CN113972667A - Power distribution network reactive power optimization method based on one-dimensional convolutional neural network - Google Patents

Power distribution network reactive power optimization method based on one-dimensional convolutional neural network Download PDF

Info

Publication number
CN113972667A
CN113972667A CN202111434416.0A CN202111434416A CN113972667A CN 113972667 A CN113972667 A CN 113972667A CN 202111434416 A CN202111434416 A CN 202111434416A CN 113972667 A CN113972667 A CN 113972667A
Authority
CN
China
Prior art keywords
neural network
distribution network
power distribution
reactive power
node
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111434416.0A
Other languages
Chinese (zh)
Inventor
王兴鑫
刘志坚
和鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kunming University of Science and Technology
Original Assignee
Kunming University of Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kunming University of Science and Technology filed Critical Kunming University of Science and Technology
Priority to CN202111434416.0A priority Critical patent/CN113972667A/en
Publication of CN113972667A publication Critical patent/CN113972667A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/18—Arrangements for adjusting, eliminating or compensating reactive power in networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/04—Power grid distribution networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/30—Reactive power compensation
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Abstract

The invention discloses a power distribution network reactive power optimization method based on a one-dimensional convolutional neural network, which comprises the following steps of: step 1, establishing a power distribution network reactive power optimization model with minimum active network loss and voltage deviation as a target function, and calculating by using a particle swarm optimization algorithm to obtain a reactive power optimization strategy corresponding to historical load data of the power distribution network; step 2, performing normalization processing on historical power distribution network load data to serve as data characteristics input by a training neural network, and using the binary coding of the reactive power optimization strategy obtained in the step 1 as an output label of the neural network; step 3, training a one-dimensional convolutional neural network to determine the structural parameters of the neural network, and obtaining a trained one-dimensional convolutional neural network model; and 4, extracting data characteristics from the actual power distribution network load data at the moment to be optimized, inputting the data characteristics into the trained one-dimensional convolutional neural network model, and outputting the binary codes corresponding to the reactive power optimization strategy. The invention can quickly obtain the corresponding reactive power optimization strategy and accelerate the operation time.

