global Tracking. Quantify the comprehensive performance indicators of each candidate solution (including economy, stability, and SOC balance), and update the global optimal solution and its dynamic tracking path.

[0098] Step 4: Employ an elite solution-guided mechanism to generate perturbation directions based on the current optimal solution, expanding the search range. Compare the quality of solutions in the attack and tracking states, and select non-dominated solutions for the next generation candidate pool. The improved position update formula for the Black-winged Kite optimization algorithm is:

[0099] in, , For the first i Only black-winged kites were present. j dimension t Subsequent t Position in +1 iteration step; ybest The optimal position; r It is a random number generated between 0 and 1; p It is a constant, with a value of 0.9; n is a nonlinear factor; C(0,1) is the Cauchy variation.

[0100] Step 5: Perform cross-regional migration operations (simulating bird migration behavior) within the solution space, and select the globally optimal allocation scheme through Pareto front.

[0101] Step 6: When the convergence threshold is met or the preset iteration limit is reached, output the optimal power allocation value of each energy storage unit; otherwise, return to step 3 to continue the optimization loop.

[0102] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, in order to efficiently and accurately allocate the total power demand of charging stations to individual electric vehicles, another microgrid day-ahead-day-intraday triggering multi-timescale scheduling system is provided. The system's optimization scheduling module also includes a microgrid and electric vehicle dual-layer control unit, which is used for: Establish fast charging rights accounts for electric vehicles participating in demand-side response to record corresponding fast charging rights points; Based on the charging characteristics of electric vehicles, the charging habits of car owners, the incomplete charging capacity, and the fast charging rights integral, an electric vehicle charging model is constructed with the goal of minimizing the charging cost of electric vehicles. The microgrid's scheduling model is constructed with the objective function of maximizing the revenue generated from electricity sales minus the equivalent cost of incomplete charging, and then minus the equivalent cost related to fast charging rights points. The electric vehicle charging model and the microgrid scheduling model constitute a two-layer optimization model; and the particle swarm optimization algorithm is used to solve the two-layer optimization model to obtain the corresponding power output of the electric vehicles participating in the demand-side response.

[0103] In this embodiment of the invention, the current execution end constructs a two-layer control model of microgrid and electric vehicle based on the fast charging power incentive mechanism, which efficiently and accurately allocates the total power demand of the charging station to individual electric vehicles.

[0104] (a) Electric vehicle charging model The model takes into account the physical characteristics of EVs, owners' charging habits, and incomplete charging capacity. Factors such as the charging cost of electric vehicles are considered, with the objective function being the minimum charging cost.

[0105] in, This refers to charging time; SOC when not fully charged; , This indicates the target SOC and the SOC at the end of charging; For the battery capacity of electric vehicles; The total fast charging power points obtained for EVs; N In The sensitivity coefficient of car owners to the right to fast charging; The frequency with which the car owner previously participated in DR; weighting coefficient and It is inversely proportional and is determined by an empirical function.

[0106] (II) Microgrid scheduling model Microgrids aim to maximize revenue from electricity sales. They adjust their total revenue by subtracting the equivalent costs of incomplete charging and the equivalent costs associated with fast-charging power points awarded as rewards. Microgrids can further adjust the weighting factors for fast-charging power point costs and incomplete charging. and To optimize the scheduling strategy. Its scheduling model is as follows:

[0107]

[0108] in, This refers to all electric vehicles; , express t The actual power of the charging station at any time and the power demand of the microgrid on the charging station; This indicates the allowable error range for the charging station's power output.

[0109] The interaction between the two optimization models mentioned above involves, during the preparation phase, building a fast-charging power database, where each user has a fast-charging power account to record their fast-charging power points. Furthermore, [the following will be implemented / implemented]. and S En ( t+ 1) Initialization. In formal operation, the first step is to determine whether EVs are needed for demand response on the microgrid. If not, then... and S En ( t+ 1) Set to 0. In this case, the electric vehicle has no charging limitations and cannot discharge to the grid. If the EV is required to participate in demand-side response (DR), the total charging power of the EV needs to meet the power demand of the microgrid. The scheduling layer determines the total power demand for the next time step. Based on the current state of the EV, determine and publish the current moment. and S En (t+ 1). Each EV is based on the published... and S En ( t+ 1) Determine the charging power at the current moment based on your own charging needs. At the end of the current moment, the state of the electric vehicle is updated, and this state information is passed to the scheduling layer. Then, the process proceeds to the next moment.

