Power distribution network planning decision optimization method and system based on big data analysis

Abstract

The invention discloses a big data analysis-based power distribution network planning decision optimization method and system, and relates to the technical field of power grids, and the method comprises the steps: constructing a system overall architecture, and carrying out the preprocessing of the constructed system overall architecture according to an improved deep learning preprocessing method; and a multi-source data fusion method is introduced, and the preprocessed system overall architecture is improved through a multi-objective optimization algorithm to complete power distribution network planning decision optimization. The training efficiency of the model is improved. The multi-target optimization algorithm processes multi-source data through a momentum modified Adam optimizer and an attention fusion network, and the feature expression ability and the model generalization ability are improved.

Description

Technical Field

[0001] The present invention relates to the field of power grid technology, and in particular to a distribution network planning decision optimization method and system based on big data analysis. Background Art

[0002] With the development of digital information technology, power grid companies continue to promote the construction, operation, and production technology upgrades of intelligent measurement and control universal terminals, with the overall goal of "improving quality and efficiency", and carry out intelligent and integrated construction of intelligent measurement and control universal terminals, giving full play to the role of data as a production factor in the substation field, and strongly supporting the upgrading of digital grids, digital enterprises, digital services, and digital industries. Among them, intelligent measurement and control universal terminals can include various types of power grid sensors.

[0003] At present, existing methods mainly rely on a single data source and fail to make full use of multi-source heterogeneous data. Traditional methods only use load data and ignore other key information. Existing technologies mostly use statistical analysis or simple machine learning methods. Traditional methods are difficult to handle nonlinear complex relationships, and decision optimization is not comprehensive. Existing planning methods often only consider a single goal and lack multi-objective collaborative optimization capabilities. Existing systems generally have data processing delays and are difficult to support rapid decision-making needs. Summary of the invention

[0004] In view of the problems existing in the existing distribution network planning decision optimization method and system based on big data analysis, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is that the existing technology mostly adopts statistical analysis or simple machine learning methods, and traditional methods are difficult to handle nonlinear complex relationships for solution.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, an embodiment of the present invention provides a distribution network planning decision optimization method based on big data analysis, which includes constructing a system overall architecture, and preprocessing the constructed system overall architecture according to an improved deep learning preprocessing method;

[0008] A multi-source data fusion method is introduced, and the overall system architecture after preprocessing is improved through a multi-objective optimization algorithm to complete the distribution network planning decision optimization.

[0009] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, wherein: the overall architecture of the construction system includes a data acquisition layer, a data preprocessing layer and an intelligent analysis layer;

[0010] The data collection layer includes a redundant design, configuring two sets of active and standby sensors at each collection point, using 5G+optical fiber dual channels, a bandwidth ≥10Gbps distributed storage system, a Hadoop cluster, and a storage capacity ≥500TB.

[0011] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, wherein: the data preprocessing layer includes a data cleaning unit, a feature extraction unit and a data fusion unit; the data cleaning unit includes support for outlier detection, missing value processing and noise filtering; the feature extraction unit includes the realization of time series feature, spatial feature and statistical feature extraction; the data fusion unit includes the completion of heterogeneous data alignment, standardization and fusion;

[0012] The intelligent analysis layer includes a deep learning modeling unit, a multi-objective optimization unit and a decision support unit; the deep learning modeling unit includes model training and optimization functions; the multi-objective optimization unit includes a multi-dimensional optimization function of the planning scheme; the decision support unit includes providing visual display and interactive operation functions.

[0013] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, the deep learning preprocessing method includes designing with a Hybrid-LSTM structure, and the specific formula is:

[0014]

[0015] Among them, N is the input feature dimension, each layer of LSTM adopts a bidirectional structure, Dropout prevents overfitting and BatchNormalization accelerates convergence;

[0016] The loss function formula of the deep learning preprocessing method is:

[0017] L=α(t)*MSE+β(t)*MAE+γ(t)*Huber

[0018] α((t)=0.5+0.3*exp(-t / T)

[0019] β(t)=0.3-0.1*exp(-t / T)

[0020] γ(t)=0.2-0.1*exp(-t / T)

[0021] Among them, t is the current training round, T is the total training round, and the weight coefficient is dynamically adjusted to adapt to the training process.

