Optimized scheduling method and system for virtual power plant

Abstract

The invention discloses an optimal scheduling method and system for a virtual power plant, and relates to the technical field of power system scheduling and operation control, and the method comprises the steps: enabling a target virtual power plant to collect load data through a connection interface of the target virtual power plant and a power grid system, and carrying out the short-term load prediction, and obtaining the short-term load prediction data; acquiring a unit start-stop and energy storage charge-discharge state feature set, and performing supply-demand matching with the short-term load prediction data; and according to a supply and demand matching result, optimizing and dispatching a generator set set and an energy storage equipment set by taking minimum energy consumption as a constraint. According to the method, the technical problems that the dynamic operation characteristics of the energy unit cannot be accurately matched, the energy consumption is high and the load balance adjustment is lagged due to the fact that the starting switching energy consumption of the generator set, the charging and discharging loss of the energy storage equipment and the power grid load balance degree are not comprehensively considered in virtual power plant scheduling in the traditional technology are solved; the technical effects of accurately matching the dynamic operation characteristics of the energy unit, reducing the energy consumption and improving the load balance adjustment real-time performance are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system dispatching and operation control, and particularly to an optimized dispatching method and system for a virtual power plant. Background Art

[0002] In a power supply circuit system, a virtual power plant needs to integrate a generator set and energy storage devices to achieve optimized dispatching. In the prior art, traditional load forecasting and equipment control methods are mostly adopted for the dispatching of virtual power plants. These methods have played a certain role in a stable power grid environment. However, with the improvement of power grid intelligence and the dynamic change requirements of load, many limitations have emerged when traditional technologies are applied to the dispatching of virtual power plants. Since the traditional dispatching methods do not fully consider multi-dimensional factors such as generator sets, energy storage devices, and the power grid, they cannot accurately match the dynamic operation characteristics of each energy unit in the virtual power plant, resulting in a high energy consumption of the dispatching scheme and a lag in load balance adjustment, and it is difficult to meet the requirements of the efficient operation of the virtual power plant and the stability of the power grid. Summary of the Invention

[0003] This application provides an optimized dispatching method and system for a virtual power plant, which is used to solve the technical problem that traditional technologies do not comprehensively consider the start-up and switching energy consumption of generator sets, the charge and discharge losses of energy storage devices, and the power grid load balance degree in the dispatching of virtual power plants, resulting in the inability to accurately match the dynamic operation characteristics of energy units, high energy consumption, and a lag in load balance adjustment.

[0004] In the first aspect of this application, an optimized dispatching method for a virtual power plant is provided. The method includes: the target virtual power plant performs time-series collection of load demand data for a preset window through a connection interface with the power grid system, and performs short-term load forecasting based on the collection result to obtain short-term load forecasting data; obtaining the start-stop state-output operation characteristic set of the generator set in the generator set set of the target virtual power plant; obtaining the charge-discharge state-equipment operation state characteristic set of the energy storage device set of the target virtual power plant; based on the start-stop state-output operation characteristic set of the generator set and the charge-discharge state-equipment operation state characteristic set, performing supply-demand matching with the short-term load forecasting data to obtain a supply-demand matching result; according to the supply-demand matching result, taking the minimization of the start-up and switching energy consumption of the generator set and the charge-discharge conversion energy consumption of the energy storage device as constraints, performing operation optimization dispatching on the generator set set and the energy storage device set.

[0005] In the second aspect of the present application, an optimized scheduling system for a virtual power plant is provided. The system includes: a short-term load prediction data acquisition module, which is used for the target virtual power plant to perform time-series acquisition of load demand data for a preset window through a connection interface with the power grid system, and perform short-term load prediction based on the acquisition results to obtain short-term load prediction data; an operating characteristic set acquisition module, which is used to obtain the start-stop state-output operating characteristic set of the generator sets in the generator set of the target virtual power plant; an operating state characteristic set acquisition module, which is used to obtain the charge-discharge state-equipment operating state characteristic set of the energy storage device set of the target virtual power plant; a supply-demand matching result acquisition module, which is based on the start-stop state-output operating characteristic set of the generator sets and the charge-discharge state-equipment operating state characteristic set, and performs supply-demand matching with the short-term load prediction data to obtain a supply-demand matching result; an operating optimization scheduling execution module, which is used to perform operating optimization scheduling on the generator set and the energy storage device set according to the supply-demand matching result, with the constraint of minimizing the start-up and switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices.

[0006] One or more technical solutions provided in the present application have at least the following technical effects or advantages: By collecting the load demand data of the target virtual power plant and the operating characteristics of each energy unit, the present application obtains the basis for scheduling decisions through processing such as long and short trend analysis and supply-demand matching, and performs optimization scheduling with the constraint of minimizing the start-up and switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices. Combining the analysis of the energy consumption function and the compensation adjustment of the scheduling plan, the multi-dimensional energy consumption optimization of the virtual power plant and the regulation of the power grid load balance are accurately realized, making the operation scheduling result of the virtual power plant more efficient and reliable, achieving the technical effects of accurately matching the dynamic operating characteristics of the energy units, reducing energy consumption and improving the real-time performance of load balance regulation. Description of the Drawings

[0007] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0008] Figure 1 It is a schematic flowchart of an optimized scheduling method for a virtual power plant provided by an embodiment of the present application.

[0009] Figure 2 It is a schematic structural diagram of an optimized scheduling system for a virtual power plant provided by an embodiment of the present application.

[0010] Explanation of the accompanying symbols: short-term load forecast data acquisition module 1, operation feature set acquisition module 2, operation status feature set acquisition module 3, supply and demand matching result acquisition module 4, operation optimization scheduling execution module 5. DETAILED DESCRIPTION

[0011] The present application provides an optimization scheduling method and system for virtual power plants, which is used to solve the technical problems that traditional technologies do not comprehensively consider the energy consumption of generator set startup and switching, energy storage equipment charging and discharging losses and grid load balance in virtual power plant scheduling, resulting in the inability to accurately match the dynamic operation characteristics of energy units, high energy consumption and lagging load balance adjustment.

[0012] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0014] Example 1, as Figure 1 As shown, an optimization scheduling method for a virtual power plant, wherein the method includes: Step A100: The target virtual power plant performs time series collection of load demand data in a preset window through a connection interface with the power grid system, and performs short-term load forecasting based on the collection results to obtain short-term load forecast data.

