Benefit evaluation method and system for resource aggregation entities participating in power grid dispatching

Abstract

The present application discloses a method and system for evaluating the benefits of a resource aggregation subject participating in power grid dispatching, the method comprising: dividing the response delay data of an energy storage device in the resource aggregation subject into a plurality of time domain data blocks, and extracting data block features of each time domain data block respectively; extracting dead zone data features from control dead zone data of a power conversion device matched with the energy storage device; inputting the dead zone data features and data block features into a mapping network in a deep learning model to obtain mapping features corresponding to the dead zone data features; inputting the mapping features and response delay data into a coding and decoding network in a deep learning model to obtain frequency modulation capacity allocation information; based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid, evaluating the benefits of the energy storage device participating in the frequency modulation process of the target power grid, the benefit evaluation result indicating the frequency modulation capacity allocated to the energy storage device by the target power grid, which enables the frequency modulation of the target power grid to take into account the balance between the power grid frequency stability and the frequency modulation cost.

Description

Technical Field

[0001] The present application relates to the technical field of power grid dispatching, and in particular to a method and system for evaluating the benefits of resource aggregation entities participating in power grid dispatching. Background Art

[0002] When the energy storage unit receives a control instruction from the corresponding power conversion device, there is usually a certain delay in responding to the instruction. Part of this delay is usually caused by the existence of a control dead zone in the power conversion device. The control dead zone refers to the area in the control system where the control system will not respond when the change of the controlled quantity is within a certain range.

[0003] The response delay of the energy storage unit will affect the effect of the energy storage unit participating in the grid frequency regulation, and thus affect the frequency stability of the grid. In addition, the cost of using different types of energy storage units to regulate the grid frequency will also be different. Therefore, how to balance the relationship between grid frequency stability and frequency regulation costs has become a current technical difficulty. Summary of the invention

[0004] In order to solve the above technical problems, the embodiment of the present application proposes a benefit evaluation method and system for resource aggregation entities participating in power grid dispatching, which can obtain more accurate benefit evaluation results, so that the frequency regulation capacity ultimately allocated to the energy storage device by the target power grid can take into account the balance between power grid frequency stability and frequency regulation costs.

[0005] In a first aspect, an embodiment of the present application provides a method for evaluating the benefits of resource aggregation entities participating in power grid dispatching, including:

[0006] Dividing the response delay data of the energy storage device in the resource aggregation subject into a plurality of time domain data blocks, and extracting data block features of each of the time domain data blocks respectively;

[0007] Extracting dead zone data features from control dead zone data of a power conversion device matched with the energy storage device;

[0008] Inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature;

[0009] Inputting the mapping features and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device;

[0010] Based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, a benefit evaluation is performed on the energy storage device's participation in the frequency regulation process of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

[0011] Optionally, the mapping network comprises N mapping units electrically connected in sequence, and the mapping feature is the output of the Nth mapping unit;

[0012] The step of inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature includes:

[0013] Using a first mapping unit, performing feature mapping based on the dead zone data feature and the data block feature to obtain an output of the first mapping unit;

[0014] Using the nth mapping unit, feature mapping is performed based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit, wherein 2≤n≤N, and n and N are both positive integers.

[0015] Optionally, each mapping unit comprises at least a cross-attention layer;

[0016] The using the first mapping unit to perform feature mapping based on the dead zone data feature and the data block feature to obtain the output of the first mapping unit includes:

[0017] Using the cross attention layer in the first mapping unit, taking the dead zone data feature as the source sequence vector and the data block feature as the target sequence vector, performing attention calculation to complete feature mapping, and obtaining the output of the first mapping unit;

[0018] The using the nth mapping unit to perform feature mapping based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit includes:

[0019] Using the cross attention layer in the nth mapping unit, the dead zone data feature is used as the source sequence vector and the output of the n-1th mapping unit is used as the target sequence vector, and attention calculation is performed to complete the feature mapping to obtain the output of the nth mapping unit.

[0020] Optionally, the codec network includes a coding network and a decoding network, and the step of inputting the mapping feature and the response delay data into the codec network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device includes:

[0021] Inputting the mapping feature and the response delay data into the encoding network for encoding to obtain a fused data feature;

[0022] The fused data features are input into the decoding network for decoding to obtain the frequency modulation capacity allocation information corresponding to the energy storage device.

