A low-carbon configuration method for an ac-dc hybrid power distribution network

Abstract

The application discloses a low-carbon configuration method for an AC-DC hybrid power distribution network, and belongs to the field of power distribution network control, which comprises the following steps: obtaining operation information of an AC-DC hybrid power distribution network and obtaining the operation information based on time series; processing the operation information to obtain an operation information sequence; reducing the dimension of the operation information sequence to obtain key features of the AC-DC hybrid power distribution network; constructing a PSO-LSTM network model based on the operation information; predicting output data of the AC-DC hybrid power distribution network based on the PSO-LSTM network model; and configuring the AC-DC hybrid power distribution network based on the output data. The method solves the problems that the data analysis of the power distribution network is narrow and the precision is insufficient in the current AC-DC hybrid power distribution network, and the influence of the green certificate-carbon trading mechanism on the AC-DC hybrid power distribution network is not fully considered for analysis, thereby fully ensuring the improvement of the rationality of the configuration of the AC-DC hybrid power distribution network.

Description

Technical Field

[0001] This application belongs to the field of distribution network control, specifically, it relates to a low-carbon configuration method for AC / DC hybrid distribution networks. Background Technology

[0002] In current power distribution systems, to meet user demand, hybrid AC / DC distribution networks are being widely constructed. In these networks, AC and DC power are often generated by different subsystems, with the DC portion typically supplied by renewable energy systems. However, renewable energy systems are significantly affected by environmental factors, necessitating the ability to predict power output based on these environmental factors and subsequently optimize the distribution network. Current configuration of hybrid AC / DC distribution networks primarily focuses on achieving economic efficiency for the system or energy storage devices, largely neglecting the impact of green certificate-carbon trading mechanisms on distribution network profitability. Furthermore, the analysis of power output often relies directly on historical output data and the correlation with meteorological factors to establish verification data segments. This results in low prediction accuracy and a narrow optimal data range, failing to meet the actual configuration requirements of hybrid AC / DC distribution networks.

[0003] Therefore, how to establish a high-precision method for low-carbon configuration of AC / DC hybrid distribution networks with a wider range of analytical data and full consideration of the green certificate-carbon trading mechanism is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] To establish a high-precision method for low-carbon configuration of AC / DC hybrid distribution networks with a wider analytical data range and considering the green certificate-carbon trading mechanism, this application discloses a method for low-carbon configuration of AC / DC hybrid distribution networks, including:

[0005] A low-carbon configuration method for an AC / DC hybrid distribution network, the configuration method comprising:

[0006] Obtain operational information of the AC / DC hybrid distribution network, and acquire operational information based on time series data;

[0007] The operation information is processed to obtain an operation information sequence;

[0008] The operational information sequence is dimensionality reduced to obtain the key characteristics of the AC / DC hybrid distribution network;

[0009] Based on the aforementioned operational information, a PSO-LSTM network model is constructed;

[0010] Based on the PSO-LSTM network model, the power output data of the AC / DC hybrid distribution network is predicted;

[0011] Based on the output data, configure an AC / DC hybrid distribution network.

[0012] Optionally, obtaining the operation information of the AC / DC hybrid distribution network and obtaining the operation information based on time series includes:

[0013] Obtain the operation information of the AC / DC hybrid distribution network, and simultaneously obtain the time and environmental information corresponding to the operation information;

[0014] Based on the time corresponding to the operational information, the operational information and corresponding environmental information of the AC / DC hybrid distribution network are sorted to obtain operational information.

[0015] Optionally, processing the running information to obtain a running information sequence includes:

[0016] Obtain the extreme points in the running information sequence, establish upper and lower envelopes for the extreme points, and obtain the mean of the upper and lower envelopes to obtain the first component;

[0017] The first component is used as the extreme point on the new running information sequence, and the upper and lower envelopes are re-established. The mean of the upper and lower envelopes re-established based on the first component is obtained to obtain the second component.

[0018] Based on the acquisition methods of the first component and the second component, all components are obtained and the components are verified until the obtained components meet the IMF conditions.

[0019] Based on the obtained components, a sequence of operational information is acquired, and the equation of the operational information sequence is:

[0020] ;

[0021] in, x ( t ) represents the sequence of runtime information. d i ( t ) indicates the first i The intrinsic moduli variables obtained from the decomposition; r i ( t () represents the remaining components after decomposition satisfying the intrinsic mode function conditions; t Indicates the time parameter.

[0022] Optionally, the dimensionality reduction of the intrinsic mode functions to obtain the key characteristics of the AC / DC hybrid distribution network includes:

[0023] Based on the intrinsic mode functions, theoretical operating data of the AC / DC hybrid distribution network are obtained;

[0024] Based on nonlinear functions, the theoretical operating data is mapped to a high-dimensional space to obtain high-dimensional space data;

[0025] The high-dimensional spatial data is reduced in dimensionality using the PCA method to obtain the reduced-dimensional data.

[0026] Based on the dimensionality-reduced data, the eigenvalues ​​and eigenvectors of the dimensionality-reduced data are obtained, and the covariance and characteristic equation are as follows:

[0027] ;

[0028] in, Cov This represents the covariance of the data after dimensionality reduction. N This represents the total amount of data after dimensionality reduction; i This represents the index of the data after dimensionality reduction; φ ( x i () represents the data sample after dimensionality reduction; λ Represents the feature values ​​of the data after dimensionality reduction; ω Represents the feature vector of the data after dimensionality reduction; x i This represents the data obtained in the high-dimensional space;

[0029] Based on the aforementioned eigenvalues ​​and eigenvectors, key characteristics of the AC / DC hybrid distribution network are obtained.

