A data-driven distribution network operation characteristic simulation method

Abstract

The present application relates to power distribution network technical field, especially to a kind of data-driven distribution network operating characteristic simulation method.The present application includes the following steps: S1, the acquisition and preprocessing of multi-source data;S2, the self-adaptive fusion of preprocessed data is formed into data set;S3, initial simulation model is constructed, training is carried out based on data set, and main simulation model is obtained;The prediction result of distribution network characteristic is output by main simulation model;S4, based on the prediction result, dynamic checking and simulation optimization are carried out, and simulation result is output.The present application covers multi-source data fusion, deep learning model construction, dynamic checking and adaptive optimization and other technical modules in work.Through the fusion of real-time load data, equipment operation data and environment data related multi-source information, combined with dynamic adjustment and incremental learning mechanism, high-precision prediction and optimization of distribution network operating state are realized.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and in particular to a data-driven method for simulating the operation characteristics of power distribution networks. Background Technology

[0002] As distribution networks continue to expand and become more complex, traditional distribution network optimization methods have shown significant limitations in handling large-scale data, real-time scheduling, and precise control.

[0003] Traditional optimization methods are mostly based on linear models and classical algorithms. While they are suitable for small-scale distribution networks, they often fail to provide timely response and efficient regulation for complex and dynamically changing modern distribution networks, especially when dealing with large amounts of variable real-time data.

[0004] If the following problems exist:

[0005] 1) Inaccurate modeling: Static data models cannot reflect the real-time dynamic changes in load, equipment status, and environmental conditions;

[0006] 2) Low scheduling efficiency: It relies on manual labor, and the manual decision-making process is long and complex, making it difficult to deal with emergencies such as high load and equipment failure.

[0007] 3) High operating costs: Single-objective optimization leads to low line losses and low utilization of backup power. Summary of the Invention

[0008] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0009] The technical solution of this invention is: a data-driven method for simulating the operation characteristics of a distribution network, comprising the following steps:

[0010] S1. Acquisition and preprocessing of multi-source data;

[0011] S2. Adaptively fuse the preprocessed data to form a dataset;

[0012] S3. Construct an initial simulation model, train it based on the dataset, and obtain the main simulation model; output the prediction results of the distribution network characteristics through the main simulation model;

[0013] S4. Based on the prediction results, perform dynamic verification and simulation optimization to output the simulation results.

[0014] In step S1,

[0015] Multi-source data includes real-time load data, equipment operation data, and environmental data;

[0016] The collected multi-source data is cleaned and standardized, including missing value imputation, data denoising, and time alignment.

[0017] Real-time load data includes the power change trend of each node;

[0018] Equipment operating data includes transformer status, voltage, and current;

[0019] Environmental data includes meteorological information and market fluctuations. Meteorological information includes temperature and humidity; market fluctuations include electricity price fluctuations, electricity demand fluctuations, and electricity supply fluctuations.

[0020] In step S2, the weights of each data dimension are dynamically adjusted according to the importance of the data, and a unified distribution network operation dataset is generated by merging them.

[0021] The importance of the data includes historical correlation, real-time volatility, and operational status sensitivity.

[0022] Historical Relevance i The following formula is used:

[0023]

[0024] In the formula, X i : Value of the i-th data dimension; Y: Key indicator of operational status; Cov(X) i (Y): X i The covariance of Y; σ Xi and σ Y :X i and the standard deviation of Y;

[0025] Real-time volatility var i The following formula is used:

[0026]

[0027] In the formula, X i,j : The value of the i-th data dimension in the j-th time window; The average value of the i-th data dimension within the current window; N: sliding window size;

[0028] Operating state sensitivity i The following formula is used:

[0029]

[0030] In the formula, Derivative function.

[0031] Dynamically adjust the weights of each data dimension, including: when the weight of a certain data source changes, the weights of other data sources need to be redistributed proportionally to keep the total weight of 1.

[0032] Let w1, w2, and w3 correspond to the weights of real-time load data, equipment operation data, and environmental data, respectively. When w1 is adjusted, w2 and w3 are adjusted according to the following formula:

[0033] w1′=w1·(1+Δ1)

[0034]

[0035] In the formula,

[0036] w1′: Adjusted load weight;

[0037] w2′ and w3′: Reassigned device operating weights and environment weights;

[0038] Δ1: The percentage increase in load weight due to fluctuations.