[0015] In the embodiments of the present application, the target virtual power plant refers to a power plant system that integrates a collection of generator sets and a collection of energy storage devices, performs time-series acquisition of load demand data, performs short-term load forecasting and supply-demand matching through a connection interface with the power grid system, and optimizes the scheduling of energy units with the constraint of minimizing energy consumption. The power grid system refers to the power supply or distribution circuit device and operating system that implements load demand data exchange and scheduling execution feedback with the target virtual power plant through a connection interface, and carries out power supply and demand balancing and regulation.

[0016] Specifically, the target virtual power plant collects the load demand data sequence of a preset window through a connection interface, determines the prediction trend characteristics through long and short trend analysis, extracts the concentrated load demand data, and superimposes the short-term load prediction trend characteristics to obtain the short-term load prediction data. The specific steps are described in detail in A110 - A140.

[0017] Step A200: Obtain the generator set start-stop state-output operation characteristic set of the generator set collection of the target virtual power plant.

[0018] Optionally, first, deploy a real-time sensor network at each key node of the generator sets in the target virtual power plant, and collect the unit operation parameters at a sampling frequency of 1 second / time, including breaker status signals, generator speed, output voltage / current, etc. For example, for a certain 200MW generator set, the breaker opening and closing signals (high level indicates startup, low level indicates shutdown) and three-phase current data (such as 1200A for phase A, 1180A for phase B, and 1210A for phase C) transmitted back by the sensor in real time are preprocessed through an edge computing unit to filter out interference noises such as voltage fluctuations, and obtain pure state raw data.

[0019] Secondly, based on the preprocessed signals, construct a start-stop state recognition model: when the breaker status signal is at a high level for more than 5 seconds, it is determined that the unit is in the startup state, otherwise it is in the shutdown state. For the units in the startup state, use the power calculation formula , where if U is the line voltage of 10.5kV, is the power factor of 0.85, and calculate the real-time output power. Exemplarily, the above-mentioned 200MW unit calculates the real-time power of 185MW at a certain moment, and at the same time records the output power fluctuation range (±2MW) and operation duration at this power. Since startup, it has been running continuously for 4 hours and 20 minutes, forming an output operation characteristic vector [startup state, 185MW, 2MW, 4h20min]; if the unit is in the shutdown state, record the shutdown duration (such as 1 hour and 15 minutes of shutdown) and the power parameters at the end of the last operation, forming [shutdown state, 0MW, 0, 1h15min].

[0020] Finally, pack and transmit the state characteristic vectors of all generator sets to the virtual power plant central controller according to a preset protocol (such as ModbusTCP), and store them in the time series database to form a start-stop state-output operation characteristic set including multiple units. For example: Unit 1: [startup, 185MW, 2MW, 4h20min], Unit 2: [shutdown, 0MW, 0, 1h15min], etc.

[0021] Through high-frequency sampling of the real-time sensor network, edge computing preprocessing, and the status recognition model, the second-level accurate acquisition of the start-stop status and output characteristics of the generating set is realized, providing real-time and accurate basic data support for the supply-demand matching and optimal dispatching of the virtual power plant, and effectively improving the timeliness and reliability of dispatching decisions.

[0022] Step A300: Obtain the charge-discharge status - equipment operation status characteristic set of the energy storage device set of the target virtual power plant.

[0023] In an embodiment of the present application, first, a high-frequency sampling module is deployed in the battery management system (BMS) and the power conversion system (PCS) of the energy storage device to collect the terminal voltage of the energy storage battery, the charge-discharge current, the battery temperature, and the PCS operation status signal at a frequency of 500 ms / time. For example, for the battery cluster of a 100 MWh energy storage power station, the voltage data (such as 3.25 V / monomer), current data (+150 A during charging, -180 A during discharging), and PCS circuit breaker status (closed indicates charging or discharging, open indicates standby) transmitted back by the sampling module in real time are denoised and filtered by the edge computing unit to eliminate interference such as temperature drift, and the original status data is obtained.

[0024] Secondly, a charge-discharge status recognition model is constructed: when the PCS circuit breaker status is closed and the absolute value of the current continues to be greater than 5% of the rated current for 10 seconds, it is determined to be in the charging or discharging state, with the current being positive for charging and negative for discharging; if the absolute value of the current is less than 5% of the rated current or the circuit breaker is open, it is determined to be in the standby state. For the charging state, using the formula , where if U is the battery terminal voltage of 3.25 V and I is the charging current of +150 A, the charging power is calculated as 487.5 W, and at the same time, the charging efficiency (such as 95%), the battery state of charge (SOC, currently 75%), and the charge-discharge cycle times (the 1200th time) are recorded to form the charging state feature vector [charging, 487.5 W, 95%, 75%, 1200 times]; if in the discharging state, the discharging power is calculated in the same way and parameters such as the depth of discharge (such as DOD 60%) are recorded to form [discharging, -540 W, 94%, 60%, 1200 times]; for the standby state, [standby, 0 W, -, 75%, 1200 times] is recorded, where "-" is a placeholder used to indicate that there are no corresponding valid parameters in this state. Because no charge-discharge behavior occurs in the standby state, parameters such as the charging efficiency (exclusive to the charging state) and the depth of discharge (DOD) that are bound to the charge-discharge action do not have actual values in the standby stage, so "-" is used to fill this position to ensure the unity of the structure of the feature vector and at the same time clarify that this parameter is meaningless in this state.

[0025] Finally, the state feature vectors of all energy storage devices are uploaded to the virtual power plant management platform through the OPCUA protocol and stored in the time series database, forming a charge-discharge state - device operation state feature set containing multiple energy storage units. For example: Energy storage unit No. 1: [Charging, 487.5W, 95%, 75%, 1200 times], Energy storage unit No. 2: [Discharging, -540W, 94%, 60%, 1200 times], etc.

[0026] Through the high-frequency sampling module, edge computing preprocessing, and state recognition model, the second-level accurate acquisition of the charge-discharge state and operation characteristics of energy storage devices is realized, providing real-time and accurate energy storage state data support for the supply-demand matching and energy consumption optimal scheduling of the virtual power plant, and effectively improving the scheduling response speed and utilization efficiency of energy storage resources.

[0027] Step A400: Based on the start-stop state - output operation feature set of the generator sets and the charge-discharge state - device operation state feature set, perform supply-demand matching with the short-term load prediction data to obtain the supply-demand matching result.