[0030] Optionally, constructing the PSO-LSTM network model based on the operational information includes:

[0031] Pre-set the basic architecture and hyperparameters of the LSTM neural network to obtain a suitable activation function;

[0032] Parameters are set for the PSO algorithm, and the set parameters are the optimization starting point of the PSO algorithm;

[0033] Based on the training dataset, the setting parameters are mapped to the parameters of the LSTM neural network, and training is performed to obtain iterative training results;

[0034] Obtain the optimal particle position from the iterative training results, or, the PSO algorithm reaches the preset maximum number of iterations, to obtain the PSO-LSTM network model.

[0035] Optionally, the prediction of power output data for the AC / DC hybrid distribution network based on the PSO-LSTM network model includes:

[0036] Obtain the constraints at the AC / DC hybrid distribution network level;

[0037] Based on the hierarchical objectives of the AC / DC hybrid distribution network, the objective function is obtained. The upper-level objective function in the hierarchical objectives of the AC / DC hybrid distribution network is:

[0038] ;

[0039] ;

[0040] in, and These represent the investment costs of wind turbine units and photovoltaic systems, respectively. This indicates the investment cost of energy storage devices; , and These represent the investment cost per unit capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the installed capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the discount rates for wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the service life of the wind turbine, photovoltaic system, and energy storage device, respectively.

[0041] The operational layer objectives in the hierarchical objectives of the AC / DC hybrid distribution network include the overall system operating cost, voltage deviation, and green certificate-carbon trading cost;

[0042] Based on the constraints of the AC / DC hybrid distribution network hierarchy and the hierarchical objectives of the AC / DC hybrid distribution network, and based on the PSO-LSTM network model, the power output data of the AC / DC hybrid distribution network is predicted.

[0043] Optionally, the step of obtaining the output data of the AC / DC hybrid distribution network based on the constraints of the AC / DC hybrid distribution network hierarchy and the hierarchical objectives of the AC / DC hybrid distribution network, and based on the PSO-LSTM network model, includes:

[0044] Historical operating information and meteorological data of the AC / DC hybrid distribution network are obtained, and the predicted output of the subsystem is obtained based on the PSO-LSTM network model.

[0045] Based on the predicted output of the subsystem, the optimal solution for the AC / DC hybrid distribution network level is obtained;

[0046] Based on the optimal solution of the AC / DC hybrid distribution network hierarchy, the objective function of the upper layer and the objective function of the operation layer are solved iteratively to obtain the model after iterative processing;

[0047] The iteratively processed model is compared with the preset convergence condition. If the iteratively processed model satisfies the preset convergence condition, the iteratively processed model is a usable model. If the iteratively processed model does not satisfy the preset convergence condition, the iteratively processed model continues to iterate until the iteratively processed model satisfies the preset convergence condition and a usable model is obtained.

[0048] Based on the available models and meteorological data, the power output data of the AC / DC hybrid distribution network is predicted.

[0049] Optionally, configuring the AC / DC hybrid distribution network based on the output data includes:

[0050] Based on the output data, obtain the output data of all subsystems in the AC / DC hybrid distribution network;

[0051] Based on the output data of all the subsystems, the AC / DC hybrid distribution network is configured.

[0052] The beneficial effects of this application include:

[0053] 1. Improved configuration accuracy. In the technical solution of this application, the operating parameters obtained during the operation of the AC / DC hybrid distribution network are decomposed based on time, meteorological data, etc., and intrinsic mode functions are obtained. Then, an output prediction network model is established based on the intrinsic mode functions. After training with the intrinsic mode functions, the accuracy of the model itself is improved. In the subsequent prediction of data processing based on this model, the configuration accuracy can be significantly improved.

[0054] 2. Increased constraints. The technical solution of this application implements the setting of all constraints in the AC / DC hybrid distribution network, especially for the green certificate-carbon trading conditions. This enables the optimization configuration of the AC / DC hybrid distribution network based on the constraints during the prediction of the power output of the AC / DC hybrid distribution network, ensuring that the network can be optimized under any circumstances.

[0055] 3. Improved configuration efficiency. The technical solution of this application realizes the determination of all operating parameters in the obtained AC / DC hybrid distribution network. In this process, a PSO-LSTM network model is established for determination. Thus, in the subsequent power output prediction process, the power output of the entire AC / DC hybrid distribution network can be predicted directly based on the meteorological data and other constraint data, thereby significantly improving the prediction efficiency of the power output of the AC / DC hybrid distribution network. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the embodiments of this application or the prior art will be briefly introduced below. Obviously, the following description is only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings are used to provide a further understanding of this disclosure and constitute a part of the specification. They are used together with the following detailed description to explain this disclosure, but do not constitute a limitation of this disclosure. In the drawings:

[0057] Figure 1 A flowchart illustrating a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment;

[0058] Figure 2 A flowchart illustrating the power output prediction process of a low-carbon configuration method for an AC / DC hybrid power distribution network provided in this application embodiment;

[0059] Figure 3 A diagram of the distribution network node system structure of a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment;

[0060] Figure 4 A photovoltaic output prediction diagram in a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment;

[0061] Figure 5 This application provides a wind power output prediction diagram in a low-carbon configuration method for an AC / DC hybrid distribution network, as shown in the embodiments of this application.

[0062] Figure 6 This is a diagram showing the wind and solar power output results in a low-carbon configuration method for an AC / DC hybrid distribution network provided in an embodiment of this application.