Description

Power distribution network reactive power optimization method based on one-dimensional convolutional neural network
Technical Field
The invention relates to a power distribution network reactive power optimization method based on a one-dimensional convolutional neural network, and belongs to the field of power distribution network reactive power optimization.
Background
Along with the continuous expansion of the scale of the distribution network, the uncertainty and the fluctuation of the load of the distribution network make the voltage stable operation of the distribution network face a serious challenge, and the improvement of the electric energy quality by ensuring the reactive power balance of the system through a certain reactive power regulation means is particularly important. The reactive power optimization of the power distribution network is a multi-objective nonlinear programming problem, the existing solving method depends on a model and parameters of the power distribution network, repeated iterative load flow calculation is needed, the calculation time is long, and the decision-making performance is poor.
Aiming at the technical problems that the existing related solving method depends on a model and parameters of a power distribution network, so that repeated iteration load flow calculation is needed, the calculation time is long, and the decision-making property is poor, an effective solution is not provided at present.
Disclosure of Invention
The invention provides a power distribution network reactive power optimization method based on a one-dimensional convolutional neural network.
The technical scheme of the invention is as follows: a power distribution network reactive power optimization method based on a one-dimensional convolutional neural network comprises the following steps:
step 1, establishing a power distribution network reactive power optimization model with minimum active network loss and voltage deviation as a target function, and calculating by using a particle swarm optimization algorithm to obtain a reactive power optimization strategy corresponding to historical load data of the power distribution network;
step 2, performing normalization processing on historical power distribution network load data to serve as data characteristics input by a training neural network, and using the binary coding of the reactive power optimization strategy obtained in the step 1 as an output label of the neural network;
step 3, training a one-dimensional convolutional neural network to determine the structural parameters of the neural network, and obtaining a trained one-dimensional convolutional neural network model;
and 4, extracting data characteristics from the actual power distribution network load data at the moment to be optimized, inputting the data characteristics into the trained one-dimensional convolutional neural network model, and outputting the binary codes corresponding to the reactive power optimization strategy.
The reactive power optimization model of the power distribution network established in the step 1 is expressed as follows:
active network loss:
Figure BDA0003381316240000011
in the formula: n islThe total number of network branches; gk(i,j)Conductance of branch k from node i to node j; u. ofi、ujVoltages of nodes i and j, respectively; thetai、θjPhase angles of voltages at nodes i and j, respectively; the node i and the node j are adjacent nodes;
node voltage deviation:
Figure BDA0003381316240000021
in the formula: u. ofmIs the voltage at load node m; u. ofStatorA specified voltage amplitude for the load node; Δ ummaxThe maximum voltage deviation allowed for the load node m; n is a radical oflThe total number of the load nodes is;
to sum up, the reactive power optimization model of the power distribution network is as follows:
minf＝floss+λfU(ii) a In the formula, λ is a penalty coefficient of the node voltage deviation.
The step 2 specifically comprises the following steps:
step 2.1: carrying out normalization processing on the node load data of the power distribution network to enable the node load data to be in a [0,1] interval, wherein the formula is as follows:
x'＝(x-xmin)/(xmax-xmin)
in the formula: x and x' are respectively the node load data of the distribution network before and after standardization, xmax、xminRespectively corresponding maximum and minimum values of the node load data;
step 2.2: and solving the reactive power optimization strategy binary code of the load data of the power distribution network node by using a particle swarm optimization algorithm as an output label of the sample.
The structure of the one-dimensional convolutional neural network comprises an input layer, a one-dimensional convolutional layer, a pooling layer, a full-link layer and an output layer, wherein the one-dimensional convolutional layer, the pooling layer and the full-link layer can be one or more.
The one-dimensional convolutional layer adopts three layers, the number of convolutional kernels is 64, 128 and 256, and the length of the convolutional kernel is 11; and adds the Dropout function and the regularization layer.
The invention has the beneficial effects that: according to the method, the reactive power optimization model of the power distribution network is established, the reactive power optimization strategy corresponding to the historical load of the power distribution network is obtained through the particle swarm optimization, the reactive power optimization strategy is subjected to binary coding, then the one-dimensional convolution neural network model is trained to map the nonlinear relation between the node load of the power distribution network and the reactive power optimization strategy, the load data of the power distribution network at the moment to be optimized are input into the neural network model, the corresponding reactive power optimization strategy can be obtained quickly, and the operation time is shortened.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a diagram of a modified IEEE33 node system;
FIG. 3 is a graph of the variation of the loss function;
FIG. 4 is a graph comparing the loss of a network according to different methods;
FIG. 5 is a graph of node voltage distributions under different approaches;
FIG. 6 is a histogram of node voltages under different methods.
Detailed Description
The invention will be further described with reference to the following figures and examples, but the scope of the invention is not limited thereto.
Example 1: as shown in fig. 1 to 6, a method for optimizing reactive power of a power distribution network based on a one-dimensional convolutional neural network includes:
step 1, establishing a power distribution network reactive power optimization model with minimum active network loss and voltage deviation as a target function, and calculating by using a particle swarm optimization algorithm to obtain a reactive power optimization strategy corresponding to historical load data of the power distribution network;
step 2, performing normalization processing on historical power distribution network load data to serve as data characteristics input by a training neural network, and using the binary coding of the reactive power optimization strategy obtained in the step 1 as an output label of the neural network;
step 3, training a one-dimensional convolutional neural network to determine structural parameters of the neural network to obtain a trained one-dimensional convolutional neural network model, and establishing a nonlinear relation from the node load of the power distribution network to a reactive power optimization strategy through the one-dimensional convolutional neural network model;
and 4, extracting data characteristics from actual load data of the power distribution network at the moment to be optimized, inputting the data characteristics into a trained one-dimensional convolutional neural network model, outputting binary codes corresponding to a reactive power optimization strategy, evaluating the network loss level and the voltage level of the power distribution network after decoding, and comparing the network loss level and the voltage level with the traditional nine-region diagram reactive power optimization.
Further, the reactive power optimization model of the power distribution network established in the step 1 can be set as follows:
step 1.1, establishing a target function with minimum active network loss and minimum voltage deviation as the following steps:
active network loss:
Figure BDA0003381316240000031
in the formula: n islThe total number of network branches; gk(i,j)Conductance of branch k from node i to node j; u. ofi、ujVoltages of nodes i and j, respectively; thetai、θjPhase angles of voltages at nodes i and j, respectively; the node i and the node j are adjacent nodesPoint;
node voltage deviation:
Figure BDA0003381316240000032
in the formula: u. ofmIs the voltage at load node m; u. ofStatorFor the specified voltage amplitude of the load node, all the load nodes adopt the same value, and the value is generally 1; Δ ummaxThe maximum voltage deviation allowed for the load node m; n is a radical oflThe total number of the load nodes is;
to sum up, the reactive power optimization model of the power distribution network is as follows:
minf＝floss+λfU(ii) a In the formula, lambda is a penalty coefficient of node voltage deviation;
step 1.2, a power flow constraint step,
Figure BDA0003381316240000041
in the formula: pi、Qi、UiRespectively injecting active power, reactive power and voltage at a node; thetaijIs the difference between the phase angles of the voltages at node i and node j; gij、BijReal and imaginary parts of elements in the node admittance matrix; h denotes the set of all nodes directly connected to the node.
Step 1.3, a safe operation restriction step,
Figure BDA0003381316240000042
in the formula:
Figure BDA0003381316240000043
and
Figure BDA0003381316240000044
respectively, the maximum and minimum voltage values allowed for node i.
Step 1.4, controlling variable constraint step:
Figure BDA0003381316240000045
in the formula:
Figure BDA0003381316240000046
the maximum reactive power which can be sent out by the static reactive power compensator; kTmax、KTminRepresenting the maximum and minimum transformation ratios of the transformer; qCmaxRepresenting the maximum reactive power that the capacitor bank can deliver.
Further, the step 2 may specifically be:
step 2.1: carrying out normalization processing on the node load data of the power distribution network to enable the node load data to be in a [0,1] interval, wherein the formula is as follows:
x′＝(x-xmin)/(xmax-xmin)
in the formula: x and x' are respectively the node load data of the distribution network before and after standardization, xmax、xminRespectively corresponding maximum and minimum values of the node load data;
step 2.2: and solving the reactive power optimization strategy binary code of the load data of the power distribution network node by using a particle swarm optimization algorithm as an output label of the sample.
Further, in the step 3, the specific method is as follows:
the structure of the one-dimensional convolutional neural network is composed of an input layer, a one-dimensional convolutional layer, a pooling layer, a full-link layer and an output layer, wherein the number of the one-dimensional convolutional layer, the pooling layer and the full-link layer can be multiple. In the one-dimensional convolutional neural network, input power distribution network node load data is firstly transformed through a series of convolutional kernels with nonlinear activation functions, and then the neural network can learn complex data characteristics by stacking a plurality of convolutional layers and outputting the convolutional layers by using the nonlinear activation functions.
The one-dimensional convolutional layer is a core part of the one-dimensional convolutional neural network, and the mathematical expression of the convolution process is as follows:
Figure BDA0003381316240000051
in the formula: l is the current layer network; j is the previous layer network;
Figure BDA0003381316240000052
mapping the jth feature in the ith layer of the current layer;
Figure BDA0003381316240000053
mapping the ith characteristic of the previous layer;
Figure BDA0003381316240000054
convolution kernels for i-th and j-th layers;
Figure BDA0003381316240000055
for biasing, f (-) is the activation function.
The pooling layer is also called a sampling layer and is generally linked behind the convolution layer, and through sampling the convolution layer, the characteristic dimension after convolution is reduced, the network complexity and the calculation workload are reduced, and the occurrence of overfitting is avoided. The pooling method frequently used is maximum pooling, and the formula is:
Figure BDA0003381316240000056
in the formula: a isl(i，t)Represents the t-th neuron of the ith map in the l-th layer.
And obtaining a plurality of one-dimensional feature layers after pooling, flattening the plurality of one-dimensional feature layers by using a Flatten layer, inputting the one-dimensional feature data into a full connection layer, and connecting all the input features by the full connection layer to play a role in classification in the whole neural network. Because the output reactive strategy uses binary coding, in the binary classification problem, the output layer adopts sigmoid as an activation function, and the mathematical expression is as follows:
Figure BDA0003381316240000057
in the formula:
Figure BDA0003381316240000058
is an output vector, representing the probability (predicted value) that the sample label is 1, h(x)Is the input vector of the fully connected layer.
One of the biggest problems with deep neural networks is that overfitting tends to occur. In order to suppress the overfitting phenomenon, the invention adopts an L2-norm regularization method and a Dropout function to increase the generalization capability of the model.
The specific steps of training the one-dimensional convolutional neural network model are as follows:
step 3.1: the method comprises the steps of initializing the structure and parameters of the one-dimensional convolutional neural network, wherein load data characteristics of nodes of the power distribution network are far smaller than picture data, and the fitting accuracy of a model can be reduced by using a pooling layer.
Step 3.2: the load data characteristics of the nodes of the power distribution network are extracted by the one-dimensional convolutional layer, a reactive power optimization strategy scheme under binary coding is output by the full connection layer, binary cross entropy is selected as a loss function, the weight and the deviation of the neural network are updated through a back propagation algorithm, the deviation of a predicted value and an actual value is continuously reduced until the loss function tends to be stable, iteration is stopped, and a determined one-dimensional convolutional neural network model is obtained.
The specific embodiment of the invention is as follows:
the simulation verification is carried out by using an improved IEEE33 node system, the specific structure of the power distribution network is shown in figure 2, the rated voltage of the power distribution network is 10kV, a node 0 is used as a balance node, the other 32 nodes are load nodes, the total number of on-load voltage regulators is 10, the adjustment step length is 0.02, and the range is 0.9-1.1. The capacitor bank C1 is arranged at the node 15, the capacitor bank C2 is arranged at the node 29, and the static var compensator is connected to the node 8. The adjusting strategy of the reactive power optimization equipment of the power distribution network uses binary codes, the adjusting capacity of C1 is 100kvar, the adjusting capacity of C2 is 100kvar, the capacity of each group is 25kvar, and the gear number of the adjusting strategy is represented by 3-bit binary numbers. The tap position of the transformer is 10 steps and can be represented by 4-bit binary numbers. The capacity of the static reactive compensator is 0-100 kvar and corresponds to 7-bit binary numbers, namely, 17-bit binary codes are used for representing a reactive power optimization strategy.
The historical load data of the power distribution network is derived from a London smart meter data set, the electricity consumption of each household of 112 blocks in 2011-2014-2 months is counted, and the collection time interval is 1 hour. The load of 112 blocks is equivalently simplified by combining an IEEE33 node system for simulation, the load of every two to three adjacent blocks is properly combined to be used as the load of one node in the power distribution network, Gaussian noise with the standard deviation of 0.1 is added to the load data of each node, load data at 5000 moments are obtained after data cleaning, a reactive power optimization model is solved by using a particle swarm algorithm to obtain a reactive power optimization strategy corresponding to each moment, and the reactive power optimization strategy is used as a label of a data set. And (3) forming 5000 sample sets, wherein 80% of samples are training sets, and 20% of samples are testing sets.
And taking the batch size as 128 according to the data size of the training, selecting a random gradient descent method to optimize the model, wherein the training is performed for 250 rounds in total, the change process of the loss function is shown in FIG. 3, and the loss function tends to be smooth and indicates that the training of the neural network model is completed.
Determining the optimal neural network structure parameters, changing the number of convolution kernels in the network, the length of the convolution kernels, and whether a Dropout layer and a regularization coefficient are added or not, training the Dropout layer and the regularization coefficient to obtain the fitting effect of the test set on different network structures, and as shown in Table 1, determining the optimal network structure parameters as convolution layer parameters, selecting (64-128-256,11), and adding a Dropout function and a regularization layer.
TABLE 1 fitting Effect of different network architectures
Figure BDA0003381316240000061
The method is used for simulating the data samples of 48 moments in two consecutive days in the test set (namely, the method in the figure) and comparing the data samples with the traditional nine-region diagram reactive power optimization method, the figure 4 shows the reactive power network loss of the power distribution network under different methods, and compared with the method without reactive power optimization, the method and the nine-region diagram can obviously reduce the network loss, but under the reactive power regulation of the method and the nine-region diagram, the network loss of the sample set is smaller.
Taking the moment with the maximum load in the sample as a typical moment, the comparison situation of the voltages of each node of the power distribution network under different reactive power control strategies is shown in fig. 5-6, fig. 5 shows the voltage magnitude of each node under different methods, and fig. 6 shows the distribution situation of the node voltages under different strategies. It can be seen from fig. 5-6 that after the reactive power regulation strategy is added, the voltage of each node of the system is improved, the voltage fluctuation range is reduced, and the voltage distribution of each node of the power distribution network is more concentrated after the method of the invention is adopted, so that the running economy of the power distribution network system is improved.
The present invention is not described in detail in the prior art.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims (5)