[0022] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, the improvement of the multi-objective optimization algorithm includes designing a momentum-corrected Adam optimizer, and the specific formula is:

[0023] m_t=β1*m_{t-1}+(1-β1)*g_t

[0024] v_t=β2*v_{t-1}+(1-β2)*g_t^2

[0025] m _t=m_t / (1-β1^t)

[0026] v _t=v_t / (1-β2^t)

[0027] η_t=η_0*(1+cos( π t / T)) / 2

[0028] θ_t=θ_{t-1}-η_t* m _t / sqrt ( v _t+ε)

[0029] Among them, β1=0.9, β2=0.999, ε=1e-8, η_0=0.001;

[0030] The method for introducing multi-source data fusion includes a feature extraction module and a feature fusion module;

[0031] The feature extraction module includes load data processing, equipment status data processing, geographic information processing and user behavior processing;

[0032] The feature fusion module includes designing an attention fusion network;

[0033] The formula for load data processing is:

[0034] Input→Conv1D(32,5)→ ReLU → MaxPool |→Conv1D(64,3)→ ReLU → MaxPool →Flatten→Dense(128)

[0035] The formula for processing the device status data is:

[0036] Input→ MultiHead (8) → LayerNorm

[0037] →FFN→ LayerNorm →Dense(128)

[0038] The formula for geographic information processing is:

[0039] Input→GCN(64)→ ReLU →GCN(32)→ ReLU → GlobalPool →Dense(128)

[0040] The formula for user behavior processing is:

[0041] Input→ BiGRU (128)→ BiGRU (64) → Attention → Dense (128)

[0042] Among them, W_q, W_k, W_v are learnable parameter matrices, d_k is the attention dimension, and n is the number of feature sources.

[0043] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, wherein: the completion of the distribution network planning decision optimization includes constructing a knowledge graph and performing configuration optimization on the constructed knowledge graph;

[0044] The configuration optimization includes making operation and maintenance decisions, and the operation and maintenance decisions include a state assessment model and a reinforcement learning strategy;

[0045] The formula of the state assessment model is:

[0046] lnput→CNN(32)→CNN(64)→ BiLSTM (128)→ BiLSTM (64) → Attention → Dense (1)

[0047] The formula for the branch and bound algorithm for the reinforcement learning strategy is:

[0048] State: [status, time, maintenance_history]|

[0049] Action: {inspect, replace, continue}

[0050] Reward: R=w1*cost+w2*time+w3*health

[0051] Q(s,a)=Q(s,a)+α[r+γ*max(Q(s′,a′))-Q(s,a)]

[0052] Among them, according to the operation and maintenance decision prediction function, each node is a feature-based test, which divides the data set into subsets ti (X strain ).

[0053] As a preferred solution of the distribution network planning decision optimization method based on big data analysis described in the present invention, the formula of the operation and maintenance decision prediction function is:

[0054]

[0055] Among them, K i represents the number of terminal nodes in the i-th decision tree, c ik represents the output value of the kth leaf node in the i-th decision tree, I is the indicator function, if X strain The region R that falls on the kth leaf node ik If it is within, it returns 1, otherwise it returns 0. ik Represents the input space region corresponding to the kth leaf node in the i-th decision tree, which is formed based on feature segmentation.

[0056] In the second aspect, an embodiment of the present invention provides a distribution network planning decision optimization system based on big data analysis, which includes: a construction module, which constructs the overall system architecture and preprocesses the constructed system overall architecture according to an improved deep learning preprocessing method; an optimization module, which introduces a multi-source data fusion method, improves the preprocessed system overall architecture through a multi-objective optimization algorithm, and completes the distribution network planning decision optimization.

[0057] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, any step of the above-mentioned distribution network planning decision optimization method based on big data analysis is implemented.

[0058] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned distribution network planning decision optimization method based on big data analysis is implemented.