[0110] Finally, the particle swarm optimization algorithm is used to solve this two-level optimization model to obtain the electric vehicles that EVA participates in scheduling and their corresponding power output.

[0111] It should be noted that there are no restrictions on the charging power of electric vehicles when there is no DR (Delivery Direction) from the microgrid. During microgrid DR, the behavior of all electric vehicles will be divided into three categories: forced fast charging mode, weak response mode, and strong response mode. When the microgrid enters DR, all electric vehicles should transition to weak response mode, in which they begin charging at a slow charging power. In weak response mode, the electricity price remains at the baseline level, equivalent to the electricity price without the DR mechanism. (See equation:)

[0112] in, for t Moment EV Electricity price; Under normal power grid conditions EV Electricity prices. Under this model, the user's fast charging credits remain unchanged.

[0113] When a car owner urgently needs to use their vehicle and requires a short period of fast charging, they can choose to enter forced fast charging mode. In this mode, users can obtain fast charging privileges in two different ways. The first method consumes the user's fast charging privilege points, allowing them to enjoy fast charging privileges at a lower electricity price. In this mode, the user's fast charging privilege points begin to decrease, as shown in the formula:

[0114] in, express t The fast charging power points earned by the EV.

[0115] Users without fast charging credits can obtain higher charging power by paying a higher electricity price. The relationship between electricity price and charging power is as follows:

[0116] in, for t Electricity price coefficient at any given time; , express t time EV The charging power and maximum charging power in weak response mode. When the charging power of an electric vehicle exceeds the maximum charging power, the electricity price increases linearly with the excess charge.

[0117] In strong response mode, users can also earn fast charging credits by discharging electricity into the grid. In this mode, the grid charges credits based on electricity prices. The system recovers the user's discharge power and rewards the user with fast charging power points based on that power. The relationship between the reward strength of fast charging power points and the user's discharge power is shown in the following formula:

[0118] in, S En ( t )for t The incentive coefficient at any given time; , S En ( t This is determined by the time-sharing mechanism on the microgrid side. and The relationship is as follows:

[0119] in, express t The remaining fast charging credits for EV users.

[0120] This invention provides a microgrid day-ahead, day-intraday triggering multi-timescale scheduling system. Compared with existing technologies, this invention constructs a multi-timescale scheduling system that includes a source-load data prediction module, a dynamic time-of-use pricing module, an optimized scheduling module, a power quality monitoring module, a triggering scheduling module, and an instruction allocation module. It integrates technologies such as source-load integrated prediction, power quality graded monitoring, multi-stakeholder game optimization, and intelligent instruction allocation, constructing a multi-timescale framework for full-chain collaboration of "prediction-monitoring-scheduling-execution." This enhances the robustness of microgrids to renewable energy consumption, reduces wind and solar curtailment rates, accelerates the response speed to power quality anomalies, prevents fault escalation, and reduces equipment wear and power outage risks. It also improves the collaborative scheduling efficiency of multiple stakeholders (central energy management platform, DES, EVA, LAS), balancing the interests of all parties. Furthermore, it promotes the engineering application of microgrid multi-timescale intelligent scheduling technology in scenarios involving diverse distributed energy sources and electric vehicle clusters.

[0121] As a response to the above Figure 1 To implement the system shown, this embodiment of the invention provides a method for multi-timescale scheduling triggered day-ahead and day-ahead in a microgrid. The method includes: The source-load data prediction module in the microgrid is used to predict user electricity load, new energy power generation and charging station load using a source-load integrated prediction method based on quadratic decomposition and feature selection, and the prediction results are obtained. A dynamic time-of-use pricing model for electric vehicles is established based on the prediction results using the dynamic time-of-use pricing module in a microgrid; the dynamic time-of-use pricing model for electric vehicles dynamically adjusts the time-of-use pricing for electric vehicles based on the difference between power supply and power consumption; The power quality monitoring module in the microgrid is used to perform hierarchical monitoring and analysis of the power quality of the microgrid, and hierarchical triggering is performed when power quality is abnormal. The optimization scheduling module and the triggering scheduling module in the microgrid are used to perform daytime optimization scheduling, intraday optimization scheduling, and power quality anomaly handling, resulting in intraday optimization scheduling strategy and power quality anomaly solution. The instruction allocation module in the microgrid is used in conjunction with the power quality anomaly solution and the intraday optimized scheduling strategy to generate scheduling instructions.

