Virtual power plant energy scheduling method, device, electronic device and readable medium

Abstract

The embodiments of the present disclosure disclose a virtual power plant energy scheduling method, device, electronic device and readable medium. A specific implementation of the method includes: obtaining a set of historical virtual power plant data groups; for each historical virtual power plant data group, performing the following steps: preprocessing the historical virtual power plant data group to obtain a processed virtual power plant data group; performing model training on each pre-set initial power load prediction model; determining the target model weight information and the target model structure information; constructing a multi-source power load prediction model, and binding the multi-source power load prediction model and the target data source label; in response to determining that the current time meets the prediction cycle conditions, obtaining a virtual power plant data group; generating each multi-source power load information; and performing energy scheduling according to the multi-source power load information. This implementation can reduce the frequency of energy scheduling, thereby reducing energy waste during the scheduling process.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a virtual power plant energy scheduling method, device, electronic device, and readable medium. Background Art

[0002] A virtual power plant is a system that uses software and intelligent control devices to centrally control and dispatch distributed energy. At present, the method commonly used in virtual power plant energy dispatch is to predict the load of the virtual power plant through a pre-trained single type of neural network model and dispatch energy according to the predicted results.

[0003] However, when the above method is used for virtual power plant energy dispatch, the following technical problems often occur:

[0004] First, the data in virtual power plants are often collected from multiple types of data sources, and the generalization ability of a single neural network model is weak, resulting in a large error between the predicted energy load and the actual load, which in turn leads to a large difference between the dispatched energy and the actual energy required. Multiple energy rescheduling is still required, which in turn leads to energy loss during multiple transmission processes, resulting in energy waste.

[0005] Second, in order to improve the prediction accuracy of the model, the neural network model used often has a higher model depth and complex model structure, which leads to a significant increase in the required computing power.

[0006] Third, during the training process of the neural network model, since the sample data in the sample data set are collected from different data sources, the effect of using conventional numerical methods to screen abnormal data in the sample data set is poor, resulting in a large number of abnormal sample data in the sample data set, causing additional computing power expenditure for model training, and then leading to a waste of computing power resources.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0008] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0009] Some embodiments of the present disclosure propose a virtual power plant energy scheduling method, device, electronic device and computer-readable medium to solve one or more of the technical problems mentioned in the above background technology section.

[0010] In a first aspect, some embodiments of the present disclosure provide a virtual power plant energy scheduling method, the method comprising: obtaining a set of historical virtual power plant data groups from a target database, wherein each historical virtual power plant data group in the above historical virtual power plant data group set corresponds to a data source label; for each historical virtual power plant data group in the above historical virtual power plant data group set, performing the following steps: performing data preprocessing on each historical virtual power plant data included in the above historical virtual power plant data group to obtain a processed virtual power plant data group; based on the above processed virtual power plant data group, performing model training on each pre-set initial power load forecasting model to generate each model weight information and each model structure information, wherein the network structures of the above each initial power load forecasting model are different, and the above each model weight information corresponds to each model training result; based on the above each model training result, the above each model weight information The method comprises the steps of: determining the target model weight information and the target model structure information based on the target model weight information and the target model structure information; constructing a multi-source power load forecasting model according to the determined target model weight information and the target model structure information, and binding the constructed multi-source power load forecasting model and the target data source tag, wherein the target data source tag is the data source tag corresponding to the historical virtual power plant data group; acquiring the virtual power plant data group in response to determining that the current time satisfies the forecasting period condition corresponding to the preset time period; generating each multi-source power load information according to the acquired virtual power plant data group and the constructed multi-source power load forecasting models; and performing energy scheduling according to the generated multi-source power load information, wherein each multi-source power load information generated can be the power load value of the corresponding power demand source within the preset unit time, and the corresponding power demand source includes but is not limited to: steel mills, chemical plants and manufacturing plants.

[0011] In a second aspect, some embodiments of the present disclosure provide a virtual power plant energy scheduling device, the device comprising: an acquisition unit, configured to acquire a set of historical virtual power plant data groups from a target database, wherein each historical virtual power plant data group in the above-mentioned set of historical virtual power plant data groups corresponds to a data source label; an execution unit, configured to execute the following steps for each historical virtual power plant data group in the above-mentioned set of historical virtual power plant data groups: performing data preprocessing on each historical virtual power plant data included in the above-mentioned historical virtual power plant data group to obtain a processed virtual power plant data group; performing model training on each pre-set initial power load forecasting model based on the above-mentioned processed virtual power plant data group to generate each model weight information and each model structure information, wherein the network structures of the above-mentioned each initial power load forecasting model are different, and the above-mentioned each model weight information corresponds to each model training result; based on the above-mentioned each model training result, the above-mentioned each model weight information and the above-mentioned each model model structure information, determine the target model weight information and the target model structure information; construct a multi-source power load forecasting model according to the determined target model weight information and the target model structure information, and bind the constructed multi-source power load forecasting model and the target data source tag, wherein the above-mentioned target data source tag is the data source tag corresponding to the above-mentioned historical virtual power plant data group; a second acquisition unit is configured to acquire the virtual power plant data group in response to determining that the current time meets the prediction period condition corresponding to the preset time period; a generation unit is configured to generate each multi-source power load information according to the acquired virtual power plant data group and each constructed multi-source power load forecasting model; an energy scheduling unit is configured to perform energy scheduling according to each generated multi-source power load information, wherein each generated multi-source power load information can be the power load value of the corresponding power demand source within the preset unit time, and the corresponding power demand source includes but is not limited to: steel mills, chemical plants and manufacturing plants.

[0012] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0013] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner of the above-mentioned first aspect is implemented.