[0059] The beneficial effects of the present invention are as follows: the present invention provides a distribution network planning decision optimization method and system based on big data analysis, which significantly improves the efficiency, accuracy and reliability of distribution network planning decisions by constructing an efficient data acquisition and processing architecture, a deep learning preprocessing method, multi-source data fusion and a multi-objective optimization algorithm. Specifically, redundant sensors and 5G+ optical fiber dual channels are used to ensure high reliability and high-speed transmission of data; the data preprocessing layer effectively handles outliers, missing values ​​and noise, extracts multi-dimensional features and completes the alignment and standardization of heterogeneous data, thereby improving data quality. The Hybrid-LSTM structure combines Dropout and Batch Normalization technology to prevent overfitting and accelerate model convergence, and the dynamically adjusted loss function weight coefficient adapts to the training process, thereby improving the training efficiency of the model. The multi-objective optimization algorithm processes multi-source data through the momentum-modified Adam optimizer and the attention fusion network, thereby improving the feature expression ability and model generalization ability. The knowledge graph is constructed and configuration optimization is performed, and the intelligent operation and maintenance decision is realized through the state evaluation model and reinforcement learning strategy, thereby improving the accuracy and real-time performance of the decision. These technical means jointly improve the scientificity and effectiveness of distribution network planning decisions and provide strong support for the stable operation of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0061] Figure 1 A flow chart of a distribution network planning decision optimization method and system based on big data analysis is provided for one embodiment of the present invention.

[0062] Figure 2 A schematic diagram of the overall method structure of a distribution network planning decision optimization method and system based on big data analysis is provided for one embodiment of the present invention.

[0063] Figure 3 A schematic diagram of a deep learning preprocessing method for a distribution network planning decision optimization method and system based on big data analysis provided by an embodiment of the present invention.

[0064] Figure 4 A schematic diagram of an improved multi-objective optimization algorithm for a distribution network planning decision optimization method and system based on big data analysis is provided as an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0066] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0068] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0069] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0070] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or 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 the present invention can be understood according to specific circumstances.

[0071] Example 1

[0072] Reference Figures 1 to 4, which is the first embodiment of the present invention, and provides a distribution network planning decision optimization method based on big data analysis, comprising:

[0073] S1: Construct the overall system architecture and preprocess the constructed system architecture according to the improved deep learning preprocessing method.

[0074] Among them, the overall architecture of the system includes data collection layer, data preprocessing layer and intelligent analysis layer;

[0075] The data collection layer includes a redundant design, configuring two sets of active and standby sensors at each collection point, using 5G+optical fiber dual channels, a bandwidth ≥10Gbps distributed storage system, a Hadoop cluster, and a storage capacity ≥500TB.

[0076] S1.1: The data preprocessing layer includes a data cleaning unit, a feature extraction unit, and a data fusion unit; the data cleaning unit includes support for outlier detection, missing value processing, and noise filtering; the feature extraction unit includes the realization of temporal features, spatial features, and statistical features extraction; the data fusion unit includes the completion of heterogeneous data alignment, standardization, and fusion;

[0077] The intelligent analysis layer includes a deep learning modeling unit, a multi-objective optimization unit and a decision support unit; the deep learning modeling unit includes model training and optimization functions; the multi-objective optimization unit includes the multi-dimensional optimization function of the planning scheme; the decision support unit includes providing visualization and interactive operation functions.

[0078] S1.2: The deep learning preprocessing method includes designing with a Hybrid-LSTM structure. The specific formula is:

[0079]

[0080] Among them, N is the input feature dimension, each layer of LSTM adopts a bidirectional structure, Dropout prevents overfitting and BatchNormalization accelerates convergence;

[0081] The loss function formula of the deep learning preprocessing method is:

[0082] L=α(t)*MSE+β(t)*MAE+γ(t)*Huber

[0083] α(t)=0.5+0.3*exp(-t / T)

[0084] β(t)=0.3-0.1*exp(-t / T)

[0085] γ(t)=0.2-0.1*exp(-t / T)

[0086] Among them, t is the current training round, T is the total training round, and the weight coefficient is dynamically adjusted to adapt to the training process.

[0087] S2: Introduce multi-source data fusion method, improve the overall system architecture after preprocessing through multi-objective optimization algorithm, and complete the distribution network planning decision optimization.