[0023] Optionally, the performing of benefit evaluation on the frequency regulation process of the energy storage device participating in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid includes:

[0024] Inputting the frequency modulation capacity allocation information and the dispatching operation data into a preset simulation model, so that the preset simulation model simulates the simulation data generated when the energy storage device is connected to the target power grid according to the frequency modulation capacity allocation information;

[0025] Based on the simulation data, a benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid.

[0026] Optionally, the performing benefit evaluation on the energy storage device participating in the frequency modulation process of the target power grid based on the simulation data includes:

[0027] According to the set power market transaction rules, determine the frequency regulation cost model corresponding to the target power grid, wherein the frequency regulation cost model is used to simulate the costs generated in the frequency regulation process, and the generated costs include the frequency regulation compensation costs of energy storage equipment;

[0028] The frequency modulation cost model is used to perform benefit evaluation based on the simulation data.

[0029] Optionally, the response delay data is at least used to characterize an average response delay of the energy storage device, wherein the energy storage device receives multiple power conversion instructions sent by the power conversion device within a historical period, and the average response delay is the average of the response time of the energy storage device to each of the power conversion instructions;

[0030] Before performing benefit evaluation on the energy storage device participating in the frequency modulation process of the target power grid based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid, the method further includes:

[0031] When the average response delay is greater than a set delay threshold, the frequency modulation capacity allocation information is updated at least according to the average response delay.

[0032] Optionally, updating the frequency modulation capacity allocation information at least according to the average response delay includes:

[0033] The frequency modulation capacity allocation information is updated based on the difference between the average response delay and the set delay threshold.

[0034] Optionally, the larger the difference is, the smaller the frequency regulation capacity allocated to the energy storage device indicated by the updated frequency regulation capacity allocation information is.

[0035] In a second aspect, an embodiment of the present application provides a benefit evaluation system for resource aggregation entities participating in power grid dispatching, including:

[0036] A data block feature acquisition module, used to divide the response delay data of the energy storage device in the resource aggregation subject into multiple time domain data blocks, and respectively extract the data block features of each of the time domain data blocks;

[0037] A dead zone data feature acquisition module, used to extract dead zone data features from control dead zone data of a power conversion device matched with the energy storage device;

[0038] A mapping feature acquisition module, used for inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature;

[0039] A frequency modulation capacity allocation information acquisition module, used to input the mapping features and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device;

[0040] A benefit evaluation module is used to perform a benefit evaluation on the energy storage device's participation in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

[0041] In summary, the embodiments of the present application have at least the following beneficial effects:

[0042] According to the embodiment of the present application, since the control dead zone data of the power conversion device can be used to characterize the factors affecting the response delay of the energy storage device, the dead zone data characteristics can be used as a prompt information, so that the deep learning model can more accurately predict the frequency regulation capacity allocation information related to the energy storage device based on the response delay data on the basis of relevant prompts, so as to further use the frequency regulation capacity allocation information and the scheduling operation data to obtain a more accurate benefit evaluation result, so that the frequency regulation capacity finally allocated to the energy storage device by the target power grid can take into account the balance between the power grid frequency stability and the frequency regulation cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for evaluating the benefits of resource aggregation entities participating in power grid dispatching provided in an embodiment of the present application;

[0044] Figure 2 It is a structural diagram of a benefit evaluation system for resource aggregation entities participating in power grid dispatching provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0046] In the description of the present application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, features defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the features. In the description of the present application, unless otherwise specified, the meaning of "multiple" is two or more. In the description of the present application, the term "including" and its variations are open inclusions, i.e., "including but not limited to". The term "based on" means "at least partially based on". The term "according to" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments".

[0047] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0048] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0049] First, see Figure 1 , shows a schematic flow chart of a method for evaluating the benefits of resource aggregation entities participating in power grid dispatching provided by an embodiment of the present application, the method comprising steps S101-S105, which are as follows:

[0050] S101, dividing the response delay data of the energy storage device in the resource aggregation subject into a plurality of time domain data blocks, and extracting data block features of each of the time domain data blocks respectively;

[0051] In one example, the data block features of each of the time domain data blocks may be extracted by a feature extraction model, wherein the feature extraction model may be constructed based on an autoregressive moving average model, a deep learning model and / or a wavelet transform.