[0014] The above-mentioned embodiments of the present disclosure have the following beneficial effects: the virtual power plant energy dispatching method of some embodiments of the present disclosure can improve the accuracy of load forecasting information, thereby reducing energy waste in the dispatching process. Specifically, the reason for the energy waste in the dispatching process is that the data in the virtual power plant is often collected from multiple types of data sources, and the generalization ability of the single neural network model is weak, resulting in a large error between the predicted energy load and the actual load, which leads to a large difference between the dispatched energy and the actual required energy, and multiple energy re-dispatching is still required, which leads to energy loss during multiple transmissions, resulting in energy waste. Based on this, the virtual power plant energy dispatching method of some embodiments of the present disclosure, first, obtains a historical virtual power plant data group set from the target database. Among them, each historical virtual power plant data group in the above historical virtual power plant data group set corresponds to a data source label. Thus, each virtual power plant data group collected from different data sources can be obtained. Then, for each historical virtual power plant data group in the above historical virtual power plant data group set, the following steps are performed: the first step is to perform data preprocessing on each historical virtual power plant data included in the above historical virtual power plant data group to obtain a processed virtual power plant data group. Therefore, the data quality of each historical virtual power plant data can be improved and noise interference can be reduced through data preprocessing. In the second step, according to the processed virtual power plant data group, the pre-set initial power load forecasting models are trained to generate each model weight information and each model structure information. Among them, the network structure of each initial power load forecasting model is different, and the weight information of each model corresponds to each model training result. Therefore, the generalization ability and prediction accuracy of each model structure for the current data source type can be determined by using the historical virtual power plant data group collected from a single data source to train models with different network structures. In the third step, according to the training results of each model, the weight information of each model and the structure information of each model, the target model weight information and the target model structure information are determined. Among them, the target data source label is the data source label corresponding to the virtual power plant data group. In practice, since different types of data sources have different index requirements for the model, the model weight information and the model structure information with a higher matching degree with the historical virtual power plant data group can be determined from the weight information of each model and the structure information of each model according to the index conditions corresponding to the target data source label, thereby improving the generalization ability of the model. The fourth step is to construct a multi-source power load forecasting model based on the determined target model weight information and target model structure information, and bind the determined multi-source power load forecasting model to the target data source label. In this way, a multi-source power load forecasting model with a better match with the data source represented by the above target data source label can be constructed, thereby improving the model prediction accuracy.Afterwards, in response to determining that the current time satisfies the prediction period condition corresponding to the preset time period, a virtual power plant data group is obtained. Next, based on the obtained virtual power plant data group and the constructed multi-source power load prediction models, each multi-source power load information is generated. Finally, energy scheduling is performed according to the generated multi-source power load information. Among them, each multi-source power load information generated can be the power load value of the corresponding power demand source within a preset unit time. The corresponding power demand sources include but are not limited to: steel mills, chemical plants, and manufacturing plants. Therefore, by performing energy scheduling with each multi-source power load information with higher accuracy, the problem of inaccurate energy scheduling caused by prediction errors can be reduced, thereby reducing the frequency of energy re-scheduling, and then reducing the loss in the energy transmission process, thereby reducing energy waste. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0016] Figure 1 is a flow chart of some embodiments of a virtual power plant energy scheduling method according to the present disclosure;

[0017] Figure 2 It is a structural schematic diagram of some embodiments of the virtual power plant energy dispatching device according to the present disclosure;

[0018] Figure 3 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0019] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0020] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0021] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0022] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0023] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0024] For the user personal information involved in this disclosure (such as the collection, storage, and use of the user's historical energy consumption data), before performing the corresponding operations, the relevant organizations or individuals shall fulfill their obligations, including conducting personal information security impact assessments, fulfilling the obligation to inform the personal information subject, and obtaining the authorization and consent of the personal information subject in advance.

[0025] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0026] Figure 1 The process 100 of some embodiments of the virtual power plant energy scheduling method according to the present disclosure is shown. The virtual power plant energy scheduling method includes the following steps:

[0027] Step 101, obtaining a set of historical virtual power plant data groups from a target database.

[0028] In some embodiments, the execution subject (such as a computing device) of the virtual power plant energy scheduling can obtain a set of historical virtual power plant data groups from the target database. Among them, each historical virtual power plant data group in the above historical virtual power plant data group set corresponds to a data source tag. The above data source tag can represent the data source of each historical virtual computer data included in the corresponding historical virtual power plant data group. For example, the data tag corresponding to a historical virtual power plant data group in the above historical virtual computer data group set is "steel plant", which can represent that each historical virtual power plant data included in the above historical virtual power plant data group comes from the steel plant (that is, collected from the steel plant). The historical virtual power plant data in the above historical virtual power plant data group set includes user historical energy consumption data, circulation source historical operation data and historical power load data. The above user historical energy consumption data can be the user energy consumption data of the corresponding data source or circulation source (operator or manufacturer) in a preset historical time period under a unit time. For example, the above historical preset time period can be "2024-01-01 to 2024-02-01". The above unit time can be every day. The above user energy consumption data can include user power consumption and user power consumption. For example, the data source label corresponding to a historical virtual power plant data group may be "steel plant", and the historical energy consumption data of each user in each historical virtual power plant data included in the above historical virtual power plant data group may be the daily power consumption and power consumption of the user in the time period of "2024-01-01 to 2024-02-01". The above-mentioned historical operation data of the transfer source may be the operation data that characterizes the operation status (operating status) of the corresponding transfer source or data source (operator or manufacturer) in a unit time within the above-mentioned preset historical time period. The above-mentioned operation data may include but is not limited to: the capacity of the transfer source (the gross production value of the operator or manufacturer) and the power consumption of the transfer source (the power consumption of the operator or manufacturer). The above-mentioned historical power load data may be the power load value of each power-consuming device in a unit time at the corresponding data source or transfer source in the above-mentioned preset historical time period. The above-mentioned historical operation data of the transfer source may be the above-mentioned target database may be a database corresponding to the above-mentioned execution subject. The above-mentioned execution subject may be a server. In practice, the above-mentioned execution subject may obtain the historical virtual power plant data group set from the target database through a wired connection method or a wireless connection method.

[0029] It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or developed in the future.

[0030] Step 102: for each historical virtual power plant data set in the historical virtual power plant data set, perform the following steps:

[0031] Step 1021, performing data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group.

[0032] In some embodiments, the execution entity may perform data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group.

[0033] In some optional implementations of some embodiments, the execution subject may perform data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group through the following steps:

[0034] In the first step, each historical virtual power plant data included in the historical virtual power plant data group is subjected to abnormal data detection to generate each abnormal data label. The abnormal data label can indicate that the corresponding historical virtual power plant data is abnormal data.

[0035] In the second step, each historical virtual power plant data corresponding to each generated abnormal data label is removed from the above historical virtual power plant data group to update the above historical virtual power plant data group.