[0088] Among them, the improvement of the multi-objective optimization algorithm includes designing a momentum-corrected Adam optimizer. The specific formula is:

[0089] m_t=β1*m_{t-1}+(1-β1)*g_t

[0090] v_t=[β2*v_{t-1}+(1-β2)*g_t^2

[0091] m _t=m_t / (1-β1^t)

[0092] v _t=v_t / (1-β2^t)

[0093] η_t=η_0*(1+cos( πt / T)) / 2

[0094] θ_t=θ_{t-1}-η_t* m _t / sqrt ( v _t+ε)

[0095] Among them, β1=0.9, β2=0.999, ε=1e-8, η_0=0.001;

[0096] The introduction of multi-source data fusion method includes feature extraction module and feature fusion module;

[0097] The feature extraction module includes load data processing, equipment status data processing, geographic information processing and user behavior processing;

[0098] The feature fusion module includes designing an attention fusion network;

[0099] The formula for load data processing is:

[0100] Input→Conv1D(32,5)→ ReLU → MaxPool |→Conv1D(64,3)→ ReLU → MaxPool →Flatten→Dense(128)

[0101] The formula for equipment status data processing is:

[0102] Input→ MultiHead (8) → LayerNorm →FFN→ LayerNorm →Dense(128)

[0103] The formula for geographic information processing is:

[0104] Input→GCN(64)→ ReLU →GCN(32)→ ReLU → GlobalPool →Dense(128)

[0105] The formula for user behavior processing is:

[0106] Input→ BiGRU (128)→ BiGRU (64) → Attention → Dense (128)

[0107] Among them, W_q, W_k, W_v are learnable parameter matrices, d_k is the attention dimension, and n is the number of feature sources.

[0108] S2.1: Completing the optimization of distribution network planning decisions includes constructing a knowledge graph and optimizing the configuration of the constructed knowledge graph;

[0109] Configuration optimization includes making operation and maintenance decisions, which include state assessment models and reinforcement learning strategies;

[0110] The formula of the state assessment model is:

[0111] lnput→CNN(32)→CNN(64)→ BiLSTM (128)→ BiLSTM (64) → Attention → Dense (1)

[0112] The formula for the branch and bound algorithm for the reinforcement learning strategy is:

[0113] State: [status, time, maintenance_history]|

[0114] Action: {inspect, replace, continue}

[0115] Reward: R=w1*cost+w2*time+w3*health

[0116] Q(s,a)=Q(s,a)+α[r+γ*max(Q(s′,a′))-Q(s,a)]

[0117] Among them, according to the operation and maintenance decision prediction function, each node is a feature-based test, which divides the data set into subsets t i (X strain ).

[0118] Furthermore, the formula of the operation and maintenance decision prediction function is:

[0119]

[0120] Among them, K i represents the number of terminal nodes in the i-th decision tree, c ik represents the output value of the kth leaf node in the i-th decision tree, I is the indicator function, if X strain The region R that falls on the kth leaf node ik If it is within, it returns 1, otherwise it returns 0. ik Represents the input space region corresponding to the kth leaf node in the i-th decision tree, which is formed based on feature segmentation.

[0121] In a preferred embodiment, a distribution network planning decision optimization system based on big data analysis includes a construction module, which constructs the system overall architecture and preprocesses the constructed system overall architecture according to an improved deep learning preprocessing method; an optimization module, which introduces a multi-source data fusion method, improves the preprocessed system overall architecture through a multi-objective optimization algorithm, and completes the distribution network planning decision optimization.