[0052] In an example, the extracted data block features may be expressed as any one of time domain features, frequency domain features, and time-frequency domain features. In this case, the expression form of the dead zone data features described below is the same as the expression form of the extracted data block features.

[0053] In actual use, since the acquired response delay data is usually historical data within a longer historical period, this embodiment divides the response delay data into multiple time domain data blocks, which can facilitate the subsequent processing of the deep learning model and reduce the computational complexity of the model operation. Furthermore, a deep learning model composed of a lightweight network structure can also be selected.

[0054] S102, extracting dead zone data features from control dead zone data of a power conversion device matched with the energy storage device;

[0055] In one example, the dead zone data feature can also be extracted by the above-mentioned feature extraction model, which will not be described in detail here. In addition, the feature extraction model is further described, and the feature extraction of data of different modalities can be realized by configuring encoding layers of different modalities in the feature extraction model.

[0056] It is worth mentioning that the relationship between energy storage devices and power conversion devices is a very important part of modern power systems and energy management. Energy storage technology can improve the efficiency, reliability and flexibility of power systems in many aspects, and power conversion devices are the key components connecting energy storage devices and power grids.

[0057] In this way, the power conversion device in this embodiment can be used to connect the energy storage device and the target power grid, and the power conversion device can control the energy storage device to perform power conversion, thereby controlling the energy storage device to participate in the frequency modulation process of the target power grid.

[0058] S103, inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature;

[0059] In one example, the mapping network may include a zero convolution layer and a feature adapter, so that in S103, the data block features and the dead zone data features can be input into the zero convolution layer for zero convolution processing; the data block features and the dead zone data features after zero convolution processing are superimposed; the superimposed features are input into the feature adapter for feature mapping; the features obtained after feature mapping are input into the zero convolution layer for zero convolution processing to obtain mapping features.

[0060] S104, inputting the mapping features and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device;

[0061] In one example, the codec network can be used to: encode the response delay data to extract the response delay data features from the response delay data; compress the mapping features and the response delay data features into a low-dimensional feature vector space to complete the fusion and obtain low-dimensional features; restore the fused low-dimensional features into output data, thereby realizing the prediction of frequency modulation capacity allocation information.

[0062] S105, based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, a benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

[0063] In one example, the frequency regulation capacity allocation information can be modified according to the benefit evaluation result, so that the modified frequency regulation capacity allocation information takes into account cost-effectiveness, so that the target power grid allocates corresponding frequency regulation capacity to the energy storage device according to the modified frequency regulation capacity allocation information.

[0064] In an example, all or part of the features in each embodiment of the present application may be expressed in the form of corresponding feature vectors.

[0065] In an optional implementation, the mapping network includes N mapping units electrically connected in sequence, and the mapping feature is the output of the Nth mapping unit.

[0066] It can be understood that, among the N mapping units in this embodiment, the output end of the (n-1)th mapping unit is electrically connected to the input end of the nth mapping unit, thereby forming N mapping units that are electrically connected in sequence.

[0067] The step of inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature includes:

[0068] Using a first mapping unit, performing feature mapping based on the dead zone data feature and the data block feature to obtain an output of the first mapping unit;

[0069] Using the nth mapping unit, feature mapping is performed based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit, wherein 2≤n≤N, and n and N are both positive integers.

[0070] In one example, the mapping network may be an adapter network, and thus the N mapping units may be N adapter units, each of which may be used to perform feature mapping.

[0071] At this time, the adapter network may include: a lower projection layer, a non-linear activation function, and an upper projection layer.

[0072] Down-projection layer: maps the input features to a lower dimensional space.

[0073] Non-linear activation function: Introduces non-linear transformation to enhance the expressive power of the model.

[0074] Up-projection layer: maps features from low-dimensional space back to high-dimensional space.

[0075] In an optional embodiment, each mapping unit includes at least a cross attention layer;

[0076] The using the first mapping unit to perform feature mapping based on the dead zone data feature and the data block feature to obtain the output of the first mapping unit includes:

[0077] Using the cross attention layer in the first mapping unit, taking the dead zone data feature as the source sequence vector and the data block feature as the target sequence vector, performing attention calculation to complete feature mapping, and obtaining the output of the first mapping unit;

[0078] The using the nth mapping unit to perform feature mapping based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit includes:

[0079] Using the cross attention layer in the nth mapping unit, the dead zone data feature is used as the source sequence vector and the output of the n-1th mapping unit is used as the target sequence vector, and attention calculation is performed to complete the feature mapping to obtain the output of the nth mapping unit.