[0036] The third step is to perform data repair processing on each of the removed historical virtual power plant data to obtain each repaired historical virtual power plant data, and insert the above-mentioned each repaired historical virtual power plant data as historical virtual power plant data into the updated historical virtual power plant data group. In practice, first, the above-mentioned execution subject can replace the abnormal values ​​in each of the removed historical virtual power plant data by mean filling to perform data repair on each of the removed historical virtual power plant data to obtain each repaired historical virtual power plant data. Then, the above-mentioned execution subject can insert the above-mentioned each repaired historical virtual power plant data as historical virtual power plant data into the updated historical virtual power plant data group.

[0037] The fourth step is to resample each historical virtual power plant data included in the updated historical virtual power plant data group to obtain a sampled historical virtual power plant data group. In practice, the above-mentioned execution subject can resample each historical virtual computer data included in the updated historical virtual computer data group by a preset resampling algorithm to obtain a sampled historical virtual power plant data group. As an example, the above-mentioned preset resampling algorithm can be a nearest neighbor interpolation algorithm or a linear interpolation algorithm.

[0038] The fifth step is to perform noise reduction processing on each sampled virtual power plant data included in the sampled virtual power plant data group to obtain a processed virtual power plant data group. In practice, the execution subject can perform noise reduction processing on each sampled virtual power plant data included in the sampled virtual power plant data group by a preset data noise reduction algorithm to obtain a processed virtual power plant data group. As an example, the preset data noise reduction algorithm can be a sliding window algorithm or a median filter algorithm.

[0039] In the process of adopting technical solutions to solve the above-mentioned technical problem one, the following technical problem three is often accompanied: in the process of neural network model training, since each sample data in the sample data set is collected from different data sources, the effect of using conventional numerical methods to screen abnormal data in the sample data set is poor, resulting in a large number of abnormal sample data in the sample data set, causing additional computing power expenditure for model training, and then leading to a waste of computing power resources.

[0040] In some optional implementations of some embodiments, the execution subject may perform abnormal data detection on each historical virtual power plant data included in the historical virtual power plant data group according to each preset power load forecast according to the processed virtual power plant data group through the following steps to generate each abnormal data label:

[0041] The first step is to determine the target number of centroids according to the number of historical virtual power plant data in the historical virtual power plant data group. In practice, the execution entity may determine the target number of centroids by taking the square root of the number of historical virtual power plant data in the historical virtual power plant data group.

[0042] In the second step, based on the historical virtual power plant data set, the following target data determination steps are performed:

[0043] The first sub-step is to determine the data difference value between each two historical virtual power plant data in the historical virtual power plant data group. In practice, the execution subject may determine the Euclidean distance between the two historical virtual power plant data as the data difference value.

[0044] The second sub-step is to sort the generated data difference values ​​to obtain a data difference value sequence. In practice, the above execution subject can sort the generated data difference values ​​from large to small to obtain a data difference value sequence.

[0045] The third sub-step is to select a data difference value that satisfies a target sequence condition from the data difference value sequence as a target data difference value. The target sequence condition may be that the target data difference value is the first data difference value in the data difference value sequence.

[0046] In a fourth sub-step, in response to determining that the number of executions of the target data determination step is equal to a preset number of times, two historical virtual power plant data corresponding to the target data difference values ​​are determined as two target historical virtual power plant data. As an example, the preset number of times may be 1.

[0047] The fifth sub-step, in response to determining that the number of executions of the above target data determination step is not equal to the preset number of times, determines the historical virtual power plant data that meets the data difference condition in the two historical virtual power plant data corresponding to the target data difference value as the target historical virtual power plant data. In practice, first, for each historical virtual power plant data corresponding to the above target difference value, the execution subject can perform the following steps: First, determine the data difference value between the above historical virtual power plant data and each determined target historical virtual power plant data. Second, use the mean of each determined data difference value as the difference mean. Finally, the execution subject can select a historical virtual power plant data with a smaller corresponding difference mean from the two historical virtual power plant data corresponding to the above target data difference value as the target historical virtual power plant data.

[0048] The sixth sub-step is to remove each determined target historical virtual power plant data from the historical virtual power plant data group to update the historical virtual power plant data group.

[0049] A seventh sub-step, in response to determining that the number of executions of the target data determination step is less than the target centroid number, executing the target data determination step again.

[0050] The third step is to cluster the historical virtual power plant data group according to the determined target historical virtual power plant data and the target centroid number to obtain a set of historical virtual power plant data clusters. In practice, the execution subject can use the target historical virtual power plant data as the cluster center and the target centroid number as the number of cluster clusters to cluster the historical virtual power plant data group using a preset clustering algorithm to obtain a set of historical virtual power plant data. As an example, the preset clustering algorithm can be a K-means algorithm.

[0051] The fourth step is to select at least one historical virtual power plant data that meets the abnormal data distance condition from the historical virtual power plant data cluster in the above-mentioned historical virtual power plant data cluster set, and to perform abnormal data annotation on the at least one historical virtual power plant data to generate each abnormal data label. Among them, the abnormal data distance condition may be that the data difference value is greater than or equal to the preset distance threshold. In practice, the above-mentioned execution subject may determine the data difference value between every two historical virtual power plant data in the above-mentioned historical virtual power plant database to obtain each data difference value. Then, the obtained data difference values ​​may be sorted in order from large to small to obtain a data difference value sequence. Afterwards, the above-mentioned execution subject may perform abnormal data annotation on at least one historical virtual power plant data corresponding to at least one data difference value that meets the abnormal data distance condition in the data difference value sequence to generate each abnormal data label representing the data abnormality. As an example, the above-mentioned abnormal data label may be "1", which represents that the corresponding historical virtual power plant data is abnormal data.

[0052] The first to fourth steps mentioned above are an inventive point of an embodiment of the present disclosure, and in combination with step 1022, solve the technical problem three mentioned above: "In the process of neural network model training, because each sample data in the sample data set is collected from different data sources, the effect of using conventional numerical methods to screen abnormal data in the sample data set is poor, so that there are still more abnormal sample data in the sample data set, resulting in additional computing power expenditure for model training, and then resulting in a waste of computing power resources." The factors that lead to the waste of computing power are often as follows: In the process of neural network model training, because each sample data in the sample data set is collected from different data sources, the effect of using conventional numerical methods to screen abnormal data in the sample data set is poor, so that there are still more abnormal sample data in the sample data set, resulting in additional computing power expenditure for model training. In order to achieve the effect of reducing computing power, the present invention adopts an improved clustering algorithm, and clusters the historical virtual power plant data included in the historical virtual power plant data group by determining the evenly distributed target historical virtual power plant data, so that the data distribution of each historical virtual power plant data can be determined through the generated historical virtual power plant data cluster set, so that abnormal historical virtual power plant data with a large deviation from normal historical virtual power plant data can be detected from each historical virtual power plant data cluster. Combined with step 1022, the amount of abnormal sample data in the sample data set can be greatly reduced, and the additional computing power expenditure in the model training process can be reduced, thereby reducing the waste of computing power resources.