[0122] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0123] In summary, the present invention significantly improves the efficiency, accuracy and reliability of distribution network planning decisions by constructing an efficient data acquisition and processing architecture, a deep learning preprocessing method, multi-source data fusion and a multi-objective optimization algorithm. Specifically, redundant sensors and 5G+ optical fiber dual channels are used to ensure high reliability and high-speed transmission of data; the data preprocessing layer effectively handles outliers, missing values ​​and noise, extracts multidimensional features and completes the alignment and standardization of heterogeneous data, thereby improving data quality. The Hybrid-LSTM structure combines Dropout and Batch Normalization techniques to prevent overfitting and accelerate model convergence, and the dynamically adjusted loss function weight coefficient adapts to the training process, thereby improving the training efficiency of the model. The multi-objective optimization algorithm processes multi-source data through the momentum-modified Adam optimizer and the attention fusion network, thereby improving the feature expression ability and model generalization ability. A knowledge graph is constructed and configuration optimization is performed, and the intelligent operation and maintenance decision-making is realized through the state evaluation model and reinforcement learning strategy, thereby improving the accuracy and real-time performance of the decision-making. These technical means jointly improve the scientificity and effectiveness of distribution network planning decisions, and provide strong support for the stable operation of the distribution network.

[0124] Example 2

[0125] Reference Figures 1 to 4 , which is the second embodiment of the present invention, and this embodiment provides a distribution network planning decision optimization method and system based on big data analysis. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0126] The overall architecture of the system is constructed, including the data collection layer, the data preprocessing layer and the intelligent analysis layer. The data collection layer uses redundant sensors and 5G+ optical fiber dual channels to ensure high reliability and high-speed transmission of data. The distributed storage system uses Hadoop clusters with a storage capacity of ≥500TB. The data preprocessing layer performs data cleaning, feature extraction and data fusion, effectively handles outliers, missing values ​​and noise, extracts multi-dimensional features and completes the alignment and standardization of heterogeneous data, and improves data quality by 20%. The intelligent analysis layer uses the Hybrid-LSTM structure for deep learning modeling, combined with Dropout and BatchNormalization technology to prevent overfitting and accelerate model convergence, reducing model training time by 30% and increasing convergence speed by 25%. The multi-objective optimization algorithm uses the momentum-modified Adam optimizer and attention fusion network to process multi-source data, and the optimization accuracy is improved by 15%. Finally, a knowledge graph is constructed and configuration optimization is performed. Through the state evaluation model and reinforcement learning strategy, the intelligent operation and maintenance decision is realized, and the accuracy of operation and maintenance decision-making reaches 95%, and the response time is shortened by 20%. The comparison between the present invention and the prior art is shown in Table 1 below:

[0127] Table 1 Comparison between the present invention and the prior art

[0128]

[0129] Table 1 shows the significant advantages of the present invention in improving the efficiency, accuracy and real-time performance of distribution network planning decisions, providing strong support for the stable operation of the distribution network.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution network planning decision optimization method based on big data analysis, characterized by: include, Constructing the overall system architecture, and preprocessing the constructed overall system architecture according to the improved deep learning preprocessing method; A multi-source data fusion method is introduced, and the overall system architecture after preprocessing is improved through a multi-objective optimization algorithm to complete the distribution network planning decision optimization.

2. The distribution network planning decision optimization method based on big data analysis according to claim 1, characterized in that: The overall architecture of the construction system includes a data acquisition layer, a data preprocessing layer and an intelligent analysis layer; The data collection layer includes a redundant design, configuring two sets of active and standby sensors at each collection point, using 5G+optical fiber dual channels, a bandwidth ≥10Gbps distributed storage system, a Hadoop cluster, and a storage capacity ≥500TB.

3. The distribution network planning decision optimization method based on big data analysis according to claim 2, characterized in that: The data preprocessing layer includes a data cleaning unit, a feature extraction unit and a data fusion unit; the data cleaning unit includes support for outlier detection, missing value processing and noise filtering; the feature extraction unit includes the realization of time series feature, spatial feature and statistical feature extraction; the data fusion unit includes the completion of heterogeneous data alignment, standardization and fusion; The intelligent analysis layer includes a deep learning modeling unit, a multi-objective optimization unit and a decision support unit; the deep learning modeling unit includes model training and optimization functions; the multi-objective optimization unit includes a multi-dimensional optimization function of the planning scheme; the decision support unit includes providing visual display and interactive operation functions.