[0080] In one example, the target sequence vector may be a query vector, and the source sequence vector may be a key-value pair vector, that is, in the cross-attention layer in the first mapping unit, the dead zone data feature is used as the key-value pair vector, and the data block feature is used as the query vector, so that the attention mechanism can be used to effectively calculate the correlation between the data block feature and the dead zone data feature to complete the feature mapping and obtain the final mapping feature.

[0081] It should be noted that the cross-attention mechanism allows the model to refer to the information of another modality when processing data of one modality. In this way, the cross-attention layer in this embodiment is a variant of the attention mechanism, in which the query vector comes from one modality (such as data block features) as the target sequence vector, and the key-value pair vector comes from another modality (such as dead zone data features) as the source sequence vector, so that the dead zone data features can be fully utilized as prompt information.

[0082] In an optional implementation, the codec network includes a coding network and a decoding network, and the mapping feature and the response delay data are input into the codec network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device, including:

[0083] Inputting the mapping feature and the response delay data into the encoding network for encoding to obtain a fused data feature;

[0084] The fused data features are input into the decoding network for decoding to obtain the frequency modulation capacity allocation information corresponding to the energy storage device.

[0085] In one example, the encoding network can be used to encode input mapping features and response delay data into a low-dimensional feature vector (i.e., fused data features), wherein the encoding network may include a first input layer, a first hidden layer, and a first output layer electrically connected in sequence.

[0086] First input layer: receives mapping features and response delay data.

[0087] First hidden layer: multiple fully connected layers or convolutional layers to extract high-level features.

[0088] First output layer: output low-dimensional fused data features.

[0089] In one example, the decoding network can be used to decode the fused data features output by the encoding network into frequency modulation capacity allocation information of the energy storage device, thereby achieving the effect of predicting the frequency modulation capacity allocation information of the energy storage device. The decoding network includes a second input layer, a second hidden layer, and a second output layer that may be electrically connected in sequence.

[0090] The second input layer: receives the fused data features output by the encoding network.

[0091] Second hidden layer: multiple fully connected layers or convolutional layers to recover high-level features.

[0092] Second output layer: output frequency modulation capacity allocation information.

[0093] In an optional implementation, the benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, including:

[0094] The frequency modulation capacity allocation information and the scheduling operation data are input into a preset simulation model, so that the preset simulation model simulates the simulation data generated when the energy storage device is connected to the target power grid according to the frequency modulation capacity allocation information.

[0095] In one example, a preset simulation model can be used to simulate the behavior of energy storage devices when they are connected to the target power grid according to the frequency modulation capacity allocation information. Here, a physical model, a data-driven model, or a hybrid model can be used for simulation. Among them, the physical model is used to simulate the dynamic behavior of the power grid based on the physical laws and mathematical models of the power system, the data-driven model is used to train the machine learning model using historical data to predict the dynamic behavior of the power grid, and the hybrid model combines the physical model and the data-driven model to improve the accuracy and robustness of the simulation.

[0096] In one example, the frequency regulation capacity allocation information is used as input to tell the simulation model the frequency regulation capacity of the energy storage device in different time periods, and the scheduling operation data is used as input to provide real-time status information of the power grid, including load, frequency deviation, generator set status, etc. The simulation model is then run to simulate the behavior of the energy storage device when it is connected to the target power grid according to the frequency regulation capacity allocation information, and generate simulation data.

[0097] In one example, the simulation data includes at least one of the following:

[0098] Frequency response: The change in grid frequency over time.

[0099] Load balancing: The change of grid load over time.

[0100] Generator set status: the start, stop and running status of the generator set.

[0101] Energy storage device status: the charge and discharge status and remaining capacity of the energy storage device.

[0102] Based on the simulation data, a benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid.

[0103] In one example, the benefit evaluation performed may include at least part of the following 1-4:

[0104] 1. Economic benefit evaluation

[0105] Frequency regulation service revenue: Revenue is calculated based on the frequency regulation service volume provided by the energy storage equipment and market pricing.

[0106] Operating costs: including charging and discharging losses of energy storage equipment, operation and maintenance costs, etc.