1. A power distribution network reactive power optimization method based on a one-dimensional convolutional neural network is characterized by comprising the following steps: the method comprises the following steps:
step 1, establishing a power distribution network reactive power optimization model with minimum active network loss and voltage deviation as a target function, and calculating by using a particle swarm optimization algorithm to obtain a reactive power optimization strategy corresponding to historical load data of the power distribution network;
step 2, performing normalization processing on historical power distribution network load data to serve as data characteristics input by a training neural network, and using the binary coding of the reactive power optimization strategy obtained in the step 1 as an output label of the neural network;
step 3, training a one-dimensional convolutional neural network to determine the structural parameters of the neural network, and obtaining a trained one-dimensional convolutional neural network model;
and 4, extracting data characteristics from the actual power distribution network load data at the moment to be optimized, inputting the data characteristics into the trained one-dimensional convolutional neural network model, and outputting the binary codes corresponding to the reactive power optimization strategy.
2. The power distribution network reactive power optimization method based on the one-dimensional convolutional neural network as claimed in claim 1, wherein: the reactive power optimization model of the power distribution network established in the step 1 is expressed as follows:
active network loss:
Figure FDA0003381316230000011
in the formula: n islThe total number of network branches; gk(i,j)Conductance of branch k from node i to node j; u. ofi、ujVoltages of nodes i and j, respectively; thetai、θjPhase angles of voltages at nodes i and j, respectively; the node i and the node j are adjacent nodes;
node voltage deviation:
Figure FDA0003381316230000012
in the formula: u. ofmIs the voltage at load node m; u. ofStatorA specified voltage amplitude for the load node; Δ ummaxThe maximum voltage deviation allowed for the load node m; n is a radical oflThe total number of the load nodes is;
to sum up, the reactive power optimization model of the power distribution network is as follows:
minf＝floss+λfU(ii) a In the formula, λ is a penalty coefficient of the node voltage deviation.
3. The power distribution network reactive power optimization method based on the one-dimensional convolutional neural network as claimed in claim 1, wherein: the step 2 specifically comprises the following steps:
step 2.1: carrying out normalization processing on the node load data of the power distribution network to enable the node load data to be in a [0,1] interval, wherein the formula is as follows:
x'＝(x-xmin)/(xmax-xmin)
in the formula: x and x' are respectively the node load data of the distribution network before and after standardization, xmax、xminRespectively corresponding maximum and minimum values of the node load data;
step 2.2: and solving the reactive power optimization strategy binary code of the load data of the power distribution network node by using a particle swarm optimization algorithm as an output label of the sample.
4. The power distribution network reactive power optimization method based on the one-dimensional convolutional neural network as claimed in claim 1, wherein: the structure of the one-dimensional convolutional neural network comprises an input layer, a one-dimensional convolutional layer, a pooling layer, a full-link layer and an output layer, wherein the one-dimensional convolutional layer, the pooling layer and the full-link layer can be one or more.
5. The power distribution network reactive power optimization method based on the one-dimensional convolutional neural network as claimed in claim 4, wherein: the one-dimensional convolutional layer adopts three layers, the number of convolutional kernels is 64, 128 and 256, and the length of the convolutional kernel is 11; and adds the Dropout function and the regularization layer.
CN202111434416.0A 2021-11-29 2021-11-29 Power distribution network reactive power optimization method based on one-dimensional convolutional neural network Pending CN113972667A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111434416.0A CN113972667A (en) 2021-11-29 2021-11-29 Power distribution network reactive power optimization method based on one-dimensional convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111434416.0A CN113972667A (en) 2021-11-29 2021-11-29 Power distribution network reactive power optimization method based on one-dimensional convolutional neural network