[0063] Figure 7 This is a system operation diagram of a green certificate-carbon trading mechanism in a low-carbon configuration method for an AC / DC hybrid distribution network provided in an embodiment of this application. Detailed Implementation

[0064] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. Furthermore, in the embodiments of this application, "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0065] In the operation of AC / DC hybrid distribution networks, the configuration of each subsystem needs to be rationally configured based on specific operating scenarios and constraints. This requires comprehensive setting of various constraints. Current configuration schemes primarily employ methods based on historical data, segmenting the data and predicting output parameters for a given period. The network configuration is then based on these predictions. This method demands high data processing accuracy, but practice shows that accurate results are often difficult to obtain. Furthermore, this method has a relatively narrow optimal data processing range, making it unsuitable for all configuration requirements. Additionally, the need for large datasets and the lengthy data acquisition time ultimately lead to low overall system efficiency, making it difficult for existing technologies to achieve efficient data analysis results.

[0066] To address the problems existing in the prior art, this application discloses a low-carbon configuration method for AC / DC hybrid distribution networks, such as... Figure 1 The diagram shown is a flowchart of a low-carbon configuration method for an AC / DC hybrid distribution network provided in an embodiment of this application. Specifically:

[0067] S110. Obtain the operation information of the AC / DC hybrid distribution network and obtain the operation information based on the time series.

[0068] S120. The running information is processed to obtain a running information sequence.

[0069] S130. Dimensionally reduce the operation information sequence to obtain the key features of the AC / DC hybrid distribution network.

[0070] S140. Based on the aforementioned operational information, construct a PSO-LSTM network model.

[0071] S150. Based on the PSO-LSTM network model, predict the power output data of the AC / DC hybrid distribution network.

[0072] S160. Based on the output data, configure the AC / DC hybrid distribution network.

[0073] The purpose of the above steps is to enable the configuration of AC / DC hybrid distribution networks to be based entirely on the historical data currently available, and to perform dimensionality reduction operations on the data after high-dimensional processing. Based on the obtained key features, a PSO-LSTM network model is constructed, an accurate model is obtained based on this model, and output prediction is performed. Based on the accurate prediction results, the distribution network is configured.

[0074] The following is a detailed explanation of all the steps mentioned above:

[0075] As described in step S110, the purpose of this step is that, for AC / DC hybrid distribution networks, a large amount of data is generated during operation, and all data has corresponding time nodes. In other words, the generated operational data of the AC / DC hybrid distribution network inherently possesses operational time characteristics. Therefore, after obtaining operational information based on these time characteristics, the data can be used in subsequent feature extraction processes. Specifically:

[0076] Obtain the operation information of the AC / DC hybrid distribution network, and simultaneously obtain the time and environmental information corresponding to the operation information;

[0077] Based on the time corresponding to the operational information, the operational information and corresponding environmental information of the AC / DC hybrid distribution network are sorted to obtain operational information.

[0078] Specifically, for the obtained AC / DC hybrid distribution network operation information, the corresponding operation status data during the operation process is obtained.

[0079] In this process, for all the obtained operational data, a correspondence needs to be established between the obtained data and the corresponding time nodes, so as to build time series information based on the data.

[0080] As described in step S120, the purpose of this step is that the amount of data obtained from the AC / DC hybrid distribution network operation process is enormous, which places extremely high demands on the data acquisition and analysis process. Therefore, in order to reduce the data processing requirements, it is necessary to preprocess the data and highlight the local characteristics of the data at different time scales in order to understand the volatility, periodicity, and operating trends of the power grid operation. Specifically:

[0081] Obtain the extreme points in the running information sequence, establish upper and lower envelopes for the extreme points, and obtain the mean of the upper and lower envelopes to obtain the first component;

[0082] The first component is used as the extreme point on the new running information sequence, and the upper and lower envelopes are re-established. The mean of the upper and lower envelopes re-established based on the first component is obtained to obtain the second component.

[0083] Based on the acquisition methods of the first component and the second component, all components are obtained and the components are verified until the obtained components meet the IMF conditions.

[0084] Based on the obtained components, a sequence of operational information is acquired, and the equation of the operational information sequence is:

[0085] ;

[0086] in, x (t ) represents the sequence of runtime information. d i ( t ) indicates the first i The intrinsic moduli variables obtained from the decomposition; r i ( t () represents the remaining components after decomposition satisfying the intrinsic mode function conditions; t Indicates the time parameter.

[0087] The above steps involve decomposing the obtained AC / DC hybrid distribution network operation data, which is processed based on Empirical Mode Decomposition (EMD) technology.

[0088] The core principle of EMD is to decompose data into multiple Intrinsic Mode Functions (IMFs) based on the characteristic time scales of the data. Each IMF corresponds to the fluctuation characteristics at different time scales. For environmental time series with a certain degree of randomness and discontinuity, EMD decomposition can not only enrich the diversity of input variables but also highlight the local characteristics of the series at different time scales, thereby reflecting its volatility, periodicity, and trend changes. The specific decomposition steps are as follows:

[0089] 1) Identify the maximum and minimum points in the original data sequence and construct their upper and lower envelopes respectively. Calculate the mean of these two envelopes to obtain the first component.

[0090] 2) Treat the first component as new data and repeat the above steps to obtain the second component.

[0091] 3) Repeat this process until the components satisfy the IMF conditions, or the remaining components become monotonic functions.

[0092] 4) Finally, name the obtained components sequentially. and a remaining component The specific representation is shown in equation (1):

[0093] ;

[0094] In the formula: For data sequences; For the first The intrinsic moduli variables obtained from the decomposition; To decompose the remaining components after satisfying the conditions of the intrinsic mode function.