[0107] Opportunity cost: The opportunity cost of using energy storage equipment for frequency regulation services, that is, the benefits that could be obtained if it were used for other purposes.

[0108] 2. Technical performance evaluation

[0109] Response speed: the response time of the energy storage device to the frequency modulation command.

[0110] Accuracy: The deviation between the output power of the energy storage device and the command power.

[0111] Reliability: Failure rate and availability of energy storage equipment during frequency regulation.

[0112] 3. Environmental Impact Assessment

[0113] Carbon emission reduction: Reduce carbon emissions by reducing the number of starts and stops of fossil fuel power generation units.

[0114] Utilization rate of renewable energy: Increase the proportion of intermittent renewable energy such as wind power and solar power connected to the grid.

[0115] 4. Grid stability assessment

[0116] Frequency deviation: The fluctuation of grid frequency after energy storage equipment participates in frequency modulation.

[0117] Frequency recovery time: the time required for the grid frequency to recover from deviation to normal range.

[0118] In an optional implementation, the performing of benefit evaluation on the energy storage device participating in the frequency regulation process of the target power grid based on the simulation data includes:

[0119] According to the set power market transaction rules, determine the frequency regulation cost model corresponding to the target power grid, wherein the frequency regulation cost model is used to simulate the costs generated in the frequency regulation process, and the generated costs include the frequency regulation compensation costs of energy storage equipment;

[0120] The frequency modulation cost model is used to perform benefit evaluation based on the simulation data.

[0121] In one example, the power market trading rules can be set to at least indicate the calculation method of the unit frequency regulation capacity cost generated when a type / one energy storage device participates in the frequency regulation process of the target power grid. In this case, the frequency regulation cost model can be a function model constructed according to the calculation method. In addition, the frequency regulation cost model can also be a model obtained by training a neural network model based on historical market trading data, wherein the historical market trading data is historical data collected according to the power market trading rules.

[0122] In an optional embodiment, the response delay data is at least used to characterize the average response delay of the energy storage device, wherein the energy storage device receives multiple power conversion instructions sent by the power conversion device within a historical period, and the average response delay is the average of the response time of the energy storage device to each of the power conversion instructions; wherein the power conversion instruction is used to instruct the energy storage device to perform power conversion, thereby adjusting the frequency modulation capacity of the energy storage device.

[0123] Before performing benefit evaluation on the energy storage device participating in the frequency modulation process of the target power grid based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid, the method further includes:

[0124] When the average response delay is greater than a set delay threshold, the frequency modulation capacity allocation information is updated at least according to the average response delay.

[0125] In an optional implementation, updating the frequency modulation capacity allocation information at least according to the average response delay includes:

[0126] The frequency modulation capacity allocation information is updated based on the difference between the average response delay and the set delay threshold.

[0127] In an optional implementation, the larger the difference is, the smaller the frequency regulation capacity allocated to the energy storage device indicated by the updated frequency regulation capacity allocation information is.

[0128] In one example, a pre-trained benefit evaluation model may be used in S105 to perform benefit evaluation. The training process of the benefit evaluation model may include the following 1-6:

[0129] 1. Data preparation

[0130] 1.1 Collecting Data

[0131] Frequency regulation capacity allocation information: including the frequency regulation capacity of energy storage equipment in different time periods.

[0132] Grid dispatching and operation data: including real-time load of the grid, frequency deviation, operating status of generator sets, etc.

[0133] Energy storage equipment parameters: such as charging and discharging efficiency, response speed, cycle life, etc.

[0134] Economic parameters: such as electricity prices, frequency regulation service fees, investment costs of energy storage equipment, etc.

[0135] Historical benefit evaluation data: If there is historical data, it can be used for model training and verification.

[0136] 1.2 Data Preprocessing

[0137] Cleaning data: removing missing values, outliers, and noise.

[0138] Normalization / Standardization: Scale the data to be in the same range so that the model converges better.

[0139] Feature engineering: extract useful features, such as time features, frequency features, etc.

[0140] 2. Feature Selection

[0141] Select characteristics related to benefit evaluation, including at least one of the following: frequency regulation capacity, grid load, frequency deviation, generator set status, energy storage equipment parameters, economic parameters, and historical benefit evaluation results.