[0028] Specifically, based on the start-stop state - output operation feature set of the generator sets and the charge-discharge state - device operation state feature set, extract the characteristics of the generator sets in the starting state and the energy storage devices in the discharging state, extract the real-time output power set and the real-time maximum discharge power set, and perform supply-demand matching with the short-term load prediction data to obtain the result. The specific steps are detailed in A410 - A420.

[0029] Step A500: According to the supply-demand matching result, with the constraint of minimizing the start-up switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices, perform operation optimization scheduling on the generator set set and the energy storage device set.

[0030] Specifically, judge whether it exceeds the preset tolerance threshold according to the supply-demand matching result. If it exceeds, trigger the operation optimization scheduling instruction and perform scheduling with the constraint of minimizing the start-up switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices. If it does not exceed, execute the acquisition of the next window load data and repeat the short-term load prediction and supply-demand matching process. The specific steps are detailed in A510 - A520.

[0031] Furthermore, step A100 in the method provided by the embodiment of the present application includes: A110: The target virtual power plant executes the load demand data sequence of the preset window through the connection interface with the power grid system.

[0032] A120: Perform long and short trend analysis on the load demand data sequence to determine the short-term load prediction trend characteristics.

[0033] A130: Traverse the load demand data sequence to extract the centralized load demand data and determine the centralized load demand data.

[0034] A140: Use the short-term load forecasting trend characteristics to perform trend superposition on the centralized load demand data to obtain the short-term load forecasting data.

[0035] In the embodiments of the present application, trend superposition is an operation of using a pre-constructed trend superposition analyzer to non-linearly fuse the centralized load demand data with the short-term load forecasting trend characteristics, thereby obtaining the short-term load forecasting data.

[0036] Specifically, first, the target virtual power plant performs a preset window load demand data acquisition operation with a time span of 15 minutes through a dedicated connection interface with the power grid system. Specifically, at a frequency of one sampling point per minute, the real-time consumption data of the power grid load is obtained in real time, and a load demand data sequence containing 15 sampling points is formed within a 15-minute window. For example, in the acquisition period from 8:00 to 8:15 in the morning, the load data of each minute node from 8:00, 8:01 to 8:15 is collected in sequence, forming a load demand data sequence such as [1.2MW, 1.25MW, 1.3MW, …, 1.4MW]. This data acquisition frequency can more finely capture the transient change characteristics of the load in a short period of time.

[0037] When performing long and short trend analysis on this data sequence, the load forecasting long and short trend characteristics are extracted from the load demand data sequence by using long and short trend analysis scales respectively. The mapping similarity analysis is performed on the corresponding sub-characteristics of the two types of characteristics to obtain a similarity set. After normalization, a trend iteration matrix is constructed, and then it is convolved and interacted with the short trend characteristics to obtain the short-term load forecasting trend characteristics. The specific steps are described in detail in A121 - A123.

[0038] In the link of extracting the centralized load demand data, first, for the 15-minute load demand data sequence, assuming it contains 15 sampling points the mean value is calculated, and the average load value is 1.2MW. Then, based on this mean value, the standard deviation of the load demand data sequence (such as 0.15MW) is calculated to determine the centralized data range. Usually, the mean value ± 1 times the standard deviation (that is, from 1.05MW to 1.35MW) is used as the centralized load interval, and the data points falling within this interval are determined as centralized load points, while the marginal data outside the interval is filtered. For example, 1.4MW and 1.45MW in this sequence exceed the upper limit of 1.35MW, and 1.0MW and 1.0MW are lower than the lower limit of 1.05MW, and they will be identified as marginal data and excluded. The remaining 11 centralized load points form a set of centralized load demand data.

[0039] Finally, in the trend superposition step, using the pre-constructed neural network trend superposition analyzer (the construction process is described in detail in A141), the above-mentioned centralized load data (such as the mean of 1.2 MW and the fluctuation characteristics within the centralized interval) and the short-term load prediction trend characteristics (such as the growth slope of 0.05 MW / 15 min every 15 minutes within the next 1 hour) are used as the model inputs. After non-linear transformation inside the model, the short-term load prediction data integrating the centralized load distribution characteristics and trend characteristics is finally output. For example, the mean of the centralized load data collected at a certain industrial park at 9:00 in the morning is 1.3 MW. After superimposing the trend characteristic of increasing by 0.2 MW per hour, the model outputs the load prediction curve from 9:15 to 10:15.

[0040] Through a combined step of multi-scale trend analysis, centralized load feature extraction, and dynamic trend superposition, a high-precision prediction of the load demand of the virtual power plant is achieved, providing a reliable data basis for subsequent supply-demand matching and energy consumption optimal scheduling, and effectively solving the problem of scheduling lag caused by large load prediction errors in the prior art.

[0041] Furthermore, step A120 in the method provided by the embodiment of the present application includes: A121: Respectively use the long-term trend analysis scale and the short-term trend analysis scale to extract trend characteristics from the load demand data sequence, and obtain the long-term load prediction trend characteristics and the short-term load prediction trend characteristics.

[0042] A122: Perform one-to-one mapping similarity analysis on the corresponding sub-characteristics of the long-term load prediction trend characteristics and the short-term load prediction trend characteristics to determine the mapping similarity set.

[0043] A123: By normalizing the mapping similarity set and constructing a trend iteration matrix, use the trend iteration matrix and the short-term load prediction trend characteristics for convolution interaction to obtain the short-term load prediction trend characteristics.

[0044] Optionally, first, according to the fluctuation variance calculation method in step A124-1, identify the fluctuation variance of the 15-minute load data sequence (assuming the total load range is 0.5 - 1.8 MW), and obtain the measured variance of 0.9 MW². The ratio to the preset standard variance of 0.6 MW² is 1.5. If the preset standard trend analysis scale is 24 hours (the specific value is determined by those skilled in the art according to the actual situation), the long-term trend analysis scale is determined to be 1.5×24 hours = 36 hours, and the short-term trend analysis scale is 36 hours / 3 = 12 hours.

[0045] Based on this dual-scale, long-term trend features and short-term trend features are extracted from the data sequence. Exemplarily, for the power grid data of an industrial park, multiple similar load patterns on working days can be captured through a 36-hour long-term trend, and the intra-day periodic fluctuations can be identified through a 12-hour short-term trend, achieving the accurate extraction of multi-time-scale features of the load data.