[0095] As described in step S130, the purpose of this step is that after data decomposition, the features of the data can be enriched, but the dimensionality of the data will increase significantly. This will lead to a decrease in the accuracy and computational speed of the model during model building, and will also easily result in overfitting. Therefore, in the specific processing, dimensionality reduction processing is also required, which is carried out using the KPCA method. Specifically:

[0096] Based on the intrinsic mode functions, theoretical operating data of the AC / DC hybrid distribution network are obtained;

[0097] Based on nonlinear functions, the theoretical operating data is mapped to a high-dimensional space to obtain high-dimensional space data;

[0098] The high-dimensional spatial data is reduced in dimensionality using the PCA method to obtain the reduced-dimensional data.

[0099] Based on the dimensionality-reduced data, the eigenvalues ​​and eigenvectors of the dimensionality-reduced data are obtained, and the covariance and characteristic equation are as follows:

[0100] ;

[0101] in, Cov This represents the covariance of the data after dimensionality reduction. N This represents the total amount of data after dimensionality reduction; i This represents the index of the data after dimensionality reduction; φ ( x i () represents the data sample after dimensionality reduction; λ Represents the feature values ​​of the data after dimensionality reduction; ω Represents the eigenvectors of the dimension-reduced matrix; x i This represents the data obtained in the high-dimensional space.

[0102] Based on the aforementioned eigenvalues ​​and eigenvectors, key characteristics of the AC / DC hybrid distribution network are obtained.

[0103] KPCA is a nonlinear data dimensionality reduction technique improved upon traditional Principal Component Analysis (PCA). It uses a nonlinear function to map data to a high-dimensional space, thereby reducing the nonlinearity of the data, and then performs PCA processing in that high-dimensional space. KPCA can preserve the global structure of the data and extract key features by maximizing the projected variance. Its main computational steps involve solving for the eigenvalues ​​and eigenvectors of the covariance matrix in the high-dimensional space.

[0104] In space K, the covariance and characteristic equation of the mapping matrix are expressed as:

[0105] ;

[0106] In the formula: φ ( x () represents the data sample after dimensionality reduction. λ This represents the feature value of the data after dimensionality reduction.

[0107] After transforming the above equation, let And introduce kernel functions The projection matrix of the mapping matrix onto space K can be obtained as follows:

[0108] ;

[0109] in, δ i Indicates the first i The proportion of high-dimensional data after dimensionality reduction δ i j Indicates the first i The proportion of data after dimensionality reduction of a high-dimensional data mapping matrix W i Indicates the first i A mapping matrix in space K The projection matrix.

[0110] As described in step S140, the purpose of this step is to process the acquired data after obtaining it. Considering that the method disclosed in this application requires data from sources with long time horizons, this type of data is clearly time-series data with long-term dependency characteristics. Therefore, in the specific processing, it is necessary to ensure that the established data can meet this characteristic, thus requiring model training. Specifically:

[0111] Pre-set the basic architecture and hyperparameters of the LSTM neural network to obtain a suitable activation function;

[0112] Parameters are set for the PSO algorithm, and the set parameters are the optimization starting point of the PSO algorithm;

[0113] Based on the training dataset, the setting parameters are mapped to the parameters of the LSTM neural network, and training is performed to obtain iterative training results;

[0114] Obtain the optimal particle position from the iterative training results, or, the PSO algorithm reaches the preset maximum number of iterations, to obtain the PSO-LSTM network model.

[0115] Long Short-Term Memory (LSTM) networks, as an improved model of recurrent neural networks, are particularly suitable for processing time-series data with long-term dependencies. However, traditional LSTM networks often rely on researchers' experience and repeated trials in hyperparameter settings and topology design. This approach is susceptible to subjective factors and makes it difficult to guarantee finding the optimal network parameter configuration. Therefore, this paper introduces the Particle Swarm Optimization (PSO) algorithm to automatically adjust the topology and hyperparameters of LSTM networks, aiming to improve the performance of LSTM networks when processing wind and solar data.

[0116] The following are the specific steps for constructing a PSO-LSTM network model:

[0117] 1) Model parameter initialization. First, the basic architecture and hyperparameters of the LSTM neural network are preset, specifically including determining the number of nodes in the input layer, hidden layer, and output layer, and selecting a suitable activation function. At the same time, key parameters for the PSO algorithm are set, such as population size, number of iterations, and spatial dimension, and initial particle positions are specified for variables such as the number of hidden layer nodes, learning rate, and number of network iterations, serving as the starting point for subsequent optimization processes.

[0118] 2) Fitness Evaluation and Model Training. Using the training dataset, the current position of the particles is mapped to the parameter configuration of the LSTM network for training. The root mean square error (RMSE) is used as the fitness evaluation metric to quantify the difference between the model's predicted output and the actual value, thereby measuring the model's prediction accuracy.

[0119] 3) Particle position update and iterative optimization. Based on the fitness evaluation results, the local optimum for each particle and the global optimum for the entire population are determined. Subsequently, the state update equation of the PSO algorithm is used to simulate the particle's motion path in space, thereby updating the particle's position and efficiently searching and optimizing the hyperparameters of the LSTM network.

[0120] 4) Iterative Optimization and Termination Judgment. The process of updating particle positions and evaluating fitness continues until the optimal particle position is found or the PSO algorithm reaches the preset maximum number of iterations. During the iteration process, the particle swarm gradually approaches the global optimum, while the hyperparameters and topology of the LSTM network are continuously optimized. When the termination condition is met, the optimization process ends, and the optimal network parameter configuration is output.