[0142] 3. Model selection

[0143] Choose an appropriate machine learning or deep learning model. Common models include:

[0144] Linear regression: Applicable to simple linear relationships.

[0145] Decision trees and random forests: suitable for nonlinear relationships and can handle complex interaction effects.

[0146] Support Vector Machines: Suitable for high-dimensional data.

[0147] Neural Networks: Suitable for complex nonlinear relationships, especially deep neural networks, convolutional neural networks, and recurrent neural networks.

[0148] 4. Model Training

[0149] 4.1 Dividing the Dataset

[0150] Divide the data into training set, validation set and test set. The usual ratio is 70% training set, 15% validation set, and 15% test set.

[0151] 4.2 Training Model

[0152] Select loss function: Choose an appropriate loss function based on the task type, such as mean square error for regression tasks.

[0153] Select optimizer: Commonly used optimizers include Adam and others.

[0154] Set hyperparameters: such as learning rate, batch size, number of training rounds, etc.

[0155] Training process: Use the training set data to train the model, regularly evaluate the model performance on the validation set, and adjust the hyperparameters.

[0156] 5. Model Evaluation

[0157] 5.1 Validation Set Evaluation

[0158] Evaluate the performance of the model on the validation set. Common evaluation indicators include:

[0159] Mean square error;

[0160] Coefficient of determination (R²);

[0161] Mean absolute error;

[0162] 5.2 Test Set Evaluation

[0163] The final performance of the model is evaluated on the test set to ensure that the model has good generalization ability.

[0164] 6. Model Optimization

[0165] 6.1 Hyperparameter Tuning

[0166] Tune hyperparameters using methods such as grid search, random search, or Bayesian optimization to improve model performance.

[0167] 6.2 Feature Selection and Engineering

[0168] Further optimize feature selection and feature engineering, remove irrelevant or redundant features, and add useful features.

[0169] On the second aspect, accordingly, the embodiments of the present application also provide a benefit evaluation system for resource aggregation entities participating in power grid dispatching, which can implement all processes of the benefit evaluation method for resource aggregation entities participating in power grid dispatching provided in the above embodiments.

[0170] See also Figure 2 , shows a schematic diagram of the structure of a benefit evaluation system for resource aggregation entities participating in power grid dispatching provided in an embodiment of the present application, the benefit evaluation system for resource aggregation entities participating in power grid dispatching includes:

[0171] The data block feature acquisition module 201 is used to divide the response delay data of the energy storage device in the resource aggregation subject into multiple time domain data blocks, and respectively extract the data block features of each of the time domain data blocks;

[0172] A dead zone data feature acquisition module 202, used to extract dead zone data features from control dead zone data of a power conversion device matched with the energy storage device;

[0173] A mapping feature acquisition module 203 is used to input the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature;

[0174] The frequency modulation capacity allocation information acquisition module 204 is used to input the mapping characteristics and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device;

[0175] The benefit evaluation module 205 is used to perform a benefit evaluation on the energy storage device's participation in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

[0176] In an optional implementation, the mapping network includes N mapping units electrically connected in sequence, and the mapping feature is the output of the Nth mapping unit.

[0177] The step of inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature includes:

[0178] Using a first mapping unit, performing feature mapping based on the dead zone data feature and the data block feature to obtain an output of the first mapping unit;

[0179] Using the nth mapping unit, feature mapping is performed based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit, wherein 2≤n≤N, and n and N are both positive integers.

[0180] In an optional embodiment, each mapping unit includes at least a cross attention layer;

[0181] The using the first mapping unit to perform feature mapping based on the dead zone data feature and the data block feature to obtain the output of the first mapping unit includes:

[0182] Using the cross attention layer in the first mapping unit, taking the dead zone data feature as the source sequence vector and the data block feature as the target sequence vector, performing attention calculation to complete feature mapping, and obtaining the output of the first mapping unit;

[0183] The using the nth mapping unit to perform feature mapping based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit includes:

[0184] Using the cross attention layer in the nth mapping unit, the dead zone data feature is used as the source sequence vector and the output of the n-1th mapping unit is used as the target sequence vector, and attention calculation is performed to complete the feature mapping to obtain the output of the nth mapping unit.