[0053] Step 1022, based on the processed virtual power plant data group, model training is performed on each pre-set initial power load prediction model to generate each model weight information and each model structure information.

[0054] In some embodiments, the execution subject may perform model training on each pre-set initial power load forecasting model according to the processed virtual power plant data group to generate each model weight information and each model structure information. Among them, the network structures of each initial power load forecasting model are different. Each model weight information corresponds to each model training result. The model weight information may be a model weight parameter obtained after the corresponding initial power load forecasting model is trained. The model structure information may be information used to describe the network structure of the trained initial power load forecasting model. The model training result may be the loss function value of the trained initial power load forecasting model. Each initial power load forecasting model may be a neural network model with virtual power plant data as input and power load forecasting information as output. As an example, each pre-set initial power load forecasting model may include but is not limited to a long short-term memory network model, a recurrent neural network model, a Transformer network model, a CNN-LSTM model and a Bi-RNN model.

[0055] In some optional implementations of some embodiments, the execution subject may perform model training on each preset initial power load forecasting model according to the processed virtual power plant data group to generate each model weight information and each model structure information through the following steps:

[0056] The first step is to determine the user's historical energy consumption data and the flow source historical operation data included in each processed virtual power plant data in the above processed virtual power plant data group as historical virtual power plant sample data to obtain a historical virtual power plant sample data set.

[0057] In the second step, for each of the preset initial power load forecasting models, based on the historical virtual power plant sample data set, the following model training steps are performed:

[0058] In the first sub-step, at least one historical virtual power plant sample data in the historical virtual power plant sample data set is input into the initial power load forecasting model to obtain forecasted power load data corresponding to each historical virtual power plant sample data in the at least one historical virtual power plant sample data input.

[0059] The second sub-step is to determine the model loss function value according to the preset loss function, the obtained predicted power load data and the historical power load data corresponding to the at least one historical virtual power plant sample data. The preset loss function may be a mean square error loss function. In practice, the execution subject may substitute the predicted power load data and the historical power load data into the loss function to determine the model loss function value.

[0060] The third sub-step is to determine whether the initial power load forecasting model reaches a preset optimization target based on the determined model loss function value. The preset optimization target may be that the model loss function value is less than or equal to a preset loss threshold.

[0061] The fourth sub-step is, in response to determining that the initial power load forecasting model achieves the above preset optimization target, performing the following model adjustment steps:

[0062] Sub-step one: determining the initial power load forecasting model as the power load forecasting model.

[0063] Sub-step two: adjusting the above-mentioned power load forecasting model to generate model weight information and model structure information.

[0064] The fifth sub-step is, in response to determining that the initial power load forecasting model does not reach the above preset optimization target, adjusting the model parameters of the initial power load forecasting model, and using the unused historical virtual power plant sample data to form a historical virtual power plant sample data set, using the adjusted initial power load forecasting model as the initial power load forecasting model, and executing the above model training step again. In practice, the above execution subject can adjust the model parameters of the initial power load forecasting model through a back propagation algorithm and a gradient descent algorithm.

[0065] In the process of adopting technical solutions to solve the above-mentioned technical problem one, the following technical problem two is often accompanied: in order to improve the prediction accuracy of the model, the neural network model used often has a high model depth and complex model structure, which leads to a significant increase in the required computing power.

[0066] In some optional implementations of some embodiments, the execution subject may adjust the power load forecasting model through the following steps to generate model weight information and model structure information:

[0067] The first step is to determine the model prediction loss value corresponding to the above power load prediction model. In practice, the above execution entity can input a historical virtual power plant sample data into the above power load prediction model, and determine the current model prediction loss value of the above power load prediction model through the preset loss function of the above power load prediction model.

[0068] The second step is to perform layer-by-layer weight export processing on the above-mentioned power load forecasting model to generate an initial model weight information sequence. Among them, each initial model weight information in the above-mentioned initial model weight information sequence includes the weight values ​​of each neuron. The above-mentioned each neuron weight value corresponds to neuron position information. The above-mentioned neuron position information can be the network layer number and layer position of the model neuron corresponding to the corresponding neuron weight value in the above-mentioned power load forecasting model. In practice, the above-mentioned execution entity can export the various neuron weight values ​​of each model neuron in each layer of the above-mentioned power load forecasting model as the initial model weight information, and sort them from shallow to deep according to the number of layers to generate an initial model weight information sequence.

[0069] Step 3: For each initial model weight information in the initial model weight information sequence, perform the following steps:

[0070] The first sub-step is to generate a non-negative average weight value based on the initial model weight information including the weight values ​​of each neuron, and determine the non-negative average weight value as the initial weight adjustment value. In practice, the execution entity may average the absolute values ​​of the weight values ​​of each neuron included in the initial model weight information to generate a non-negative average weight value.

[0071] The second sub-step is to generate a weight adjustment value according to the determined non-negative average weight value. In practice, the execution subject may square the ratio of the non-negative average weight value to the preset adjustment coefficient to obtain the weight adjustment value. As an example, the preset adjustment value may be e.

[0072] The third sub-step is to perform the following model structure adjustment steps based on the initial weight adjustment value:

[0073] Sub-step 1: Selecting each neuron weight value that meets the weight comparison condition from the initial model weight information as each target neuron weight value. The weight comparison condition may be that the target neuron weight value is less than or equal to the weight adjustment value.

[0074] Sub-step 2: Delete each model neuron corresponding to each target neuron weight value from the power load forecasting model to obtain an adjusted power load forecasting model. In practice, the above-mentioned execution subject can delete each model neuron corresponding to each target neuron weight value from the power load forecasting model according to the corresponding position information of each neuron to obtain an adjusted power load forecasting model.

[0075] Sub-step three: perform prediction verification on the adjusted power load forecasting model to generate a predicted loss value. In practice, the above-mentioned execution entity can input the historical virtual power plant sample data used when "determining the model predicted loss value corresponding to the above-mentioned power load forecasting model" into the adjusted power load forecasting model, and determine the current model predicted loss value of the adjusted power load forecasting model through the preset loss function of the adjusted power load forecasting model as the predicted loss value.