[0122] This invention provides a multi-timescale scheduling method for microgrids triggered within the day-ahead period. Compared with existing technologies, this invention constructs a multi-timescale scheduling system that includes a source-load data prediction module, a dynamic time-of-use pricing module, an optimized scheduling module, a power quality monitoring module, a triggered scheduling module, and an instruction allocation module. It integrates technologies such as source-load integrated prediction, power quality graded monitoring, multi-stakeholder game optimization, and intelligent instruction allocation, constructing a multi-timescale framework for full-chain collaboration of "prediction-monitoring-scheduling-execution." This enhances the robustness of microgrids to renewable energy consumption, reduces wind and solar curtailment rates, accelerates the response speed to power quality anomalies, prevents fault escalation, and reduces equipment wear and power outage risks. It also improves the collaborative scheduling efficiency of multiple stakeholders (central energy management platform, DES, EVA, LAS), balancing the interests of all parties. Furthermore, it promotes the engineering application of multi-timescale intelligent scheduling technology for microgrids in scenarios involving diverse distributed energy sources and electric vehicle clusters.

[0123] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0124] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A microgrid day-ahead-intraday multi-timescale scheduling system, characterized in that, include: The source-load data prediction module uses a source-load integrated prediction method with secondary decomposition and feature selection to predict user electricity load, new energy power generation and electric vehicle load, and outputs the prediction results to the dynamic time-of-use pricing module and the optimized scheduling module. The dynamic time-of-use pricing module is used to transmit the prediction results of the source-load data prediction module and the dynamic time-of-use pricing of electric vehicles to the optimization scheduling module. The optimized scheduling module includes a day-ahead optimized scheduling unit and an intraday optimized scheduling unit; the day-ahead optimized scheduling unit performs day-ahead optimized scheduling processing based on the data transmitted by the dynamic time-of-use pricing module to obtain the day-ahead optimized scheduling strategy; The intraday optimization scheduling unit performs intraday optimization scheduling processing based on the day-ahead optimization scheduling strategy obtained by the day-ahead optimization scheduling unit to obtain the intraday optimization scheduling strategy, and transmits the intraday optimization scheduling strategy to the instruction allocation module; The power quality monitoring module analyzes historical monitoring data of power quality indicators at each site to determine the threshold for triggering power quality anomalies and classify the trigger levels of quality anomalies. The monitoring mode employs edge computing to determine the target quality anomaly trigger level based on the power quality anomaly trigger threshold, and transmits the target quality anomaly trigger level to the trigger scheduling module. The trigger scheduling module generates a corresponding power quality anomaly solution based on the target power quality anomaly trigger level output by the power quality monitoring module, and transmits the power quality anomaly solution to the instruction allocation module. The instruction allocation module generates scheduling instructions by combining the power quality anomaly solution and the intraday optimized scheduling strategy.

2. The system according to claim 1, characterized in that, The source load data prediction module includes a secondary decomposition unit, a feature selection unit, a nested cross-validation unit, and an ensemble learning prediction unit. The secondary decomposition unit is used to decompose the time-series load data of user electricity load, new energy power generation and electric vehicle load into multiple intrinsic mode components using CEEMDAN. The nonlinear complexity is quantified by calculating the permutation entropy of each component, and high-frequency non-stationary components are selected as key modes. Furthermore, wavelet packet decomposition is used to perform a secondary decomposition on the remaining residual terms to obtain the secondary decomposition results; The feature selection unit is used to extract low-frequency features and sudden fluctuation features from the secondary decomposition results, detect energy mutation points based on HHT transform, generate time-varying feature vectors, calculate the HSIC value of each influencing factor, and sort the corresponding influencing factors according to the HSIC value from largest to smallest to obtain a candidate feature set. The nested cross-validation unit has an outer layer used to divide the 5-fold test set in chronological order, and an inner layer that uses a Bayesian optimizer to synchronously search for parameters on the training set, iteratively eliminates redundant features, and finally outputs a stable feature set and the optimal parameter combination. The integrated learning prediction unit is used to decompose the obtained stable feature set into high-frequency components, mid-frequency components and low-frequency components, which are modeled by two-layer LSTM / TCN, single-layer LSTM / TCN and BPNN base learners, respectively, to capture multi-scale features. The prediction results of each component are integrated through a fully connected layer of the meta-learner, and the final prediction result is obtained after optimizing the weights.