[0039] The data dimensions are merged according to their weights to generate a unified power distribution network operation dataset, as follows:

[0040]

[0041] In the formula, D fused : The merged, integrated dataset; X i : The value of the i-th data dimension; w i : Weight of the i-th data dimension; n: Number of data points.

[0042] In step S3,

[0043] An initial simulation model based on a deep neural network is constructed, and the main simulation model is obtained after training. The formula of the main simulation model is:

[0044] F = W (L) ·f(W (L-1) ·f(···f(W (L) ·X+b (L) )···)

[0045] +b (L-1) )b (L)

[0046] In the formula, F: main simulation model function; X: input feature matrix; W (l) W: The weight matrix of the l-th layer; (L-1) b: The weight matrix of the (l-1)th layer; (l) : The bias vector of the l-th layer; b (L-1) : Bias vector of layer l-1; f: Activation function; L: Total number of layers.

[0047] This also includes model optimization, which uses adaptive L2 regularization to constrain the output weight matrix of the deep neural network, using the following formula:

[0048] L total =L loss +γ·R(W)

[0049] In the formula, L total : Total objective function value; L loss : Prediction error; R(W): Regularization term, used to constrain the size of the weight matrix W; γ: Regularization strength coefficient.

[0050] In step S4, the dynamic verification includes:

[0051] The prediction results are compared with real-time monitoring data, using the following relative error:

[0052]

[0053] In the formula, E rel : Relative error; Y pred Model predicted value; Y real Real-time monitoring data;

[0054] Error verification includes:

[0055] The error is relatively small: E rel If the result is less than 5%, the prediction is considered acceptable, and the model directly outputs the simulation result.

[0056] Medium error: 5% ≤ E rel <10%, indicating that the model needs to be verified;

[0057] The error is relatively large: E rel ≥10% triggers model parameter adjustment or data input optimization.

[0058] When the error is large, the model parameters are dynamically adjusted through incremental learning;

[0059] This includes: adding new data samples and updating model weights;

[0060] Weight update formula: For the weight matrix W of each layer (l) The updated formula is:

[0061]

[0062] In the formula, V: loss function; μ: learning rate; ΔW (l) β: Incremental adjustment of weights for new samples; Incremental learning strength;

[0063] Or reallocate data source weights;

[0064] The redistribution formula is: data source weights are based on historical relevance (rel). iVolatility var i and sensitivity i Recalculate:

[0065]

[0066] This invention encompasses technical modules such as multi-source data fusion, deep learning model construction, dynamic verification, and adaptive optimization. By fusing multi-source information related to real-time load data, equipment operation data, and environmental data, and combining dynamic adjustment and incremental learning mechanisms, it achieves high-precision prediction and optimization of the distribution network's operating status. Attached Figure Description

[0067] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In the drawings, the parts are not necessarily drawn to scale.

[0068] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0069] The present invention is as follows Figure 1 As shown, a data-driven simulation method for the operation characteristics of a distribution network includes the following steps:

[0070] S1. Acquisition and preprocessing of multi-source data;

[0071] S2. Adaptively fuse the preprocessed data to form a dataset;

[0072] This invention generates a unified, high-quality input dataset through multi-source data preprocessing and fusion, providing support for model building and verification.

[0073] S3. Construct an initial simulation model, train it based on the dataset to obtain the main simulation model; output the prediction results of distribution network characteristics through the main simulation model; construct a deep learning model based on the fused data, and output the key simulation results of distribution network operation characteristics.

[0074] S4. Based on the prediction results, perform dynamic verification and simulation optimization to output the simulation results. Through verification and simulation optimization, the real-time performance and adaptability of the model are enhanced, ensuring the accuracy and practicality of the simulation results.

[0075] This invention encompasses technical modules such as multi-source data fusion, deep learning model construction, dynamic verification, and adaptive optimization. By fusing multi-source information related to real-time load data, equipment operation data, and environmental data, and combining dynamic adjustment and incremental learning mechanisms, it achieves high-precision prediction and optimization of the distribution network's operating status.

[0076] In step S1,

[0077] Multi-source data includes real-time load data, equipment operation data, and environmental data;

[0078] The collected multi-source data is cleaned and standardized, including missing value imputation, data denoising, and time alignment.

[0079] Missing value imputation: Negative missing values ​​are filled in using time series interpolation algorithms;

[0080] Data denoising: Noise is removed using a sliding window filtering algorithm;

[0081] Time alignment: unify timestamps of data from different sources to ensure data synchronization.