[0076] Sub-step four, in response to determining that the predicted loss value is less than or equal to the model predicted loss value, taking the sum of the weight adjustment value and the initial weight adjustment value as the initial weight adjustment value, and determining the adjusted power load prediction model as the target power load prediction model, and executing the above-mentioned model structure adjustment step again.

[0077] Sub-step five, in response to determining that the predicted loss value is greater than the model predicted loss value and the number of executions of the above-mentioned model structure adjustment step is less than the preset number of executions, the difference between the weight adjustment value and the initial weight adjustment value is used as the initial weight adjustment value, and the power load prediction model is used to execute the above-mentioned model structure adjustment step again.

[0078] Sub-step six, in response to determining that the execution number of the above-mentioned model structure adjustment step is greater than the preset execution number, the power load forecasting model is determined as the target power load forecasting model.

[0079] The fourth step is to perform weight export processing on the determined target power load forecasting model to generate model weight information. In practice, the above-mentioned execution subject can export the model weight parameters of the determined target power load forecasting model through relevant library functions (for example, the model.save_weights() function in the TensorFlow library in Python) to generate model weight information.

[0080] The fifth step is to generate model structure information according to the determined target power load forecasting model. In practice, the execution subject can export the network structure of the target power load forecasting model through relevant library functions (for example, the model.to_json() function in the TensorFlow library in Python) to obtain model structure information describing the network structure of the target power load forecasting model.

[0081] The first to fifth steps mentioned above are an inventive point of an embodiment of the present disclosure, and in combination with the step, solve the technical problem 2 mentioned above: "In order to improve the prediction accuracy of the model, the neural network model used often has a high model depth and a complex model structure, which leads to a significant increase in the required computing power." The factors that lead to a significant increase in computing power are often as follows: In order to improve the prediction accuracy of the model, the neural network model used often has a high model depth and a complex model structure. In order to achieve the effect of reducing computing power, the present disclosure adopts a model structure adjustment method of deleting each neuron with a smaller contribution in the neural network model. First, by determining the prediction loss value of the model, a benchmark reference is provided for the model structure adjustment. Then, by deriving the model weight information layer by layer and calculating the non-negative average weight value to determine the weight adjustment value, it is possible to determine and delete each neuron with a smaller contribution to the prediction loss in the above-mentioned power load prediction model. Afterwards, by iteratively adjusting and verifying the model structure, it is ensured that the adjusted model still maintains or improves the prediction accuracy, thereby reducing the complexity and computing requirements of the model, thereby effectively reducing the computing power requirements during the model task process, and maintaining the original prediction performance of the model.

[0082] Step 1023, determine the target model weight information and the target model structure information according to the training results of each model, the weight information of each model and the structure information of each model.

[0083] In some embodiments, the execution subject may determine the target model weight information and the target model structure information according to the training results of each model, the weight information of each model and the structure information of each model. The target data source tag is the data source tag corresponding to the virtual power plant data group.

[0084] In some optional implementations of some embodiments, the execution subject may determine the target model weight information and the target model structure information according to the training results of the above-mentioned models, the weight information of the above-mentioned models, and the structure information of the above-mentioned models through the following steps:

[0085] In the first step, a model training result that satisfies a model training condition is selected from the above-mentioned model training results as a target model training result. The above-mentioned model training condition can be that the target model training result is the best model training result among the above-mentioned model training results (that is, the loss function value is the smallest).

[0086] In the second step, the model weight information corresponding to the above target model training result is determined as the target model weight information.

[0087] In the third step, the model structure information corresponding to the above target model training result is determined as the target model structure information.

[0088] Step 1024, constructing a multi-source power load forecasting model according to the determined target model weight information and target model structure information, and binding the determined multi-source power load forecasting model and the target data source label.

[0089] In some embodiments, the above-mentioned execution entity can construct a multi-source power load prediction model according to the determined target model weight information and target model structure information, and bind the determined multi-source power load prediction model and the target data source label.

[0090] In some optional implementations of some embodiments, the execution subject may construct a multi-source power load forecasting model according to the determined target model weight information and target model structure information, and bind the determined multi-source power load forecasting model to the target data source label through the following steps:

[0091] The first step is to determine the initial power load forecasting model corresponding to the above target model training results.

[0092] The second step is to adjust the structure of the initial power load forecasting model according to the target model structure information to update the initial power load forecasting model. In practice, the execution subject can delete or add the model structure of the initial power load forecasting model according to the model network structure represented by the target model structure information through relevant library functions (such as the model.add() method or model.layers() method in the TensorFlow library in Python) to update the initial power load forecasting model. The model may include but is not limited to the number of network layers of the model and the number of neurons in each layer of the model.

[0093] The third step is to weight the updated initial power load forecasting model according to the above target model weight information to construct a multi-source power load forecasting model. In practice, the above execution subject can import the above target model weight information into the updated initial power load forecasting model through relevant library functions (such as the tf.keras.models.load() function in the TensorFlow library in Python) to weight the updated initial power load forecasting model and construct a multi-source power load forecasting model.

[0094] The fourth step is to bind the constructed multi-source power load forecasting model with the target data source tag. In practice, the execution subject may associate the target data source tag with the constructed multi-source power load forecasting model to bind the tag and the model.

[0095] Step 103, in response to determining that the current time satisfies a prediction period condition corresponding to a preset time period, obtaining a virtual power plant data group.

[0096] In some embodiments, in response to determining that the current time satisfies the prediction cycle condition corresponding to the preset time period, the execution subject may obtain a virtual power plant data group. The prediction cycle time condition corresponding to the preset time period may be that the current time is the cycle time. For example, the preset time period may be 1 hour. The initial time may be 8 o'clock, and the cycle time may be, but is not limited to, 9 o'clock, 10 o'clock, and 11 o'clock. The virtual power plant data in the virtual power plant data group may be virtual power plant data obtained after being collected by sensors and smart meters, and stored in the target database. Each virtual power plant data in the virtual power plant data group corresponds to a data source label. The data source label is similar to the data source label, and can characterize the data source of the corresponding virtual power plant data. The sensors include, but are not limited to, voltage sensors, current sensors, and electric power sensors. In practice, the user may obtain a virtual power plant data group from the target database.

[0097] Step 104 , generating various multi-source power load information according to the acquired virtual power plant data group and the constructed various multi-source power load prediction models.