4. The distribution network planning decision optimization method based on big data analysis according to claim 3 is characterized in that: The deep learning preprocessing method includes designing with a Hybrid-LSTM structure, and the specific formula is: Among them, N is the input feature dimension, each layer of LSTM adopts a bidirectional structure, Dropout prevents overfitting and BatchNormalization accelerates convergence; The loss function formula of the deep learning preprocessing method is: L=α(t)*MSE+β(t)*MAE+γ(t)*Huber α(t)0.5+0.3*exp(-t / T) β(t)=0.3-0.1*exp(-t / T) γ(t)=0.2-0.1*exp(-t / T) Among them, t is the current training round, T is the total training round, and the weight coefficient is dynamically adjusted to adapt to the training process.

5. The distribution network planning decision optimization method based on big data analysis according to claim 4, characterized in that: The improvement of the multi-objective optimization algorithm includes designing a momentum-corrected Adam optimizer, and the specific formula is: m_t=β1*m_{t-1}+(1-β1)*g_t v_t=β2*v_{t-1}+(1-β2)*g_t∧2 m_ t=m_t / (1-β1∧t) v_ t=v_t / (1-β2∧t) η_t=η_0*(1+cos( πt / T)) / 2 θ_t=θ_{t-1}-η_t* m_ t / sqrt(v _t+e) Among them, β1=0.9, β2=0.999, ε=1e-8, η_0=0.001; The method for introducing multi-source data fusion includes a feature extraction module and a feature fusion module; The feature extraction module includes load data processing, equipment status data processing, geographic information processing and user behavior processing; The feature fusion module includes designing an attention fusion network; The formula for load data processing is: Input→Conv1D(32,5)→ ReLU → MaxPool |→Conv1D(64,3))→ ReLU → MaxPool →Flatten→Dense(128) The formula for processing the device status data is: Input→ MultiHead (8)→ LayerNorm →FFN→ LayerNorm→ Dense(128) The formula for geographic information processing is: Input→GCN(64)→ ReLU →GCN(32)→ ReLU → GlobalPool →Dense(128) The formula for user behavior processing is: Input→ BiGRU (128)→ BiGRU (64)→Attention→Dense(128) Among them, W_q, W_k, W_v are learnable parameter matrices, d_k is the attention dimension, and n is the number of feature sources.

6. The distribution network planning decision optimization method based on big data analysis according to claim 5, characterized in that: The completion of the distribution network planning decision optimization includes constructing a knowledge graph and performing configuration optimization on the constructed knowledge graph; The configuration optimization includes making operation and maintenance decisions, and the operation and maintenance decisions include a state assessment model and a reinforcement learning strategy; The formula of the state assessment model is: Input→CNN(32)→CNN(64)→ BiLSTM (128)→ BiLSTM (64)→Attention→Dense(1) The formula for the branch and bound algorithm for the reinforcement learning strategy is: State:[status,time,maintenance_history] Action:{inspect,replace,continue} Reward:R=w1*cost+w2*time+w3*health Q(s,a)=Q(s,a)+α[r+γ*max(Q(s',a'))-Q(s,a)] Among them, according to the operation and maintenance decision prediction function, each node is a feature-based test, which divides the data set into subsets t i (X strain ).

7. The distribution network planning decision optimization method based on big data analysis according to claim 6, characterized in that: The formula of the operation and maintenance decision prediction function is: Among them, K i represents the number of terminal nodes in the i-th decision tree, c ik represents the output value of the kth leaf node in the i-th decision tree, I is the indicator function, if X strain The region Ri that falls on the kth leaf node k If it is within, it returns 1, otherwise it returns 0. ik Represents the input space region corresponding to the kth leaf node in the i-th decision tree, which is formed based on feature segmentation.

8. A distribution network planning decision optimization system based on big data analysis, based on the distribution network planning decision optimization method based on big data analysis according to any one of claims 1 to 7, characterized in that: include, A construction module, which constructs the overall system architecture and preprocesses the constructed overall system architecture according to the improved deep learning preprocessing method; The optimization module introduces a multi-source data fusion method, improves the overall system architecture after preprocessing through a multi-objective optimization algorithm, and completes the distribution network planning decision optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution network planning decision optimization method based on big data analysis described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distribution network planning decision optimization method based on big data analysis described in any one of claims 1 to 7 are implemented.

Images