[0046] Next, when performing sub-feature mapping similarity analysis on the two types of features, the long-term trend features are divided into 4 time sub-segments, each with a duration of 9 hours, and 3 sub-features, namely the mean, slope, and fluctuation amplitude, are extracted from each sub-segment; the short-term trend features are correspondingly divided into 48 15-minute sub-segments, and 3 sub-features are also extracted. Then, the similarity degree of the corresponding sub-segment feature vectors is calculated through cosine similarity, that is, the correlation degree between features is measured by calculating the cosine value of the vector angle, obtaining a set containing 48 similarity values, and calculating its mean. This process can more accurately reflect the correlation of the load trend at different time scales, providing data support for subsequent trend fusion.

[0047] After normalizing the similarity set to the [0, 1] interval, we get [0.89, 0.84, 0.99, 0.92], and construct a 4×12 trend iteration matrix. The matrix elements are filled with the normalized similarities in chronological order to match the mapping relationship between the long-term and short-term sub-segments. When using a graph convolutional network to perform convolutional interaction on this matrix and the short-term trend features, the convolution kernel size is set to 3×3, the stride is 1, and through 2-layer graph convolution operations, the output dimension is (4 - 3 + 1)×(12 - 3 + 1) = 2×10. After two-layer convolution, multi-dimensional features can still be retained, and finally, the short-term load prediction trend features integrating the long-term and short-term trends are obtained.

[0048] Through the combined steps of dual-scale feature extraction, sub-feature similarity analysis, normalized matrix construction, and graph convolutional interaction, the deep fusion of long-term and short-term trends in the load data is achieved, solving the problem of prediction lag caused by single-scale analysis in the prior art, providing a prediction data basis that better fits the real-time load changes for the virtual power plant, and improving the short-term load prediction accuracy.

[0049] Furthermore, step A124 in the method provided in the embodiment of the present application includes: A124-1: Identify the fluctuation variance of the load demand data sequence to obtain the fluctuation variance identification result, calculate the ratio of the fluctuation variance identification result to the preset standard fluctuation variance, and multiply the ratio by the preset standard trend analysis scale to determine the long-term trend analysis scale, and take one-third of the long-term trend analysis scale as the short-term trend analysis scale.

[0050] In the embodiments of the present application, the fluctuation variance is the result obtained after identifying the fluctuation variance of the load demand data sequence, and is used to reflect the fluctuation degree of the load data. The preset standard fluctuation variance is a benchmark value set in advance, and is used to calculate the ratio with the result of the fluctuation variance identification to determine the long-term trend analysis scale.

[0051] Specifically, first, identify the fluctuation variance of the 15-minute load data sequence (assuming the load value range is 0.6 - 1.9 MW). By calculating the average value of the squared differences of the load at adjacent sampling points, the result of the fluctuation variance identification is 0.75 MW². The preset standard fluctuation variance is 0.5 MW², and the ratio of the two is 1.5. The preset standard trend analysis scale is 24 hours, and thus the long-term trend analysis scale is determined to be 1.5 × 24 hours = 36 hours, and the short-term trend analysis scale is 36 hours ÷ 3 = 12 hours. This dynamic scale determination method can more accurately match the load fluctuation characteristics compared with the traditional fixed 24-hour scale, that is, when the load fluctuation variance is larger, such as during the peak period of industrial production, the long-term trend analysis scale automatically expands to capture the longer-term change law; conversely, it shortens to improve the identification accuracy of short-term fluctuations.

[0052] Taking the power grid of an industrial park as an example, under the traditional fixed 24-hour scale, the load prediction error for the period from 14:00 to 14:15 in the afternoon is relatively high. After using the dynamically determined 36-hour long-term trend and 12-hour short-term trend analysis, the prediction error for the same period is reduced. By calculating the ratio of the fluctuation variance to the standard value, the long-term trend analysis scale can be automatically adjusted according to the load fluctuation intensity. For example, when the load variance exceeds the standard value, the scale expands to 36 hours to cover the similar load patterns of multiple working days, while the 12-hour short-term trend can capture the intra-day periodic fluctuations. The combination of the two realizes the accurate extraction of the multi-time scale characteristics of the load data.

[0053] Through the combined steps of load data fluctuation variance identification, ratio calculation with the standard value, and dynamic scale division, the problem in the prior art that the fixed time scale cannot adapt to the dynamic change of the load is solved, enabling the long-term and short-term trend analysis scales to be adaptively adjusted according to the load fluctuation characteristics, reducing the load prediction error, and providing a more accurate time scale basis for the supply-demand matching and optimal dispatching of virtual power plants.

[0054] Furthermore, step A140 in the method provided by the embodiments of the present application includes: A141: Pre-build a trend superposition analyzer, and use the trend superposition analyzer to perform trend superposition analysis on the centralized load demand data and the short-term load prediction trend characteristics to obtain the short-term load prediction data.

[0055] Specifically, the trend superposition analyzer is pre-built as a three-layer neural network model: the input layer contains 20 nodes, including 10 concentrated load data features + 10 short-term trend features; the hidden layer uses 16 ReLU activation function nodes; the output layer is 1 linear node, corresponding to the 1-hour load prediction value. During the model training phase, the load data of the past 3 months is used, with a sampling interval of 15 minutes, a total of 8,640 samples, of which 70% is used as the training set, 20% as the validation set, and 10% as the test set. During training, the mean squared error (MSE) is used as the loss function, and the Adam optimizer (learning rate 0.001) is iterated 500 rounds. When the MSE of the validation set is lower than 0.04 MW², the training stops. Finally, the standard deviation of the prediction error of the test set is 0.08 MW.

[0056] In practical applications, the concentrated load demand data extracted by traversing the load data sequence, such as the peak data that exceeds 130% of the average load continuously for 5 minutes within a certain period, with values of 1.4 MW, 1.5 MW, etc., and the short-term load prediction trend features obtained through double-scale analysis, such as the fluctuation slope of 0.05 MW / 15 min every 15 minutes within the next 1 hour, are used as the input vectors of the model. After the non-linear transformation of the hidden layer, the output layer generates short-term load prediction data that integrates the characteristics of sudden increase in concentrated load and trend features.

[0057] Through the trend superposition analyzer with a pre-built neural network structure, the non-linear fusion of concentrated load data and short-term trend features is realized, solving the problem that the fixed-weight superposition in the prior art cannot adapt to load mutations, reducing the short-term load prediction error, providing high-precision prediction data support for the supply-demand matching of virtual power plants, and effectively improving the real-time performance and accuracy of dispatching decisions.