[0121] As described in step S150, the purpose of this step is to analyze the data obtained from the historical data and network model obtained during the configuration of the AC / DC hybrid distribution network, and then perform characteristic analysis on the operating parameters of the entire AC / DC hybrid distribution network to establish the output characteristic model of the entire distribution network. Specifically:

[0122] Obtain the constraints at the AC / DC hybrid distribution network level;

[0123] Based on the hierarchical objectives of the AC / DC hybrid distribution network, the objective function is obtained. The upper-level objective function in the hierarchical objectives of the AC / DC hybrid distribution network is:

[0124] ;

[0125] ;

[0126] in, and These represent the investment costs of wind turbine units and photovoltaic systems, respectively. This indicates the investment cost of energy storage devices; , and These represent the investment cost per unit capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the installed capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the discount rates for wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the service life of the wind turbine, photovoltaic system, and energy storage device, respectively. F 1 This indicates the output result from the upper layer;

[0127] The operational layer objectives in the hierarchical objectives of the AC / DC hybrid distribution network include the overall system operating cost, voltage deviation, and green certificate-carbon trading cost;

[0128] Based on the constraints of the AC / DC hybrid distribution network hierarchy and the hierarchical objectives of the AC / DC hybrid distribution network, and based on the PSO-LSTM network model, the power output data of the AC / DC hybrid distribution network is predicted.

[0129] like Figure 2The diagram shown is a flowchart of a low-carbon configuration method for AC / DC hybrid distribution networks provided in this application, illustrating the power output prediction process. First, EMD technology is used to decompose photovoltaic power output data into multiple intrinsic mode functions (IMFs). Then, the KPCA method is employed to reduce the dimensionality of these IMFs and extract key features. Finally, these features are input into an LSTM network optimized using the PSO algorithm to construct a wind and solar power output characteristic prediction model.

[0130] The specific construction process of the model can be summarized as follows:

[0131] 1) Data preprocessing: The collected photovoltaic power data and its corresponding environmental factor data are preliminarily cleaned to remove abnormal data points caused by equipment failure, etc.

[0132] 2) Empirical Mode Decomposition: The preprocessed environmental data sequence is input into the EMD algorithm to decompose the intrinsic mode functions at different frequencies and a residual term in order to capture the intrinsic characteristics of the data.

[0133] 3) Kernel Principal Component Analysis (KPCA) dimensionality reduction: The IMF components obtained from EMD decomposition are processed by KPCA. Through nonlinear transformation, high-dimensional data is mapped to a low-dimensional space, and key feature factors are extracted, thereby effectively reducing data dimensionality and eliminating redundant information.

[0134] 4) Dataset Construction and Partitioning: The KPCA-reduced data and historical photovoltaic power data were normalized to construct a dataset suitable for training the PSO-LSTM network. Subsequently, the dataset was divided into training and test sets according to a certain ratio for subsequent model training and validation.

[0135] 5) PSO-LSTM model training: Initialize the parameters of the LSTM model and input the training set data into the model for training. Utilize the PSO algorithm to optimize the hyperparameters of the LSTM model to improve its predictive performance.

[0136] 6) Model evaluation and testing: The trained PSO-LSTM model is evaluated using test set data. Evaluation metrics such as mean absolute error, root mean square error, mean absolute percentage error, and goodness of fit are calculated and output to verify the model's predictive performance and generalization ability.

[0137] For the established model, it is necessary to determine the objective function to obtain the information objects to be processed. Specifically, for the hierarchical objectives of the AC / DC hybrid distribution network, the objective function is obtained. The upper-level objective function in the hierarchical objectives of the AC / DC hybrid distribution network is:

[0138] ;

[0139] ;

[0140] in, and These represent the investment costs of wind turbine units and photovoltaic systems, respectively. This indicates the investment cost of energy storage devices; , and These represent the investment cost per unit capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the installed capacity of wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the discount rates for wind turbines, photovoltaic systems, and energy storage devices, respectively. , and These represent the service life of the wind turbine, photovoltaic system, and energy storage device, respectively. F 1 This indicates the output result from the upper layer;

[0141] The operational layer objectives of the AC / DC hybrid distribution network include the overall system operating cost, voltage deviation, and green certificate-carbon trading cost.

[0142] Among them, the constraints at the upper level include:

[0143] (2) Constraints

[0144] 1) Constraints on the installation location of wind, solar and energy storage systems

[0145] ;

[0146] In the formula, , , and These include wind turbines, photovoltaic systems, energy storage devices, and upstream grid connection nodes. For distribution network nodes.

[0147] 2) Configure capacity constraints

[0148] ;

[0149] In the formula, and These represent the maximum and minimum installed power of the wind turbine unit, respectively. and These represent the maximum and minimum installed power of the photovoltaic system, respectively. and These represent the maximum and minimum installed power of the energy storage device, respectively.

[0150] Among them, the constraints for the runtime layer include:

[0151] 1) Overall system operating cost

[0152] ;

[0153] ;

[0154] In the formula, The cost of purchasing electricity for the system; Cost of electricity loss; Costs associated with curtailing wind and solar power; Cost per unit of purchased electricity; , and These are the unit power generation costs for wind turbines, photovoltaics, and energy storage, respectively. , , and They are respectively Real-time power purchase capacity, wind turbine power generation capacity, photovoltaic power generation capacity, and energy storage power generation capacity; For the line Current; Line resistance; Cost per unit of wind and solar power discard; and The power discarded by wind and solar power within the time period; c LOSS This refers to the unit transmission loss cost of the line.

[0155] 2) Voltage deviation

[0156] ;

[0157] In the formula, and They are respectively Actual voltage and rated voltage of the node; V pc This indicates voltage deviation.

[0158] The obtained voltage deviation assessment results are used to perform precise discharge compensation operations.

[0159] 3) Green Certificates - Carbon Trading Costs

[0160] ;

[0161] ;

[0162] In the formula, For green certificate trading coefficient; for Line load power at any given time; The base price for carbon trading; and These are the carbon emission factors per unit of purchased power from the upstream power grid and the carbon emission allowance per unit of load, respectively.