[0098] In some embodiments, the execution subject may generate various multi-source power load information based on the acquired virtual power plant data group and the constructed multi-source power load forecasting models. The generated multi-source power load information may be a power load forecast value obtained by forecasting the energy consumption data at the corresponding power demand source (i.e., the data source represented by the data source tag).

[0099] In some optional implementations of some embodiments, the execution subject may generate various multi-source power load information according to the acquired virtual power plant data group and the constructed various multi-source power load prediction models through the following steps:

[0100] The first step is to preprocess the acquired virtual power plant data group to obtain a processed power plant data group. Among them, each processed power plant data in the above-mentioned processed power plant data group corresponds to a data source label. The above-mentioned data source label can characterize the data source of the above-mentioned virtual power plant data group. It should be noted that the implementation method of "preprocessing the acquired virtual computer data group to obtain a processed power plant data group" can refer to the above-mentioned implementation method of "preprocessing the various historical virtual power plant data included in the above-mentioned historical virtual power plant data group to obtain a processed virtual power plant data group", which will not be repeated here.

[0101] In the second step, for each processed power plant data in the above processed power plant data group, the following steps are performed:

[0102] In the first sub-step, the data source tag corresponding to the above-mentioned processed power plant data is determined as the target data source tag.

[0103] In the second sub-step, a multi-source power load forecasting model that satisfies the data source matching condition corresponding to the target data source tag is selected from the above-mentioned multi-source power load forecasting models as the target multi-source power load forecasting model. The above-mentioned data matching condition can be that the target data source tag is the same as the data source tag corresponding to the multi-source power load forecasting model.

[0104] The third sub-step is to input the processed power plant data into the target multi-source power load forecasting model to generate multi-source power load information.

[0105] Step 105: Perform energy dispatch according to the generated multi-source power load information.

[0106] In some embodiments, the above-mentioned execution subject may perform energy scheduling according to the generated multi-source power load information. Among them, each multi-source power load information generated may be the power load value of the corresponding power demand source (i.e., the data source represented by the corresponding data source tag) within a preset unit time. The default unit of the above-mentioned power load value may be kilowatt-hour. The corresponding power demand source (i.e., the data source represented by the corresponding data source tag) includes but is not limited to: steel mills, chemical plants and manufacturing plants. In practice, for each multi-source power load information in the above-mentioned multi-source power load information, the above-mentioned execution subject may multiply the power load value represented by the above-mentioned multi-source power load information by the above-mentioned preset unit time to determine the target power. Then, power equal to the above-mentioned target power may be dispatched to the power demand source corresponding to the above-mentioned multi-source power load information (i.e., the data source represented by the corresponding data source tag).

[0107] Optionally, the above virtual power plant energy scheduling method may further include the following steps:

[0108] The first step is to parse the load data visualization request in response to receiving the load data visualization request to obtain time range information. The load data visualization request may be a request for visualizing the predicted multi-source power load information. In practice, the execution entity may parse the load data visualization request by extracting the Jason attribute to obtain the time range information. The time range information may be information that characterizes the time range. For example, the time range information may be "2024-03-01—2024-03-10".

[0109] The second step is to obtain, from the target database, a set of multi-source power load information groups that meet the time range conditions corresponding to the time range information according to the time range information. The time range condition may be that the predicted time corresponding to the multi-source power load information in the target database is within the time range represented by the time range information. In practice, the execution subject may obtain each multi-source power load information that meets the time range condition from the target database, and then divide it according to the data source label corresponding to each multi-source power load information to obtain a set of multi-source power load information groups.

[0110] The third step is to visualize the obtained multi-source power load information group set on an associated display device. In practice, the execution subject may visualize each multi-source power load information group in the obtained multi-source power load information group set in a preset chart form and display it on an associated display device. The preset chart form may be a line chart. The display device may be a display screen.

[0111] The above-mentioned embodiments of the present disclosure have the following beneficial effects: the virtual power plant energy dispatching method of some embodiments of the present disclosure can improve the accuracy of load forecasting information, thereby reducing energy waste in the dispatching process. Specifically, the reason for the energy waste in the dispatching process is that the data in the virtual power plant is often collected from multiple types of data sources, and the generalization ability of the single neural network model is weak, resulting in a large error between the predicted energy load and the actual load, which leads to a large difference between the dispatched energy and the actual required energy, and multiple energy re-dispatching is still required, which leads to energy loss during multiple transmissions, resulting in energy waste. Based on this, the virtual power plant energy dispatching method of some embodiments of the present disclosure, first, obtains a historical virtual power plant data group set from the target database. Among them, each historical virtual power plant data group in the above historical virtual power plant data group set corresponds to a data source label. Thus, each virtual power plant data group collected from different data sources can be obtained. Then, for each historical virtual power plant data group in the above historical virtual power plant data group set, the following steps are performed: the first step is to perform data preprocessing on each historical virtual power plant data included in the above historical virtual power plant data group to obtain a processed virtual power plant data group. Therefore, the data quality of each historical virtual power plant data can be improved and noise interference can be reduced through data preprocessing. In the second step, according to the processed virtual power plant data group, the pre-set initial power load forecasting models are trained to generate each model weight information and each model structure information. Among them, the network structure of each initial power load forecasting model is different, and the weight information of each model corresponds to each model training result. Therefore, the generalization ability and prediction accuracy of each model structure for the current data source type can be determined by using the historical virtual power plant data group collected from a single data source to train models with different network structures. In the third step, according to the training results of each model, the weight information of each model and the structure information of each model, the target model weight information and the target model structure information are determined. Among them, the target data source label is the data source label corresponding to the virtual power plant data group. In practice, since different types of data sources have different index requirements for the model, the model weight information and the model structure information with a higher matching degree with the historical virtual power plant data group can be determined from the weight information of each model and the structure information of each model according to the index conditions corresponding to the target data source label, thereby improving the generalization ability of the model. The fourth step is to construct a multi-source power load forecasting model based on the determined target model weight information and target model structure information, and bind the determined multi-source power load forecasting model to the target data source label. In this way, a multi-source power load forecasting model with a better match with the data source represented by the above target data source label can be constructed, thereby improving the model prediction accuracy.Afterwards, in response to determining that the current time satisfies the prediction period condition corresponding to the preset time period, a virtual power plant data group is obtained. Next, based on the obtained virtual power plant data group and the constructed multi-source power load prediction models, each multi-source power load information is generated. Finally, energy scheduling is performed according to the generated multi-source power load information. Among them, each multi-source power load information generated can be the power load value of the corresponding power demand source within a preset unit time. The corresponding power demand sources include but are not limited to: steel mills, chemical plants, and manufacturing plants. Therefore, by performing energy scheduling with each multi-source power load information with higher accuracy, the problem of inaccurate energy scheduling caused by prediction errors can be reduced, thereby reducing the frequency of energy re-scheduling, and then reducing the loss in the energy transmission process, thereby reducing energy waste.