[0058] Furthermore, step A400 in the method provided in the embodiments of the present application includes: A410: Extract the features in the generator set start-stop state-output operation feature set and the charge-discharge state-equipment operation state feature set that are in the generator set start state and discharge state, and extract the real-time output power set and the real-time maximum discharge power set.

[0059] A420: Perform supply-demand matching on the real-time output power set and the real-time maximum discharge power set with the short-term load prediction data to obtain a supply-demand matching result.

[0060] Specifically, first, select the units in the start-up state from the start-up and shutdown state-output operation characteristic set of the generator sets, and extract their real-time output power. For example, a virtual power plant includes 10 generator sets, 8 of which are in the start-up state, and the real-time output powers are 185MW, 190MW, 178MW, etc., forming a real-time output power set [185MW, 190MW, 178MW, 188MW, 192MW, 180MW, 175MW, 182MW]. At the same time, extract the devices in the discharging state from the charge-discharge state-device operation state characteristic set of the energy storage devices, and extract their real-time maximum discharge power. For example, 15 out of 20 energy storage units are in the discharging state, and the real-time maximum discharge powers are 50MW, 45MW, 60MW, etc., forming a real-time maximum discharge power set .

[0061] Next, match the above real-time output power set, real-time maximum discharge power set with the short-term load prediction data to obtain the supply-demand matching result. Assume that the short-term load prediction data shows that the load in the next 15 minutes is 1200MW, the total sum of the real-time output power set is 1470MW, and the total sum of the real-time maximum discharge power set is 781MW. When matching supply and demand, it is necessary to calculate the difference between the total power supply capacity (1470MW + 781MW = 2251MW) and the load prediction value to determine that the supply-demand surplus is 1051MW and judge whether it meets the grid balance demand. If the predicted load is 2500MW, then the supply-demand gap is 249MW, and subsequent optimal dispatching needs to be triggered.

[0062] By high-frequency collecting the real-time operation characteristics of the generator sets and energy storage devices, constructing an accurate power set, and dynamically matching with the short-term load prediction data, the real-time and accurate connection between the power supply capacity of the virtual power plant and the load demand is realized, providing reliable data support for the optimal dispatching decision-making, and effectively improving the stability of the power grid operation.

[0063] Further, step A500 in the method provided by the embodiment of the present application includes: A510: Judge whether the supply-demand matching result exceeds the preset supply-demand matching tolerance threshold. If so, trigger an operation optimization dispatching instruction, and according to the operation optimization dispatching instruction, with minimizing the start-up and switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices as constraints, perform operation optimization dispatching on the generator set set and the energy storage device set.

[0064] A520: If not, then perform the time-series acquisition of the load demand data of the next preset window, and perform short-term load prediction and supply-demand matching again.

[0065] In one embodiment, first, a dynamic evaluation mechanism for the tolerance threshold of supply-demand matching is established. Based on the historical operation data of the virtual power plant, combined with the start-up and switching energy consumption characteristics of the generating units (for example, the cold start energy consumption of a certain 300MW coal-fired unit is 500kWh per time) and the charge-discharge conversion energy consumption parameters of the energy storage devices (for example, the converter of a 1000-degree energy storage power station consumes 48kWh for each charge-discharge conversion), the initial threshold is set at ±5%. When the real-time supply-demand matching result exceeds this range, an optimized dispatch is triggered. For example, at a certain moment, the short-term load forecast value is 1500MW, and the real-time power supply capacity (generating unit output + energy storage discharge) is 1420MW. The gap of 80MW exceeds the 5% threshold of 75MW, so the optimization process is started. The specific steps are described in detail in A511 - A513.

[0066] If the supply-demand matching result does not exceed the threshold, for example, the gap is within ±5%, the load data collection for the next preset window is performed. For example, the load demand data is collected at a 15-minute interval, and the short-term load forecast and supply-demand matching are carried out again to form a closed-loop monitoring mechanism. This design can avoid the energy consumption loss caused by frequent dispatching.

[0067] Through the dynamic threshold judgment mechanism and energy consumption constraint optimization, the efficient utilization of the virtual power plant resources and the improvement of the applicability of the power grid operation are realized, providing a reliable technical support for the multi-energy collaborative dispatch.

[0068] Furthermore, step A510 in the method provided by the embodiment of the present application includes: A511: Use the switching energy consumption function to traverse the generating unit start-stop state-output operation characteristic set for state switching energy consumption analysis to obtain the state switching energy consumption set.

[0069] A512: Use the charge-discharge conversion energy consumption function to traverse the charge-discharge state-equipment operation state characteristic set for charge-discharge conversion energy consumption analysis to obtain the charge-discharge energy consumption set.

[0070] A513: With the goal of compensating the supply-demand matching result and minimizing the start-up and switching energy consumption of the generating units and the charge-discharge conversion energy consumption of the energy storage devices as constraints, combine the state switching energy consumption set and the charge-discharge energy consumption set to determine the operation optimization dispatch plan, and perform operation optimization dispatch on the generating unit set and the energy storage device set based on the operation optimization dispatch plan.

[0071] In the embodiment of the present application, the switching energy consumption refers to the energy consumption generated during the start-stop state switching of the generating units, and its calculation is based on the switching energy consumption function constructed by the start-up time, shutdown time, and output fluctuation range. The charge-discharge energy consumption refers to the energy consumption generated during the charge-discharge state conversion of the energy storage devices, and its calculation is based on the charge-discharge conversion energy consumption function constructed by the charge efficiency, discharge power, and discharge cycle.

[0072] Optionally, first, construct a switching energy consumption and charge-discharge conversion energy consumption function and generate particle coding. The switching energy consumption function is constructed based on the start-up time, shutdown time, and output fluctuation range. For example, the cold start energy consumption function of a 300MW coal-fired unit is as follows: , where is the switching energy consumption of the generator set, that is, the energy loss generated when the unit switches between start-stop states (such as from shutdown to start-up, or a large adjustment of the output during operation); is the start-up time, which refers to the time required for the generator set to transition from the shutdown state to the stable operation state, usually in minutes; is the shutdown time, that is, the time taken for the unit to transition from the operating state to a complete shutdown, the duration of the last shutdown, which is used to reflect the impact of the equipment standby state on the start-up energy consumption; is the output fluctuation range, which refers to the deviation value between the actual output and the rated output during the start-up or adjustment process of the unit.