[0185] In an optional implementation, the codec network includes a coding network and a decoding network, and the mapping feature and the response delay data are input into the codec network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device, including:

[0186] Inputting the mapping feature and the response delay data into the encoding network for encoding to obtain a fused data feature;

[0187] The fused data features are input into the decoding network for decoding to obtain the frequency modulation capacity allocation information corresponding to the energy storage device.

[0188] In an optional implementation, the benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, including:

[0189] Inputting the frequency modulation capacity allocation information and the dispatching operation data into a preset simulation model, so that the preset simulation model simulates the simulation data generated when the energy storage device is connected to the target power grid according to the frequency modulation capacity allocation information;

[0190] Based on the simulation data, a benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid.

[0191] In an optional implementation, the performing of benefit evaluation on the energy storage device participating in the frequency regulation process of the target power grid based on the simulation data includes:

[0192] According to the set power market transaction rules, determine the frequency regulation cost model corresponding to the target power grid, wherein the frequency regulation cost model is used to simulate the costs generated in the frequency regulation process, and the generated costs include the frequency regulation compensation costs of energy storage equipment;

[0193] The frequency modulation cost model is used to perform benefit evaluation based on the simulation data.

[0194] In an optional implementation, the response delay data is at least used to characterize the average response delay of the energy storage device, wherein the energy storage device receives multiple power conversion instructions sent by the power conversion device within a historical period, and the average response delay is the average of the response time of the energy storage device to each of the power conversion instructions;

[0195] The system further comprises:

[0196] A frequency modulation capacity allocation information updating module is used to update the frequency modulation capacity allocation information at least according to the average response delay before performing a benefit evaluation on the energy storage device participating in the frequency modulation process of the target power grid based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid, when the average response delay is greater than a set delay threshold.

[0197] In an optional implementation, updating the frequency modulation capacity allocation information at least according to the average response delay includes:

[0198] The frequency modulation capacity allocation information is updated based on the difference between the average response delay and the set delay threshold.

[0199] In an optional implementation, the larger the difference is, the smaller the frequency regulation capacity allocated to the energy storage device indicated by the updated frequency regulation capacity allocation information is.

[0200] In summary, the embodiments of the present application have at least the following beneficial effects:

[0201] According to the embodiment of the present application, since the control dead zone data of the power conversion device can be used to characterize the factors affecting the response delay of the energy storage device, the dead zone data characteristics can be used as a prompt information, so that the deep learning model can more accurately predict the frequency regulation capacity allocation information related to the energy storage device based on the response delay data on the basis of relevant prompts, so as to further use the frequency regulation capacity allocation information and the scheduling operation data to obtain a more accurate benefit evaluation result, so that the frequency regulation capacity finally allocated to the energy storage device by the target power grid can take into account the balance between the power grid frequency stability and the frequency regulation cost.

[0202] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary hardware platform, and of course, it can also be implemented entirely by hardware. Based on such an understanding, all or part of the contribution of the technical solution of the present application to the background technology can be embodied in the form of a software product, and the computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application or some parts of the embodiments.

[0203] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications are also considered to be within the scope of protection of the present application.

Claims

1. A method for evaluating the benefits of resource aggregation entities participating in power grid dispatching, characterized in that: include: Dividing the response delay data of the energy storage device in the resource aggregation subject into a plurality of time domain data blocks, and extracting data block features of each of the time domain data blocks respectively; Extracting dead zone data features from control dead zone data of a power conversion device matched with the energy storage device; Inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature; Inputting the mapping features and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device; Based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, a benefit evaluation is performed on the energy storage device's participation in the frequency regulation process of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

2. The benefit evaluation method of resource aggregation subject participating in power grid dispatching according to claim 1 is characterized in that: The mapping network comprises N mapping units electrically connected in sequence, and the mapping feature is the output of the Nth mapping unit; The step of inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature includes: Using a first mapping unit, performing feature mapping based on the dead zone data feature and the data block feature to obtain an output of the first mapping unit; Using the nth mapping unit, feature mapping is performed based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit, wherein 2≤n≤N, and n and N are both positive integers.