[0082] Real-time load data includes the power change trend of each node;

[0083] Equipment operating data includes transformer status, voltage, and current;

[0084] Environmental data includes meteorological information and market fluctuations. Meteorological information includes temperature and humidity; market fluctuations include electricity price fluctuations, electricity demand fluctuations, and electricity supply fluctuations.

[0085] In step S2, the weights of each data dimension are dynamically adjusted according to the importance of the data, and a unified distribution network operation dataset is generated by merging them.

[0086] Among them, the importance of data is a measure of the degree of influence of a certain data dimension on the operating status of the distribution network, including historical correlation, real-time fluctuation, and operating status sensitivity.

[0087] Historical Relevance i The contribution of a specific data dimension to key indicators of distribution network operation (such as voltage stability and load distribution) in historical data is calculated through correlation analysis (such as the Pearson correlation coefficient). The formula is as follows:

[0088]

[0089] In the formula, X i : Value of the i-th data dimension; Y: Key indicator of operational status; Cov(X) i (Y): X i The covariance of Y; σ Xi and σ Y :X i and the standard deviation of Y;

[0090] Real-time volatility var iThis represents the degree of drastic change in a certain data dimension at the current moment. The magnitude of data change is calculated using a sliding window, as shown in the following formula:

[0091]

[0092] In the formula, X i,j : The value of the i-th data dimension in the j-th time window; The average value of the i-th data dimension within the current window; N: sliding window size;

[0093] Operating state sensitivity i This reflects the direct impact of fluctuations in a certain data dimension on key operational status indicators, and is defined as the partial derivative of the operational status indicator with respect to the data dimension; it is expressed by the following formula:

[0094]

[0095] In the formula, Derivative function.

[0096] When the load fluctuates drastically at a certain moment (such as during peak periods), var i To improve, the weight w1 of real-time load data increases;

[0097] When a device becomes abnormal, it leads to operational status sensitivity. i As a result, the weight w2 of the equipment operation data increases;

[0098] The current temperature has risen sharply, causing var i The weight w1 of real-time load data is increased.

[0099] The weights of each data dimension are dynamically adjusted, including: when the weight of a certain data source changes due to changes in its characteristics (such as increased volatility or increased sensitivity), the weights of other data sources need to be redistributed proportionally to keep the total weight of 1.

[0100] Let w1, w2, and w3 (corresponding to real-time load data, equipment operation data, and environmental data, respectively). When w1 (the weight of real-time load data) increases due to increased volatility, the other weights w2 and w3 are adjusted according to the following formula:

[0101] w1′=w1·(1+Δ1)

[0102]

[0103] in:

[0104] w1′: Adjusted load weight;

[0105] w2′ and w3′: Reassigned device operating weights and environment weights;

[0106] Δ1: The percentage increase in load weight due to fluctuations.

[0107] The data dimensions are merged according to their weights to generate a unified power distribution network operation dataset, as follows:

[0108]

[0109] In the formula, D fused : The merged, integrated dataset; X i : The value of the i-th data dimension; w i : Weight of the i-th data dimension; n: Number of data points.

[0110] This invention generates a highly consistent and comprehensive power distribution network operation dataset through multi-source data fusion, thereby improving the accuracy and timeliness of data input and providing a reliable foundation for subsequent simulation models.

[0111] In step S3,

[0112] An initial simulation model based on a deep neural network is constructed, and the main simulation model is obtained after training. The formula of the main simulation model is:

[0113] F = W (L) ·f(W (L-1) ·f(···f(W (L) ·X+b (L) )···)+b (L-1) )b (L)

[0114] In the formula, F: main simulation model function; X: input feature matrix; W (l) W: The weight matrix of the l-th layer; (L-1) b: The weight matrix of the (l-1)th layer; (l) : The bias vector of the l-th layer; b (L-1) : Bias vector of layer l-1; f: Activation function; L: Total number of layers;

[0115] The main simulation model is used to capture the nonlinear relationships of the distribution network operation characteristics:

[0116] Input features: Feature matrix from fused data, containing multi-dimensional features of distribution network operation (such as load, voltage, current, equipment status, environmental conditions, etc.);

[0117] Output targets: node voltage, current, power flow, and other characteristics.