[0073] Exemplarily, when the start-up time is 20 minutes, the shutdown time is 30 minutes, and the output fluctuation is 15MW, the single start-up energy consumption is 0.7×20 + 0.3×30 + 0.04×15 = 23.6kWh. After traversing the start-stop states of 10 units, a state switching energy consumption set [23.6, 15.8, 0, 22.3, 0,...]kWh is formed (0 indicates no switching state).

[0074] The charge-discharge conversion energy consumption function is constructed based on the charge efficiency, discharge power, and discharge cycle. For example, the conversion energy consumption function of a 500kWh energy storage device is as follows: , where is the charge-discharge conversion energy consumption of the energy storage device, which is the energy consumption generated due to equipment loss, energy conversion efficiency, etc. when the energy storage switches between charge and discharge modes (such as from charge to discharge); is the discharge power, which is the output power of the energy storage device during discharge; is the discharge cycle, which is the duration of a complete discharge process of the energy storage; is the charge efficiency, which is the ratio of the actual stored energy to the input energy when the energy storage device is charging.

[0075] For example, when the discharge power is 80 kW, the discharge cycle is 45 minutes, and the charging efficiency is 92%, the energy consumption per conversion is 0.05 × 80 + 0.03 × 45 / 0.92 ≈ 5.47 kWh. After traversing 20 energy storage units, the charge and discharge energy consumption set is [5.47, 0, 3.92, 4.47, 0, ...] kWh (0 indicates no conversion). The generator start and stop states (10 dimensions) and the energy storage charge and discharge states (20 dimensions) are encoded as 30-dimensional particles, each representing a scheduling scheme. For example, the particle [1, 1, 0, ..., 1, -1, 0] represents the start of the first two units, the discharge of the first energy storage unit, and the charging of the second energy storage unit.

[0076] Next, perform iterative optimization using the particle swarm algorithm. Initialize 30 particles and randomly generate their initial positions and velocities within the range (-0.5, 0.5). For example, the initial position of particle 1 corresponds to a total energy consumption of 258 kWh. The fitness function is defined as: ,in It is the fitness value, which is used to measure the quality of the scheduling scheme. The smaller the value, the better the scheme. It is a weight coefficient used to balance the importance of supply-demand matching error and energy consumption, ranging from 0 to 1.

[0077] For example, Taking 0.6, when the load forecast is 1800MW and the real-time power supply capacity is 1720MW, the fitness of a particle is 0.6×80+0.4×258=151.2. During the iteration process, each particle updates the individual best (pbest) and the global best (gbest). For example, the fitness of particle 7 drops from 180 to 135 at the 12th iteration, and the pbest is updated; the fitness of the global best particle is 102.5 at the 18th iteration, corresponding to a supply-demand gap of 10MW and a total energy consumption of 170.8kWh. The speed and position update formula is: 、 ,in For example, particle 3 adjusts its update rate to start three units and discharge five energy storage devices, reducing its fitness from 160 to 118. The solution terminates after 50 iterations or when the fitness change rate is less than 1%. The optimal solution is obtained: start six units, discharge eight energy storage devices, consume a total of 145.3 kWh, and maintain a supply-demand gap of 0 MW.

[0078] Finally, dispatch is executed based on the optimal solution. For example, if the load suddenly increases by 120MW during a certain period, the algorithm prioritizes dispatching the operating units to increase output (increasing energy consumption by 25kWh), then activates one cold unit (consuming 20.6kWh), and simultaneously adjusts the discharge of six energy storage devices (converting energy consumption 4.47 × 6 = 26.82kWh), for a total energy consumption of 72.42kWh.

[0079] Through the global optimization of the energy consumption function by the particle swarm algorithm, the minimum energy consumption scheduling of the virtual power plant under the constraints of supply-demand matching is realized, providing an efficient solution for the grid balance control and applicable operation.

[0080] Furthermore, step A514 in the method provided by the embodiments of the present application includes: A514-1: The switching energy consumption function is constructed based on the start-up time, shutdown time, and output fluctuation range, and the charge-discharge conversion energy consumption function is constructed based on the charging efficiency, discharge power, and discharge cycle.

[0081] In one embodiment, the construction of the specific switching energy consumption function and the charge-discharge conversion energy consumption function has been described in detail in steps A511 - A513, and the definitions of their energy consumption function parameters are shown in Table 1.

[0082] Switching energy consumption function parameters: The start-up time reflects the time taken for the unit to transition from shutdown to stable operation. The longer the time, the higher the energy consumption generated by energy losses during the equipment start-up process, unit preheating, etc., directly affecting the calculation of the start-up switching energy consumption. The shutdown time reflects the duration of the unit's shutdown. If the shutdown is long, the equipment state changes significantly during restart (such as temperature reduction requiring re-preheating), which will additionally increase the switching energy consumption and is used to correct the start-up energy consumption. The output fluctuation range represents the deviation between the actual output and the rated output when the unit starts up or its state changes. A large fluctuation means unstable equipment operation and requires additional energy consumption to maintain, which is a key dynamic indicator for measuring the switching energy consumption.

[0083] Charge-discharge conversion energy consumption function parameters: The charging efficiency measures the energy utilization rate of the energy storage device during charging. A low efficiency means more energy loss during charging, which will be directly reflected in the total charge-discharge conversion energy consumption and determines the basic loss in the energy storage link. The discharge power refers to the output power of the energy storage during discharge. The higher the power, the greater the possible losses in the device's internal circuit, converter, etc. (the device has a power-loss characteristic), affecting the energy consumption during the discharge process. The discharge cycle is the duration of a complete discharge of the energy storage. Different durations result in differences in the thermal loss and battery cycle loss during continuous discharge of the device, and are used to calculate the cumulative energy consumption during the entire discharge process.

[0084] Through these parameters, the energy consumption calculation under different working conditions is refined from dimensions such as the time, power, and efficiency of equipment operation. When scheduling the virtual power plant, it can accurately evaluate the energy consumption costs of unit start-stop and energy storage charge-discharge, providing accurate data for the particle swarm algorithm to find the optimal solution and realizing cost-reducing scheduling.