[0163] After obtaining the model, it is necessary to acquire the output parameters of each subsystem in the AC / DC hybrid distribution network based on the established model. Subsequent adjustments can then be made based on these parameters. Determining the specific output parameters requires defining various objective functions and constraint functions to obtain accurate output data. Specifically:

[0164] Historical operating information and meteorological data of the AC / DC hybrid distribution network are obtained, and the predicted output of the subsystem is obtained based on the PSO-LSTM network model.

[0165] Based on the predicted output of the subsystem, the optimal solution for the AC / DC hybrid distribution network level is obtained;

[0166] Based on the optimal solution of the AC / DC hybrid distribution network hierarchy, the objective function of the upper layer and the objective function of the operation layer are solved iteratively to obtain the model after iterative processing;

[0167] The iteratively processed model is compared with the preset convergence condition. If the iteratively processed model satisfies the preset convergence condition, the iteratively processed model is a usable model. If the iteratively processed model does not satisfy the preset convergence condition, the iteratively processed model continues to iterate until the iteratively processed model satisfies the preset convergence condition and a usable model is obtained.

[0168] Based on the available models and meteorological data, the power output data of the AC / DC hybrid distribution network is predicted.

[0169] In addition, the constraints also include:

[0170] 1) Power balance constraints

[0171] Discussion Area:

[0172] ;

[0173] In the formula: for The set of parent nodes of a node; For A node is the set of end nodes of a branch that has a starting node; VSC AC side power; For the exchange area branch road The active power; For the exchange area branch road reactive power; This is a collection of branch roads in the communication area.

[0174] DC region:

[0175] ;

[0176] In the formula, For the exchange area branch road The active power; For VSC DC-side power; This is a collection of DC regional branches.

[0177] 2) Voltage and current constraints

[0178] ;

[0179] In the formula, and These are the upper and lower limits of the node voltage; and These are the upper and lower limits of the line current, respectively.

[0180] 3) Output constraints of distributed renewable energy sources

[0181] ;

[0182] In the formula, and These represent the maximum output values ​​of the wind turbine and the photovoltaic system, respectively.

[0183] 4) VSC constraints

[0184] ;

[0185] 5) Energy storage constraints

[0186] ;

[0187] In the formula, To improve the charging and discharging efficiency of energy storage devices; and Energy storage devices Capacity and output power at any given time; and These represent the maximum and minimum output power of the energy storage device, respectively. and These represent the maximum and minimum values ​​of the state of charge of the energy storage device, respectively. and This represents the initial state of charge and the final state of charge.

[0188] As described in step S160, the purpose of this step is to determine the output data of each subsystem in the AC / DC hybrid distribution network after obtaining the objective function and constraints. Only then can the AC / DC hybrid distribution network be configured based on this data. Specifically:

[0189] Based on the output data, obtain the output data of all subsystems in the AC / DC hybrid distribution network;

[0190] Based on the output data of all the subsystems, the AC / DC hybrid distribution network is configured.

[0191] After obtaining the objective function and constraints, it is necessary to solve the model of the entire system. Specifically:

[0192] 1) Input power distribution network parameters:

[0193] First, EMD and KPCA methods are used to process historical data and meteorological data of new energy sources such as wind power and photovoltaics. PSO-LSTM is used to predict the output of wind and solar power. The prediction results will serve as the data basis for the configuration model input.

[0194] 2) Initialize particle position and velocity:

[0195] For the upper-level model, the location and capacity of wind, solar, and energy storage are initialized as the position and velocity of particles. These parameters will serve as the initial values ​​for optimization of the upper-level model and as the basic input for optimization of the lower-level model. For the lower-level model, to meet the multi-objective optimization requirements of the lower-level model, the particle swarm of the MOPSO algorithm is initialized, with each particle representing a set of possible wind, solar, and energy storage output schemes.

[0196] 3) Optimization of lower-level models:

[0197] Based on multiple optimization objectives and constraints of the lower-level model, the MOPSO algorithm is used to optimize the wind-solar-storage capacity output in different time periods. The MOPSO algorithm can handle multiple conflicting objectives and find a set of Pareto optimal solutions. The optimization results will be fed back to the upper-level model, providing a basis for further optimization of the upper-level model.

[0198] 4) Upper-level model update:

[0199] The upper-level model receives the Pareto optimal solution set from the lower-level MOPSO algorithm. Based on the single-objective optimization goal and constraints of the upper-level model, it recalculates and selects the optimal wind-solar-storage configuration scheme. Using the PSO algorithm, the location and capacity of wind, solar, and storage systems are optimized and updated to approximate the optimal solution of the upper-level model.

[0200] 5) Update the particle population and iteratively solve:

[0201] For the upper-level model, the position and velocity of the particle population are continuously updated to gradually approach the global optimum. For the lower-level model, the particle population is also updated in each iteration to explore more Pareto optima. Steps 3 and 4, i.e., lower-level model optimization and upper-level model update, are repeated iteratively to solve the problem. Through parameter interaction and iterative optimization between the two models, the optimal wind-solar-storage configuration and system operation output strategy are gradually approximated.

[0202] 6) Convergence condition judgment:

[0203] If the preset convergence condition is met, the algorithm terminates and outputs the optimal wind-solar-storage configuration scheme and system operation output strategy. If the convergence condition is not met, the above steps are repeated, i.e., the particle population is updated and the solution is iterated, until the convergence condition is met.

[0204] To further illustrate the beneficial effects of this application, specific numerical examples are provided, specifically:

[0205] 1) Based on system nodes, configure a simulation model of AC / DC hybrid distribution network.