[0112] Further references Figure 2 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of a virtual power plant energy dispatching device. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the virtual power plant energy scheduling device can be specifically applied to various electronic devices.

[0113] like Figure 2As shown, the virtual power plant energy dispatching device 200 of some embodiments includes: a first acquisition unit 201, an execution unit 202, a second acquisition unit 203, a generation unit 204 and an energy dispatching unit 205. The acquisition unit 201 is configured to acquire a set of historical virtual power plant data groups from a target database, wherein each historical virtual power plant data group in the above historical virtual power plant data group set corresponds to a data source label; the execution unit 202 is configured to perform the following steps for each historical virtual power plant data group in the above historical virtual power plant data group set: perform data preprocessing on each historical virtual power plant data included in the above historical virtual power plant data group to obtain a processed virtual power plant data group; perform model training on each pre-set initial power load forecasting model according to the above processed virtual power plant data group to generate each model weight information and each model structure information, wherein the network structures of the above initial power load forecasting models are different, and the above each model weight information corresponds to each model training result; determine the target model weight information and Target model structure information; constructing a multi-source power load forecasting model according to the determined target model weight information and target model structure information, and binding the constructed multi-source power load forecasting model and the target data source tag, wherein the above-mentioned target data source tag is the data source tag corresponding to the above-mentioned historical virtual power plant data group; the second acquisition unit 203 is configured to acquire the virtual power plant data group in response to determining that the current time satisfies the prediction period condition corresponding to the preset time period; the generation unit 204 is configured to generate each multi-source power load information according to the acquired virtual power plant data group and each constructed multi-source power load forecasting model; the energy dispatching unit 205 is configured to perform energy dispatching according to each generated multi-source power load information, wherein each generated multi-source power load information can be the power load value of the corresponding power demand source within the preset unit time, and the corresponding power demand source includes but is not limited to: steel mills, chemical plants and manufacturing plants.

[0114] It can be understood that the units recorded in the virtual power plant energy dispatching device 200 are similar to those in the reference Figure 1 The various steps in the described method correspond to each other. Therefore, the operations, features and beneficial effects described above for the method are also applicable to the virtual power plant energy scheduling device 200 and the units contained therein, and will not be repeated here.

[0115] Reference below Figure 3 , which shows a structural schematic diagram of an electronic device 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0116] like Figure 3 As shown, the electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory 302 or a program loaded from a storage device 308 into a random access memory 303. Various programs and data required for the operation of the electronic device 300 are also stored in the random access memory 303. The processing device 301, the read-only memory 302, and the random access memory 303 are connected to each other via a bus 304. An input / output interface 305 is also connected to the bus 304.

[0117] Typically, the following devices may be connected to the I / O interface 305: input devices 306 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 308 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 309. The communication devices 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as required.

[0118] In particular, according to some embodiments of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from the network through the communication device 309, or installed from the storage device 308, or installed from the read-only memory 302. When the computer program is executed by the processing device 301, the above-mentioned functions defined in the method of some embodiments of the present disclosure are executed.

[0119] It should be noted that the computer-readable medium recorded in some embodiments of the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In some embodiments of the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0120] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (Hyper Text Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0121] The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or it may exist independently without being assembled into the electronic device. The above-mentioned computer-readable medium carries one or more programs. When the above-mentioned one or more programs are executed by the electronic device, the electronic device: obtains a set of historical virtual power plant data groups from the target database, wherein each historical virtual power plant data group in the above-mentioned set of historical virtual power plant data groups corresponds to a data source label; for each historical virtual power plant data group in the above-mentioned set of historical virtual power plant data groups, performs the following steps: performs data preprocessing on each historical virtual power plant data included in the above-mentioned historical virtual power plant data group to obtain a processed virtual power plant data group; performs model training on each pre-set initial power load forecasting model based on the above-mentioned processed virtual power plant data group to generate each model weight information and each model structure information, wherein the network structure of each of the above-mentioned initial power load forecasting models is different, and each of the above-mentioned model weight information corresponds to each model training result; based on each of the above-mentioned model training results, the above The target model weight information and the target model structure information are determined based on the weight information of each model and the structure information of each model; a multi-source power load forecasting model is constructed based on the determined target model weight information and target model structure information, and the constructed multi-source power load forecasting model is bound to the target data source tag, wherein the target data source tag is the data source tag corresponding to the historical virtual power plant data group; in response to determining that the current time satisfies the forecasting period condition corresponding to the preset time period, a virtual power plant data group is obtained; based on the obtained virtual power plant data group and the constructed multi-source power load forecasting models, each multi-source power load information is generated; energy scheduling is performed based on the generated multi-source power load information, wherein each multi-source power load information generated can be the power load value of the corresponding power demand source within a preset unit time, and the corresponding power demand source includes but is not limited to: steel mills, chemical plants and manufacturing plants.

[0122] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0123] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0124] The units described in some embodiments of the present disclosure may be implemented by software or by hardware. The described units may also be provided in a processor, for example, may be described as: a processor comprising a first acquisition unit, an execution unit, a second acquisition unit, and an energy scheduling unit. The names of these units do not, in some cases, constitute limitations on the units themselves, for example, the first acquisition unit may also be described as "a unit for acquiring a set of historical virtual power plant data groups from a target database".

[0125] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0126] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with the technical features with similar functions disclosed in the embodiments of the present disclosure (but not limited to) and the technical solutions formed.