3. The system according to claim 1, characterized in that, The day-ahead optimization scheduling unit in the optimization scheduling module is also used for: The central energy management platform is taken as the leader, and a day-ahead optimization scheduling model for the leader is established with the goal of maximizing the daily operating revenue of the leader. Taking the distributed energy system as the first follower, and establishing a day-ahead optimization scheduling model for the first follower with the goal of maximizing the daily operating revenue of the first follower; The electric vehicle cluster is regarded as the second follower, and the day-ahead optimization scheduling model of the second follower is established with the optimization objective of maximizing the sum of the consumer surplus of the second follower and the revenue of electric vehicle users. A day-ahead optimization scheduling model for the third follower is established by taking the regular load aggregator as the third follower and maximizing the total consumer surplus of the third follower as the optimization objective. By combining the leader's day-ahead optimization scheduling model, the first follower's day-ahead optimization scheduling model, the second follower's day-ahead optimization scheduling model, and the third follower's day-ahead optimization scheduling model, a master-slave Stackelberg game model is established, so that the day-ahead optimization scheduling strategy after game equilibrium is generated based on the master-slave Stackelberg game model.

4. The system according to claim 3, characterized in that, The day-ahead optimization scheduling strategy generated based on the master-slave Stackelberg game model after generating the game equilibrium in the day-ahead optimization scheduling unit includes: The central energy management platform generates an electricity price strategy set S based on the energy supply and demand forecast of the entire network; and encodes multiple pricing schemes through quantum superposition states. The distributed energy system initializes the corresponding strategy set. S DES Based on the central government's energy purchase price, the output of each unit is adjusted through the quantum rotating door to optimize energy production costs and energy sales revenue. The electric vehicle cluster initializes the corresponding strategy set SEVA; and dynamically adjusts the charging power and discharging power of electric vehicles according to the central charging and discharging price, combining quantum cross-balancing user demand and grid constraints. The strategy set corresponding to the initialization of the regular load aggregator. S LAS Furthermore, the quantum Grover algorithm is used to search for the optimal load combination and minimize the energy cost under the central electricity price. The central energy management platform receives the optimization results from the distributed energy system, the electric vehicle cluster, and the conventional load aggregator, generates a hyperlocation strategy pool through quantum entanglement fusion, and uses quantum measurement to screen for Pareto optimal solutions to obtain the day-ahead optimized scheduling strategy.

5. The system according to claim 1, characterized in that, The intraday optimized scheduling unit in the optimized scheduling module is also used for: The load adjustment cost is determined based on the adjustment amount of the power of each unit in the day-ahead optimized scheduling strategy. The heat load adjustment cost is determined based on the adjustment amount of the heat power of each unit in the day-ahead optimized scheduling strategy. The emergency operation costs of energy storage and electric vehicles are determined based on the emergency charging and discharging power of energy storage devices and the emergency charging and discharging power of electric vehicle clusters, respectively. The cost of reserve capacity is determined by a hybrid model that combines deterministic basic reservation with dynamic adjustments based on historical events. The minimum value of the sum of the electrical load adjustment cost, the thermal load adjustment cost, the energy storage emergency operation cost, the electric vehicle emergency operation cost, and the reserve capacity reservation cost is determined as the objective function for intraday optimal scheduling.