[0085] Table 1: Definition Table of Energy Consumption Function Parameters In summary, the optimization scheduling method for a virtual power plant provided by the embodiments of the present application has the following technical effects: In this application, time-series collection of load demand data is performed in a preset window of the target virtual power plant. Through processing such as long-term and short-term trend analysis and extraction of concentrated load data, short-term load prediction data is obtained. By combining the start-stop state-output operation characteristic set of the generator set and the charge-discharge state-equipment operation state characteristic set of the energy storage device for supply-demand matching, and taking the minimization of the start-up and switching energy consumption of the generator set and the charge-discharge conversion energy consumption of the energy storage device as constraints for operation optimization scheduling, the multi-energy optimization scheduling of the virtual power plant is accurately realized, making the scheduling decision of the virtual power plant more accurate and efficient, achieving the technical effects of accurately matching the dynamic operation characteristics of energy units, reducing energy consumption, and improving the real-time performance of load balance regulation.

[0086] Embodiment 2. As Figure 2 shown, based on the same inventive concept as the foregoing Embodiment 1, an optimization scheduling system for a virtual power plant provided by an embodiment of this application includes: A short-term load prediction data acquisition module 1, which is used for the target virtual power plant to perform time-series collection of load demand data in a preset window through a connection interface with the power grid system, and perform short-term load prediction based on the collection result to obtain short-term load prediction data.

[0087] An operation characteristic set acquisition module 2, which is used to obtain the start-stop state-output operation characteristic set of the generator sets in the generator set collection of the target virtual power plant.

[0088] An operation state characteristic set acquisition module 3, which is used to obtain the charge-discharge state-equipment operation state characteristic set of the energy storage device collection of the target virtual power plant.

[0089] A supply-demand matching result acquisition module 4, which is based on the start-stop state-output operation characteristic set of the generator set and the charge-discharge state-equipment operation state characteristic set, performs supply-demand matching with the short-term load prediction data, and obtains a supply-demand matching result.

[0090] An operation optimization scheduling execution module 5, which is used to perform operation optimization scheduling on the generator set collection and the energy storage device collection according to the supply-demand matching result, with the minimization of the start-up and switching energy consumption of the generator set and the charge-discharge conversion energy consumption of the energy storage device as constraints.

[0091] Furthermore, the short-term load prediction data acquisition module 1 is used to perform the following steps: The target virtual power plant executes the load demand data sequence of a preset window through the connection interface with the power grid system; performs long and short trend analysis on the load demand data sequence to determine the short-term load forecasting trend characteristics; traverses the load demand data sequence to extract the concentrated load demand data and determines the concentrated load demand data; uses the short-term load forecasting trend characteristics to perform trend superposition on the concentrated load demand data to obtain the short-term load forecasting data.

[0092] Further, the short-term load forecasting data acquisition module 1 is used to execute the following steps: Respectively use the long-term trend analysis scale and the short-term trend analysis scale to extract trend characteristics from the load demand data sequence to obtain the long-term load forecasting trend characteristics and the short-term load forecasting trend characteristics; perform one-to-one mapping similarity analysis on the corresponding sub-characteristics of the long-term load forecasting trend characteristics and the short-term load forecasting trend characteristics to determine the mapping similarity set; perform normalization processing on the mapping similarity set, construct a trend iteration matrix, and use the trend iteration matrix and the short-term load forecasting trend characteristics to perform convolution interaction to obtain the short-term load forecasting trend characteristics.

[0093] Further, the short-term load forecasting data acquisition module 1 is used to execute the following steps: Identify the fluctuation variance of the load demand data sequence to obtain the fluctuation variance identification result, calculate the ratio of the fluctuation variance identification result to the preset standard fluctuation variance, and multiply the ratio by the preset standard trend analysis scale to determine the long-term trend analysis scale, and take one-third of the long-term trend analysis scale as the short-term trend analysis scale.

[0094] Further, the short-term load forecasting data acquisition module 1 is used to execute the following steps: Pre-construct a trend superposition analyzer, and use the trend superposition analyzer to perform trend superposition analysis on the concentrated load demand data and the short-term load forecasting trend characteristics to obtain the short-term load forecasting data.

[0095] Further, the supply-demand matching result acquisition module 4 is used to execute the following steps: Extract the characteristics in the generator set start-stop state-output operation characteristic set and the charge-discharge state-equipment operation state characteristic set that are in the generator set start state and the discharge state, and extract the real-time output power set and the real-time maximum discharge power set; perform supply-demand matching on the real-time output power set and the real-time maximum discharge power set with the short-term load forecasting data to obtain the supply-demand matching result.

[0096] Further, the operation optimization scheduling execution module 5 is used to execute the following steps: Determine whether the supply-demand matching result exceeds a preset supply-demand matching tolerance threshold. If so, trigger the operation optimization scheduling instruction. According to the operation optimization scheduling instruction, with minimizing the start-up and switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices as constraints, perform operation optimization scheduling on the generator set collection and the energy storage device collection. If not, execute the time-series acquisition of the load demand data for the next preset window, and perform short-term load forecasting and supply-demand matching again.

[0097] Further, the operation optimization scheduling execution module 5 is used to execute the following steps: Use the switching energy consumption function to traverse the generator set start-stop state-output operation characteristic set for state switching energy consumption analysis to obtain the state switching energy consumption set; use the charge-discharge conversion energy consumption function to traverse the charge-discharge state-device operation state characteristic set for charge-discharge conversion energy consumption analysis to obtain the charge-discharge energy consumption set; with the goal of compensating the supply-demand matching result and minimizing the start-up and switching energy consumption of the generator sets and the charge-discharge conversion energy consumption of the energy storage devices as constraints, combine the state switching energy consumption set and the charge-discharge energy consumption set to determine the operation optimization scheduling scheme, and perform operation optimization scheduling on the generator set collection and the energy storage device collection based on the operation optimization scheduling scheme.

[0098] Further, the operation optimization scheduling execution module 5 is used to execute the following steps: The switching energy consumption function is constructed based on the start-up time, shutdown time, and output fluctuation range, and the charge-discharge conversion energy consumption function is constructed based on the charging efficiency, discharge power, and discharge cycle.