Claims

1. A virtual power plant energy dispatching method, comprising: Acquire a historical virtual power plant data set from a target database, wherein each historical virtual power plant data set in the historical virtual power plant data set corresponds to a data source tag; For each historical virtual power plant data set in the historical virtual power plant data set, the following steps are performed: Performing data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group; According to the processed virtual power plant data group, model training is performed on each preset initial power load forecasting model to generate each model weight information and each model structure information, wherein the network structure of each initial power load forecasting model is different, and each model weight information corresponds to each model training result; Determine target model weight information and target model structure information according to the training results of each model, the weight information of each model and the structure information of each model; According to the determined target model weight information and target model structure information, a multi-source power load forecasting model is constructed, and the constructed multi-source power load forecasting model is bound to a target data source tag, wherein the target data source tag is a data source tag corresponding to the historical virtual power plant data group; In response to determining that the current time satisfies a prediction period condition corresponding to a preset time period, acquiring a virtual power plant data group; Generate various multi-source power load information according to the acquired virtual power plant data group and the constructed various multi-source power load prediction models; Energy scheduling is performed based on the generated multi-source power load information, wherein each generated multi-source power load information is the power load value of the corresponding power demand source within a preset unit time, and the corresponding power demand sources include: steel mills, chemical plants and manufacturing plants.

2. The method according to claim 1, wherein: The method further comprises: In response to receiving the load data visualization request, parsing the load data visualization request to obtain time range information; According to the time range information, acquiring from the target database a set of multi-source power load information groups that meet the time range condition corresponding to the time range information; The acquired multi-source power load information group set is visualized in an associated display device.

3. The method according to claim 1, wherein: The performing data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group includes: Performing abnormal data detection on each historical virtual power plant data included in the historical virtual power plant data group to generate each abnormal data label; Eliminate each historical virtual power plant data corresponding to each generated abnormal data tag from the historical virtual power plant data group to update the historical virtual power plant data group; Performing data cleaning on each historical virtual power plant data in the updated historical virtual power plant data group to obtain a cleaned historical virtual power plant data group; Resampling each historical virtual power plant data included in the updated historical virtual power plant data group to obtain a sampled historical virtual power plant data group; The various sampled virtual power plant data included in the sampled virtual power plant data group are subjected to noise reduction processing to obtain a processed virtual power plant data group.

4. The method according to claim 2, wherein: The historical virtual power plant data in the historical virtual power plant data set includes historical energy consumption data of users, historical operation data of transfer sources and historical power load data; and the model training is performed on each pre-set initial power load prediction model according to the processed virtual power plant data set to generate each model weight information and each model structure information, including: Determine the user's historical energy consumption data and the flow source historical operation data included in each processed virtual power plant data in the processed virtual power plant data group as historical virtual power plant sample data to obtain a historical virtual power plant sample data set; For each of the preset initial power load forecasting models, based on the historical virtual power plant sample data set, the following model training steps are performed: Inputting at least one historical virtual power plant sample data in the historical virtual power plant sample data set into the initial power load forecasting model to obtain forecasted power load data corresponding to each historical virtual power plant sample data in the at least one historical virtual power plant sample data input; Determine a model loss function value according to a preset loss function, each predicted power load data obtained, and each historical power load data corresponding to at least one historical virtual power plant sample data; Determining whether the initial power load forecasting model reaches a preset optimization target according to the determined model loss function value; In response to determining that the initial power load forecasting model achieves the preset optimization target, the following model adjustment steps are performed: determining the initial power load forecasting model as the power load forecasting model; Adjusting the power load forecasting model to generate model weight information and model structure information; In response to determining that the initial power load forecasting model does not achieve the preset optimization target, the model parameters of the initial power load forecasting model are adjusted, and unused historical virtual power plant sample data are used to form a historical virtual power plant sample data set, and the adjusted initial power load forecasting model is used as the initial power load forecasting model, and the model training step is performed again.

5. The method according to claim 4, wherein: Determining target model weight information and target model structure information according to the training results of each model, the weight information of each model and the structure information of each model includes: Selecting a model training result that meets the model training condition from the various model training results as the target model training result; Determine the model weight information corresponding to the target model training result as the target model weight information; The model structure information corresponding to the target model training result is determined as the target model structure information.

6. The method according to claim 1, wherein: The method of constructing a multi-source power load forecasting model according to the determined target model weight information and target model structure information, and binding the constructed multi-source power load forecasting model to the target data source label, comprises: Determining an initial power load forecasting model corresponding to the target model training result; According to the target model structure information, the initial power load forecasting model is structurally adjusted to update the initial power load forecasting model; According to the target model weight information, weighting processing is performed on the updated initial power load forecasting model to construct a multi-source power load forecasting model; The constructed multi-source power load forecasting model is bound to the target data source label.

7. The method according to claim 4, wherein: The generating of each multi-source power load information according to the acquired virtual power plant data group and each multi-source power load prediction model constructed includes: Preprocessing the acquired virtual power plant data group to obtain a processed power plant data group, wherein each processed power plant data in the processed power plant data group corresponds to a data source label; For each processed power plant data in the processed power plant data group, the following steps are performed: Determine the data source tag corresponding to the processed power plant data as the target data source tag; Selecting a multi-source power load forecasting model that meets the data source matching condition corresponding to the target data source label from the multi-source power load forecasting models as the target multi-source power load forecasting model; The processed power plant data is input into the target multi-source power load forecasting model to generate multi-source power load information.

8. A virtual power plant energy dispatching device, comprising: An acquisition unit is configured to acquire a set of historical virtual power plant data groups from a target database, wherein each historical virtual power plant data group in the set of historical virtual power plant data groups corresponds to a data source tag; The execution unit is configured to perform the following steps for each historical virtual power plant data group in the historical virtual power plant data group set: perform data preprocessing on each historical virtual power plant data included in the historical virtual power plant data group to obtain a processed virtual power plant data group; perform model training on each pre-set initial power load forecasting model according to the processed virtual power plant data group to generate each model weight information and each model structure information, wherein the network structure of each initial power load forecasting model is different, and each model weight information corresponds to each model training result; determine the target model weight information and the target model structure information according to each model training result, each model weight information and each model structure information; construct a multi-source power load forecasting model according to the determined target model weight information and target model structure information, and bind the constructed multi-source power load forecasting model to a target data source label, wherein the target data source label is a data source label corresponding to the historical virtual power plant data group; A second acquisition unit is configured to acquire a virtual power plant data group in response to determining that the current time satisfies a prediction period condition corresponding to a preset time period; A generating unit configured to generate various multi-source power load information according to the acquired virtual power plant data group and the constructed various multi-source power load prediction models; The energy dispatching unit is configured to perform energy dispatching according to the generated multi-source power load information, wherein each generated multi-source power load information is the power load value of the corresponding power demand source within a preset unit time, and the corresponding power demand sources include: steel mills, chemical plants and manufacturing plants.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images