[0118] This also includes model optimization, employing adaptive L2 regularization to constrain the output weight matrix of the deep neural network to improve the model's generalization ability and adapt to different scenarios and operating conditions. The following formula is used:

[0119] L total =L loss +γ·R(W)

[0120] In the formula, L total : Total objective function value; L loss : Prediction error; R(W): Regularization term, used to constrain the size of the weight matrix W; γ: Regularization strength coefficient.

[0121] The main simulation model is the core of this invention, used to accurately capture the complex nonlinear relationships of distribution network operation characteristics. It is a key component that, after training with datasets, can ultimately process multi-dimensional data inputs and generate high-precision simulation results of the distribution network's operating status.

[0122] The initial simulation model relies on fixed parameters and is suitable for prediction in small-scale, simple scenarios; it is difficult to cope with the modeling needs of dynamic, multi-source, and nonlinear relationships.

[0123] Main simulation model: A dynamically optimized deep learning model; capable of handling multi-dimensional complex inputs and performing real-time prediction and optimization; supports dynamic verification and adaptive adjustment, adapting to various scenarios.

[0124] In step S4, the dynamic verification includes:

[0125] The prediction results are compared with real-time monitoring data, using the following relative error:

[0126]

[0127] In the formula, E rel : Relative error; Y pred Model predicted value; Y real Real-time monitoring data;

[0128] Error is the deviation between the model's predicted value and the real-time monitoring data;

[0129] Error verification includes:

[0130] The error is relatively small: E rel If the result is less than 5%, the prediction is considered acceptable, and the model directly outputs the simulation result.

[0131] Medium error: 5% ≤ E rel <10%, indicating that the model needs to be verified;

[0132] The error is relatively large: E rel ≥10% triggers model parameter adjustment or data input optimization.

[0133] Under conditions of moderate error, the model is calibrated, including:

[0134] 1) Recalculate the weights using the weighting formula:

[0135] If the error primarily originates from the load data, increase w1;

[0136] If the device status is highly sensitive, adjust w2.

[0137] 2) Regularization parameter verification

[0138] Check if the model is overfitting or underfitting:

[0139] Overfitting: The weight parameters are too large; adjust the regularization strength γ.

[0140] Underfitting: The model is not complex enough. Increase the number of hidden layers or neurons appropriately.

[0141] By dynamically adjusting γ, the problem of excessively large or small weights can be suppressed.

[0142] When the error is large, the model parameters are dynamically adjusted through incremental learning;

[0143] This includes: adding new data samples and updating model weights; or reallocating data source weights;

[0144] Specifically as follows:

[0145] 1) Add data samples with high volatility or abnormal scenarios and update the model weights;

[0146] (1) Incremental data sample update:

[0147] Add new samples from highly volatile or abnormal scenarios to the training set;

[0148] High volatility refers to a sudden change in a data source within a short period of time, which is usually quantified by calculating the rate of change using a sliding window.

[0149] Abnormal scenarios refer to situations in the operation of a power distribution network that deviate significantly from historical trends or normal conditions.

[0150] The training weights are increased based on the importance of the samples.

[0151] (2) Weight update formula: For the weight matrix W of each layer (l) The updated formula is:

[0152]

[0153] In the formula, V: loss function; μ: learning rate; ΔW (l) β: Incremental adjustment of weights for new samples; β: Incremental learning intensity.

[0154] By adjusting model weights based on real-time monitoring data and updating model parameters through an incremental learning mechanism, the adaptability of the model under high volatility and abnormal scenarios has been significantly improved.

[0155] Through real-time verification and dynamic optimization, optimization solutions can be quickly generated in load fluctuation or abnormal scenarios, significantly shortening response time and reducing power outages and energy losses.

[0156] 2) Reallocate data source weights to increase the proportion of key data.

[0157] Key data refers to data dimensions that have a significant impact on the operation of the distribution network, and typically includes:

[0158] Load data during peak load periods;

[0159] Abnormal equipment operation data;

[0160] Meteorological data under extreme weather conditions.

[0161] Reassignment formula: Data source weights are based on historical relevance (rel). i Volatility var i and sensitivity i Recalculate:

[0162]

[0163] Finally, based on the verified model, simulation results are output (such as node voltage change curves, line power distribution diagrams, power flow optimization suggestions, etc.).

[0164] This invention improves the real-time performance and accuracy of simulation results through dynamic verification; adaptive adjustment enhances the stability of the model in complex environments.