[0206] like Figure 3 The diagram shows the distribution network node system structure of a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment. The system allows 2-33 nodes to connect to wind turbines, photovoltaics, and energy storage, with a maximum installed power of 3000 kW. The particle swarm optimization algorithm for solving the model has a population size, storage pool size, and maximum number of iterations set to 100, 50, and 100, respectively. The economic service life of energy storage is 12 years, with a minimum and maximum state of charge (SOC) of 0.2 and 0.9, respectively. The cost per unit of charge / discharge power is 0.25 yuan / kW, and the energy storage operating loss and charge / discharge efficiency are set to 20% and 90%, respectively. The installation costs per unit capacity for wind turbines, photovoltaics, and energy storage are 8000 yuan / kW, 2000 yuan / kW, and 2000 yuan / kW, respectively. The system purchases electricity from the upstream grid using time-of-use pricing, with a wind power price of 0.3 yuan / kWh and a photovoltaic power price of 0.25 yuan / kWh.

[0207] 2) Wind and solar forecasting based on the acquired model

[0208] Three models were used for comparative prediction: a traditional LSTM neural network model, an EMD-LSTM neural network model, and an EMD-KPCA-LSTM prediction model. The data used for model prediction came from historical records of wind and solar phenomena and corresponding meteorological events at a wind and solar base in Gansu Province from September 1st to September 10th, 2022. The data was updated every 15 minutes, totaling 960 sets. The data was divided into training and testing sets in a 7:3 time series ratio, with 70% used for training and 30% for testing model performance. All data were normalized to eliminate the influence of differences in units of measurement on the results. Figure 4 The image shown is a photovoltaic output prediction diagram in a low-carbon configuration method for an AC / DC hybrid distribution network provided in an embodiment of this application; as shown... Figure 5 The figure shown is a wind power output prediction diagram in a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment. Further analysis reveals that the relevant models all show good fitting results with the actual values, as shown in Table 1.

[0209] ;

[0210] 3) Consider the output data of AC / DC hybrid distribution networks under the green certificate-carbon trading mechanism.

[0211] This section uses the wind and solar forecast results based on the aforementioned EMD-KPCA-PSO-LSTM model as data foundation, and designs several different planning schemes for systematic comparative analysis: 1) Scenario 1: Wind-solar-storage configuration method without considering green certificate-carbon trading. 2) Scenario 2: Wind-solar-storage configuration method considering green certificate trading. 3) Scenario 3: Wind-solar-storage configuration method considering carbon trading. 4) Scenario 4: Wind-solar-storage configuration method considering green certificate-carbon trading. The results are as follows: Figure 6 The diagram shown illustrates the system operation scheme of the green certificate-carbon trading mechanism in a low-carbon configuration method for an AC / DC hybrid distribution network provided in this application embodiment. The configuration scheme and results after introducing this condition are determined according to the following table:

[0212] ;

[0213] From a comprehensive perspective of economics and environmental protection, Scenario 4 demonstrates significant advantages. Its investment cost is 16.4% lower than Scenario 1, operating costs are reduced by 24.1%, and carbon emissions are sharply reduced by 30.3%. This is mainly due to the synergistic effect of dual market mechanisms: green certificate trading offsets part of the initial investment through renewable energy certificate revenue, while the carbon trading mechanism incentivizes the system to actively reduce carbon emissions through carbon emission quota trading. Both mechanisms jointly drive the system towards optimizing for low-capacity, high-efficiency configurations; for example, energy storage capacity is reduced by 40% compared to Scenario 3, but carbon emission efficiency is superior.

[0214] Scenario 2 and Scenario 3, both relying on single market mechanisms, exhibit a trade-off between the two. While Scenario 2, which only considers green certificate trading, achieves lower investment costs than Scenario 3 through a 1400kW photovoltaic configuration, the lack of carbon trading results in carbon emissions that are 18.8% higher than Scenario 4. Scenario 3, which only considers carbon trading, relies on 3000kW of large-scale energy storage for deep peak shaving during operation, reducing carbon emissions to 20.94 tons, but over-configuration leads to a 34.5% surge in investment costs compared to Scenario 4. Neither scenario overcomes the limitations of a single policy tool and fails to achieve the optimal balance between economic and environmental benefits.

[0215] From a reliability perspective, the voltage deviation indicators across different scenarios converge, indicating that the impact of node configuration on voltage stability has reached saturation. Scenario 4, while maintaining equivalent voltage quality, reduces energy storage capacity by 12.5% ​​compared to Scenario 1 by lowering wind and solar capacity configuration, achieving a 58.7% reduction in carbon reduction costs, demonstrating the market mechanism's economic constraint on technological solutions. This configuration optimization avoids the over-configuration trap of Scenario 3 while overcoming the limitations of insufficient energy storage capacity in Scenario 2, showcasing the value of multiple policy tools in enhancing overall system efficiency. Regarding the system operation scheme for Scenario 4, the results are as follows... Figure 7 The diagram shown is a system operation scheme diagram of the green certificate-carbon trading mechanism in a low-carbon configuration method for an AC / DC hybrid distribution network provided in an embodiment of this application.

[0216] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to computer program instructions. The aforementioned computer program can be stored in a non-volatile storage medium, and when executed, it performs the steps of the above method embodiments. Alternatively, if the integrated unit of the present invention is implemented as a software functional module and sold or used as an independent product, it can also be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the embodiments of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a non-volatile storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention.