3. The benefit evaluation method of resource aggregation subject participating in power grid dispatching according to claim 2 is characterized in that: Each mapping unit includes at least a cross-attention layer; The using the first mapping unit to perform feature mapping based on the dead zone data feature and the data block feature to obtain the output of the first mapping unit includes: Using the cross attention layer in the first mapping unit, taking the dead zone data feature as the source sequence vector and the data block feature as the target sequence vector, performing attention calculation to complete feature mapping, and obtaining the output of the first mapping unit; The using the nth mapping unit to perform feature mapping based on the dead zone data feature and the output of the n-1th mapping unit to obtain the output of the nth mapping unit includes: Using the cross attention layer in the nth mapping unit, the dead zone data feature is used as the source sequence vector and the output of the n-1th mapping unit is used as the target sequence vector, and attention calculation is performed to complete the feature mapping to obtain the output of the nth mapping unit.

4. The benefit evaluation method of resource aggregation subject participating in power grid dispatching according to claim 1 is characterized in that: The codec network includes a coding network and a decoding network. The mapping feature and the response delay data are input into the codec network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device, including: Inputting the mapping feature and the response delay data into the encoding network for encoding to obtain a fused data feature; The fused data features are input into the decoding network for decoding to obtain the frequency modulation capacity allocation information corresponding to the energy storage device.

5. The benefit evaluation method of resource aggregation subject participating in power grid dispatching according to claim 1 is characterized in that: The benefit evaluation of the energy storage device participating in the frequency modulation process of the target power grid based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid includes: Inputting the frequency modulation capacity allocation information and the dispatching operation data into a preset simulation model, so that the preset simulation model simulates the simulation data generated when the energy storage device is connected to the target power grid according to the frequency modulation capacity allocation information; Based on the simulation data, a benefit evaluation is performed on the energy storage device participating in the frequency regulation process of the target power grid.

6. The benefit evaluation method of resource aggregation subject participating in power grid dispatching according to claim 5 is characterized in that: The benefit evaluation of the energy storage device participating in the frequency modulation process of the target power grid based on the simulation data includes: According to the set power market transaction rules, determine the frequency regulation cost model corresponding to the target power grid, wherein the frequency regulation cost model is used to simulate the costs generated in the frequency regulation process, and the generated costs include the frequency regulation compensation costs of energy storage equipment; The frequency modulation cost model is used to perform benefit evaluation based on the simulation data.

7. The benefit evaluation method for resource aggregation entities participating in power grid dispatching according to any one of claims 1 to 6, characterized in that: The response delay data is at least used to characterize the average response delay of the energy storage device, wherein the energy storage device receives multiple power conversion instructions sent by the power conversion device within a historical period, and the average response delay is the average of the response time of the energy storage device to each of the power conversion instructions; Before performing benefit evaluation on the energy storage device participating in the frequency modulation process of the target power grid based on the frequency modulation capacity allocation information and the dispatching operation data of the target power grid, the method further includes: When the average response delay is greater than a set delay threshold, the frequency modulation capacity allocation information is updated at least according to the average response delay.

8. The benefit evaluation method for resource aggregation entities participating in power grid dispatching according to claim 7 is characterized in that: The updating of the frequency modulation capacity allocation information at least according to the average response delay comprises: The frequency modulation capacity allocation information is updated based on the difference between the average response delay and the set delay threshold.

9. The benefit evaluation method for resource aggregation entities participating in power grid dispatching according to claim 8 is characterized in that: The larger the difference is, the smaller the frequency regulation capacity allocated to the energy storage device indicated by the updated frequency regulation capacity allocation information is.

10. A benefit evaluation system for resource aggregation entities participating in power grid dispatching, characterized in that: include: A data block feature acquisition module, used to divide the response delay data of the energy storage device in the resource aggregation subject into multiple time domain data blocks, and respectively extract the data block features of each of the time domain data blocks; A dead zone data feature acquisition module, used to extract dead zone data features from control dead zone data of a power conversion device matched with the energy storage device; A mapping feature acquisition module, used for inputting the dead zone data feature and the data block feature into a mapping network in a deep learning model to obtain a mapping feature corresponding to the dead zone data feature; A frequency modulation capacity allocation information acquisition module, used to input the mapping features and the response delay data into the encoding and decoding network in the deep learning model to obtain the frequency modulation capacity allocation information corresponding to the energy storage device; A benefit evaluation module is used to perform a benefit evaluation on the energy storage device's participation in the frequency regulation process of the target power grid based on the frequency regulation capacity allocation information and the dispatching operation data of the target power grid, wherein the benefit evaluation result is used to indicate the frequency regulation capacity allocated to the energy storage device by the target power grid.

Images