6. The system according to claim 5, characterized in that, The method for determining the reserve capacity reservation cost using a hybrid model that combines deterministic basic reservation with dynamic adjustments based on historical events includes: A deterministic basic reserved model is constructed based on the shortest duration of independent load support by energy storage devices and electric vehicle clusters, and the peak load power of microgrids; Based on the number of extreme events occurring within the statistical period and the actual energy used by the charging station within the statistical period, the deterministic basic reserved model is dynamically corrected to obtain the historical event dynamic correction model. The reserve cost of the standby capacity is determined based on the difference between the output of the historical event dynamic correction model and the minimum state of charge of the energy storage under normal operating conditions.

7. The system according to claim 1, characterized in that, The power quality monitoring module, by analyzing historical monitoring data of power quality indicators from various power stations, determines the power quality anomaly trigger threshold and classifies the anomaly trigger levels, including: Collect historical monitoring data of power quality indicators for each power station and integrate the historical monitoring data of each indicator into an independent indicator dataset; The K-means clustering method based on the nearest neighbor density matrix is ​​used to cluster the index datasets corresponding to each power quality index, and k-class index clustering results corresponding to each power quality index are obtained. The clustering results of the k-type indicators are determined as k-type trigger objects, and the k-type trigger objects are divided into different quality anomaly trigger levels according to the degree of anomaly; Based on the clustering results of the k-type indicators, the average value of the indicators of the same type of triggering objects is calculated to obtain the trigger threshold of the k-type triggering objects.

8. The system according to claim 1, characterized in that, The triggering and scheduling module is also used for: Obtain the duration of each power quality indicator exceeding the limit, as well as the penalty coefficient for different trigger levels, and determine the cost of power quality limit exceeding penalty. Obtain the emergency operation cost of each energy storage device and the emergency operation cost of each electric vehicle in the electric vehicle cluster to determine the equipment adjustment cost; When the local resources of a microgrid cannot quickly restore power quality, in order to avoid purchasing electricity from the grid at a high price due to power shortage, the emergency power purchase cost is calculated. When the output of new energy sources increases suddenly, in order to prevent equipment overload, it is necessary to reduce the output of wind and solar power and calculate the cost of curtailment penalties. Based on the power quality violation penalty cost, the equipment adjustment cost, the emergency power purchase cost, and the power abandonment penalty cost, a trigger scheduling objective function is constructed; An improved Black-winged Kite optimization algorithm is used to solve the triggering scheduling objective function to obtain the power quality anomaly solution.

9. The system according to any one of claims 1 to 8, characterized in that, The optimized scheduling module further includes a dual-layer control unit for microgrids and electric vehicles, which is used for: Establish fast charging rights accounts for electric vehicles participating in demand-side response to record corresponding fast charging rights points; Based on the charging characteristics of electric vehicles, the charging habits of car owners, the incomplete charging capacity, and the fast charging rights integral, an electric vehicle charging model is constructed with the goal of minimizing the charging cost of electric vehicles. The microgrid's scheduling model is constructed with the objective function of maximizing the revenue generated from electricity sales minus the equivalent cost of incomplete charging, and then minus the equivalent cost related to fast charging rights points. The electric vehicle charging model and the microgrid scheduling model constitute a two-layer optimization model; and the particle swarm optimization algorithm is used to solve the two-layer optimization model to obtain the corresponding power output of the electric vehicles participating in the demand-side response.

10. A method for multi-timescale scheduling triggered within the day-ahead period of a microgrid, characterized in that, include: The source-load data prediction module in the microgrid is used to predict user electricity load, new energy power generation and charging station load using a source-load integrated prediction method based on quadratic decomposition and feature selection, and the prediction results are obtained. A dynamic time-of-use pricing model for electric vehicles is established based on the prediction results using the dynamic time-of-use pricing module in a microgrid; the dynamic time-of-use pricing model for electric vehicles dynamically adjusts the time-of-use pricing for electric vehicles based on the difference between power supply and power consumption; The power quality monitoring module in the microgrid is used to perform hierarchical monitoring and analysis of the power quality of the microgrid, and hierarchical triggering is performed when power quality is abnormal. The optimization scheduling module and the triggering scheduling module in the microgrid are used to perform daytime optimization scheduling, intraday optimization scheduling, and power quality anomaly handling, resulting in intraday optimization scheduling strategy and power quality anomaly solution. The instruction allocation module in the microgrid is used in conjunction with the power quality anomaly solution and the intraday optimized scheduling strategy to generate scheduling instructions.

Images