[0165] The present invention has the following advantages:

[0166] By fusing multi-source data and adaptive weight adjustment, the impact of load fluctuations, equipment malfunctions and environmental changes can be dynamically captured.

[0167] Based on nonlinear modeling capabilities, it captures complex relationships between multidimensional features, thereby improving prediction accuracy.

[0168] The combination of dynamic verification mechanism and incremental learning enhances the model's adaptability to highly volatile and abnormal scenarios.

[0169] Dynamically adjust model weights and parameters to generate optimization results in real time. Integrate simulation and decision support for rapid response to load fluctuations, equipment malfunctions, and environmental changes.

[0170] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data-driven simulation method for the operational characteristics of a distribution network, characterized in that, Includes the following steps: S1. Acquisition and preprocessing of multi-source data; multi-source data includes real-time load data, equipment operation data, and environmental data. S2. Adaptively fuse the preprocessed data to form a dataset; S3. Construct an initial simulation model with a deep neural network as its core, train it based on the dataset, and obtain the main simulation model; The main simulation model outputs the predicted results of the distribution network characteristics; S4. Based on the prediction results, perform dynamic verification and simulation optimization to output the simulation results; In step S2, the weights of each data dimension are dynamically adjusted according to the importance of the data, and a unified distribution network operation dataset is generated by merging them. The importance of the data includes historical correlation, real-time volatility, and operational status sensitivity. Historical relevance The following formula is used: In the formula, : No. The value of the data dimension; Key indicators of operational status; : and covariance; and : and Standard deviation; Real-time volatility The following formula is used: In the formula, : No. Data dimension in the first The value of a time window; : No. The average value of the data dimension within the current window; : Slide window size; Operating state sensitivity The following formula is used: In the formula, : Derivative function; Dynamically adjust the weights of each data dimension, including: when the weight of a certain data source changes, the weights of other data sources need to be redistributed proportionally to keep the total weight of 1. set up , , The weights corresponding to real-time load data, equipment operation data, and environmental data are respectively, when During adjustment, and Adjust according to the following formula: ， ， In the formula, Adjusted load weight; and : The reassigned equipment operating weights and environmental weights; The incremental proportion of load weight changes due to volatility; The data dimensions are merged according to their weights to generate a unified power distribution network operation dataset, as follows: In the formula, The merged, integrated dataset; : No. The value of the data dimension; : No. Weights of data dimensions; n: number of data points; In step S4, the dynamic verification includes: The prediction results are compared with real-time monitoring data, using the following relative error: In the formula, Relative error; Model predictions; Real-time monitoring data; Error verification includes: The error is small: If the result is less than 5%, the prediction is considered acceptable, and the model directly outputs the simulation result. Medium error: 5% ≤ <10%, indicating that the model needs to be verified; The error is relatively large: ≥10% triggers model parameter adjustment or data input optimization; When the error is large, the model parameters are dynamically adjusted through incremental learning; This includes: adding new data samples and updating model weights; Weight update formula: For the weight matrix of each layer The updated formula is: In the formula, V: loss function; Learning rate; Incremental adjustment of weights for new samples; Incremental learning intensity; Or reallocate data source weights; The redistribution formula is as follows: Data source weights are based on historical relevance. volatility and sensitivity Recalculate: 。 2. The data-driven simulation method for distribution network operation characteristics according to claim 1, characterized in that, In step S1, The collected multi-source data is cleaned and standardized, including missing value imputation, data denoising, and time alignment.

3. The data-driven simulation method for distribution network operation characteristics according to claim 2, characterized in that, Real-time load data includes the power change trend of each node; Equipment operating data includes transformer status, voltage, and current; Environmental data includes meteorological information and market fluctuations. Meteorological information includes temperature and humidity; market fluctuations include electricity price fluctuations, electricity demand fluctuations, and electricity supply fluctuations.

4. The data-driven simulation method for distribution network operation characteristics according to claim 1, characterized in that, In step S3, The master simulation model is obtained after training. The formula for the master simulation model is: In the formula, F: main simulation model function; Input feature matrix; : No. Layer weight matrix; : No. Layer bias vector; Activation function; L: Total number of floors.

5. The data-driven simulation method for distribution network operation characteristics according to claim 4, characterized in that, It also includes model optimization, using adaptive L2 regularization to constrain the output weight matrix of the deep neural network, using the following formula: In the formula, : Overall objective function value; Prediction error; Regularization term, used to constrain the weight matrix. Size; Regularization intensity coefficient.

Images