[0099] The optimization scheduling system for a virtual power plant provided by an embodiment of the present invention can execute the optimization scheduling method for a virtual power plant provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0100] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The included individual units and modules are only divided according to the functional logic, but are not limited to the above division as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0101] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present application shall be included within the protection scope of the present application. In some cases, the actions or steps recited in the present application can be executed in a sequence different from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or consecutive order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An optimal scheduling method for a virtual power plant, characterized in that The method includes: The target virtual power plant performs time-series acquisition of load demand data for a preset window through the connection interface with the power grid system, and performs short-term load forecasting based on the acquisition results to obtain short-term load forecasting data; Obtain the generator set start-stop state-output operation characteristic set of the generator set set of the target virtual power plant; Obtain the charge-discharge state-equipment operation state characteristic set of the energy storage device set of the target virtual power plant; Based on the generator set start-stop state-output operation characteristic set and the charge-discharge state-equipment operation state characteristic set, perform supply-demand matching with the short-term load forecasting data to obtain a supply-demand matching result; According to the supply-demand matching result, with the minimization of the generator set start-up switching energy consumption and the energy storage device charge-discharge conversion energy consumption as constraints, perform operation optimization scheduling on the generator set set and the energy storage device set.

2. The optimization scheduling method for a virtual power plant according to claim 1, characterized in that, The target virtual power plant performs time-series acquisition of load demand data for a preset window through the connection interface with the power grid system, and performs short-term load forecasting based on the acquisition results to obtain short-term load forecasting data, including: The target virtual power plant executes a load demand data sequence for a preset window through the connection interface with the power grid system; Perform long and short trend analysis on the load demand data sequence to determine the short-term load forecasting trend characteristics; Traverse the load demand data sequence to extract concentrated load demand data and determine the concentrated load demand data; Use the short-term load forecasting trend characteristics to perform trend superposition on the concentrated load demand data to obtain the short-term load forecasting data.

3. The optimal scheduling method for a virtual power plant according to claim 2, wherein Performing long and short trend analysis on the load demand data sequence to determine the short-term load forecasting trend characteristics includes: Respectively use the long trend analysis scale and the short trend analysis scale to extract trend characteristics from the load demand data sequence to obtain the load forecasting long trend characteristics and the load forecasting short trend characteristics; Perform one-to-one mapping similarity analysis on the corresponding sub-characteristics of the load forecasting long trend characteristics and the load forecasting short trend characteristics to determine the mapping similarity set; By performing normalization processing on the mapping similarity set and constructing a trend iteration matrix, use the trend iteration matrix and the load forecasting short trend characteristics for convolution interaction to obtain the short-term load forecasting trend characteristics.

4. The optimized scheduling method for a virtual power plant according to claim 3, characterized in that Perform fluctuation variance identification on the load demand data sequence to obtain a fluctuation variance identification result, calculate the ratio of the fluctuation variance identification result to the preset standard fluctuation variance, and multiply the ratio by the preset standard trend analysis scale to determine the long trend analysis scale, and take one-third of the long trend analysis scale as the short trend analysis scale.

5. The optimal scheduling method for a virtual power plant according to claim 2, characterized in that, Pre-construct a trend superposition analyzer, and use the trend superposition analyzer to perform trend superposition analysis on the concentrated load demand data and the short-term load forecasting trend characteristics to obtain the short-term load forecasting data.

6. The optimization scheduling method for a virtual power plant according to claim 1, wherein Based on the generator set start-stop state-output operation characteristic set and the charge-discharge state-equipment operation state characteristic set, perform supply-demand matching with the short-term load forecasting data to obtain a supply-demand matching result, including: Extract the features in the generator set start-stop state-output operation feature set and the charge-discharge state-equipment operation state feature set that are in the generator set startup state and discharge state, and extract the real-time output power set and the real-time maximum discharge power set; Match the supply and demand of the real-time output power set and the real-time maximum discharge power set with the short-term load prediction data to obtain the supply-demand matching result.

7. The optimized scheduling method for a virtual power plant according to claim 1, characterized in that According to the supply-demand matching result, with the constraint of minimizing the generator set startup switching energy consumption and the charge-discharge conversion energy consumption of the energy storage device, perform operation optimization scheduling on the generator set set and the energy storage device set, including: Judge whether the supply-demand matching result exceeds the preset supply-demand matching tolerance threshold. If so, trigger the operation optimization scheduling instruction, and based on the operation optimization scheduling instruction, with the constraint of minimizing the generator set startup switching energy consumption and the charge-discharge conversion energy consumption of the energy storage device, perform operation optimization scheduling on the generator set set and the energy storage device set; If not, execute the time-series acquisition of the load demand data in the next preset window, and perform short-term load prediction and supply-demand matching again.

8. The optimized scheduling method for a virtual power plant according to claim 7, characterized in that, Including: Use the switching energy consumption function to traverse the generator set start-stop state-output operation feature set for state switching energy consumption analysis to obtain the state switching energy consumption set; Use the charge-discharge conversion energy consumption function to traverse the charge-discharge state-equipment operation state feature set for charge-discharge conversion energy consumption analysis to obtain the charge-discharge energy consumption set; With the goal of compensating the supply-demand matching result, with the constraint of minimizing the generator set startup switching energy consumption and the charge-discharge conversion energy consumption of the energy storage device, combine the state switching energy consumption set and the charge-discharge energy consumption set to determine the operation optimization scheduling plan, and perform operation optimization scheduling on the generator set set and the energy storage device set based on the operation optimization scheduling plan.

9. The optimal scheduling method for a virtual power plant according to claim 8, characterized in that, The switching energy consumption function is constructed based on the startup time, shutdown time, and output power fluctuation range, and the charge-discharge conversion energy consumption function is constructed based on the charging efficiency, discharge power, and discharge cycle.

10. An optimized scheduling system for a virtual power plant, characterized in that, For implementing an optimization scheduling method for a virtual power plant according to any one of claims 1-9, the system includes: A short-term load prediction data acquisition module, configured to perform time-series acquisition of load demand data in a preset window through a connection interface with the power grid system for the target virtual power plant, and perform short-term load prediction according to the acquisition result to obtain short-term load prediction data; An operation feature set acquisition module, configured to acquire the generator set start-stop state-output operation feature set of the generator set set of the target virtual power plant; An operation state feature set acquisition module, configured to acquire the charge-discharge state-equipment operation state feature set of the energy storage device set of the target virtual power plant; A supply-demand matching result acquisition module, based on the generator set start-stop state-output operation feature set and the charge-discharge state-equipment operation state feature set, performs supply-demand matching with the short-term load prediction data to obtain the supply-demand matching result; An operation optimization scheduling execution module, configured to perform operation optimization scheduling on the generator set set and the energy storage device set according to the supply-demand matching result, with the constraint of minimizing the generator set startup switching energy consumption and the charge-discharge conversion energy consumption of the energy storage device.

Images