[0217] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A low-carbon configuration method for an AC-DC hybrid power distribution network, characterized in that, The configuration method comprises: obtaining operation information of an AC-DC hybrid power distribution network, and obtaining the operation information based on a time sequence; processing the operation information to obtain an operation information sequence; dimensionally reducing the operation information sequence to obtain key features of the AC-DC hybrid power distribution network; based on the operation information, constructing a PSO-LSTM network model; based on the PSO-LSTM network model, predicting output data of the AC-DC hybrid power distribution network; based on the output data, configuring the AC-DC hybrid power distribution network; the PSO-LSTM network model, the output data of the AC-DC hybrid power distribution network, comprising: obtaining constraint conditions of the AC-DC hybrid power distribution network level; based on the AC-DC hybrid power distribution network level target, obtaining a target function, the upper target function in the AC-DC hybrid power distribution network level target is: ， ， wherein, and respectively represent the wind turbine unit and photovoltaic system investment cost; represents the energy storage device investment cost; , and respectively represent the wind turbine unit, photovoltaic system and energy storage device unit capacity investment cost; , and respectively represent the wind turbine unit, photovoltaic system and energy storage device installed capacity; , and respectively represent the wind turbine unit, photovoltaic system and energy storage device discount rate; , and respectively represent the wind turbine unit, photovoltaic system and energy storage device service life; the operation layer target in the AC-DC hybrid power distribution network level target includes system comprehensive operation cost, voltage deviation and green certificate-carbon trading cost; based on the constraint conditions of the AC-DC hybrid power distribution network level and the AC-DC hybrid power distribution network level target, and based on the PSO-LSTM network model, the output data of the AC-DC hybrid power distribution network is predicted; based on the constraint conditions of the AC-DC hybrid power distribution network level and the AC-DC hybrid power distribution network level target, and based on the PSO-LSTM network model, the output data of the AC-DC hybrid power distribution network is obtained, comprising: obtaining historical operation information and meteorological data of the AC-DC hybrid power distribution network, and obtaining subsystem predicted output based on the PSO-LSTM network model; based on the subsystem predicted output, obtaining the optimal solution of the AC-DC hybrid power distribution network level; based on the optimal solution of the AC-DC hybrid power distribution network level, the upper target function and the operation layer target function are iteratively solved to obtain an iteration processed model; comparing the iteration processed model with a preset convergence condition, if the iteration processed model meets the preset convergence condition, the iteration processed model is a usable model, if the iteration processed model does not meet the preset convergence condition, the iteration processed model is continuously iterated until the iteration processed model meets the preset convergence condition and a usable model is obtained; based on the usable model and the meteorological data, the output data of the AC-DC hybrid power distribution network is predicted.

2. The low-carbon configuration method of an AC / DC hybrid power distribution network according to claim 1, characterized in that, the operation information of the AC-DC hybrid power distribution network is obtained, and the time and environment information corresponding to the operation information is obtained at the same time; based on the time corresponding to the operation information, the operation information and the corresponding environment information of the AC-DC hybrid power distribution network are sorted to obtain the operation information. the operation information is processed to obtain an operation information sequence, comprising:

3. The low-carbon configuration method for AC / DC hybrid power distribution network according to claim 1, characterized in that, obtaining extreme points in the operation information sequence, establishing upper and lower envelope lines for the extreme points, and obtaining the mean value of the upper and lower envelope lines to obtain a first component; ​ The first component is taken as an extreme point on a new operation information sequence, and upper and lower envelope lines are re-established, and a mean value based on the re-established upper and lower envelope lines of the first component is obtained to obtain a second component; Based on the obtaining method of the first component and the second component, all components are obtained, and the components are verified until the obtained components meet the IMF condition; Based on the obtained components, an operation information sequence is obtained, and the operation information sequence equation is: ， wherein, x t represents a running information sequence, d i t represents the eigenmode variable obtained by decomposing the i r n t represents the remaining component after the decomposition satisfies the eigenmode function condition; t represents a time parameter.​​​​ 4. The low-carbon configuration method of an AC / DC hybrid power distribution network according to claim 1, characterized in that, The operation information sequence is reduced in dimension to obtain the key features of the AC-DC hybrid distribution network, including: Based on the eigenfunction, the theoretical operation data of the AC-DC hybrid distribution network are obtained; Based on the nonlinear function, the theoretical operation data are mapped to a high-dimensional space to obtain high-dimensional space data; Based on the PCA method, the high-dimensional space data are reduced in dimension to obtain reduced data; Based on the reduced data, the eigenvalues and eigenvectors of the reduced data are obtained, and the covariance and characteristic equation are: ， wherein, Cov represents the covariance of the reduced dimension data; N represents the total amount of the reduced dimension data; i represents the index of the reduced dimension data; φ x i represents the sample of the reduced dimension data; λ represents the eigenvalue of the reduced dimension data; ω represents the eigenvector of the reduced dimension data; x i represents the data in the high dimension space obtained;​ Based on the eigenvalues and eigenvectors, the key features of the AC-DC hybrid distribution network are obtained.

5. The low-carbon configuration method of an AC / DC hybrid power distribution network according to claim 1, characterized in that, The PSO-LSTM network model is constructed based on the operation information, including: The basic framework and hyperparameters of the LSTM neural network are preset to obtain a suitable activation function; The PSO algorithm is set with parameters, and the setting parameters are the optimization starting points of the PSO algorithm; Based on the training data set, the setting parameters are mapped to the parameters of the LSTM neural network, and training is performed to obtain an iterative training result; The optimal particle position in the iterative training result is obtained, or the PSO algorithm reaches a preset maximum number of iterations to obtain a PSO-LSTM network model.

6. The low-carbon configuration method of an AC / DC hybrid power distribution network according to claim 1, characterized in that, The AC-DC hybrid distribution network is configured based on the output data, including: Based on the output data, the output data of all subsystems in the AC-DC hybrid distribution network are obtained; Based on the output data of all subsystems, the AC-DC hybrid distribution network is configured.

Images