Publications (1)

Publication Number Publication Date
CN113972667A true CN113972667A (en) 2022-01-25

Family

ID=79590338

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111434416.0A Pending CN113972667A (en) 2021-11-29 2021-11-29 Power distribution network reactive power optimization method based on one-dimensional convolutional neural network

Country Status (1)

Country Link
CN (1) CN113972667A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115017735A (en) * 2022-07-05 2022-09-06 重庆大学 Multi-agent probability voltage stability calculation method for high-dimensional system
CN115632406A (en) * 2022-12-21 2023-01-20 国网天津市电力公司城东供电分公司 Reactive voltage control method and system based on digital-mechanism fusion drive modeling
EP4254705A1 (en) * 2022-03-30 2023-10-04 Fundación Tecnalia Research & Innovation Computer-implemented method or assessment of a power distribution grid
CN117526443A (en) * 2023-11-07 2024-02-06 北京清电科技有限公司 Novel power system-based power distribution network optimization regulation and control method and system
CN117526443B (en) * 2023-11-07 2024-04-26 北京清电科技有限公司 Power system-based power distribution network optimization regulation and control method and system

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103219732A (en) * 2013-04-24 2013-07-24 上海电力学院 reactive voltage controlling method of power distribution network with variable speed constant frequency wind farm
US20190288508A1 (en) * 2018-03-16 2019-09-19 Siemens Aktiengesellschaft Method for operating an electrical energy supply network, and control device for controlling devices of an electrical distribution network
WO2019190036A1 (en) * 2018-03-27 2019-10-03 경상대학교산학협력단 Method and apparatus for transmission power control based on neural network in wireless communication system
CN110535146A (en) * 2019-08-27 2019-12-03 哈尔滨工业大学 The Method for Reactive Power Optimization in Power of Policy-Gradient Reinforcement Learning is determined based on depth
CN111756049A (en) * 2020-06-18 2020-10-09 国网浙江省电力有限公司电力科学研究院 Data-driven reactive power optimization method considering loss of real-time measurement information of power distribution network
US20210112073A1 (en) * 2020-09-25 2021-04-15 Intel Corporation Broadcast remote sealing for scalable trusted execution environment provisioning
CN113078641A (en) * 2021-04-29 2021-07-06 国网山东省电力公司经济技术研究院 Power distribution network reactive power optimization method and device based on evaluator and reinforcement learning

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103219732A (en) * 2013-04-24 2013-07-24 上海电力学院 reactive voltage controlling method of power distribution network with variable speed constant frequency wind farm
US20190288508A1 (en) * 2018-03-16 2019-09-19 Siemens Aktiengesellschaft Method for operating an electrical energy supply network, and control device for controlling devices of an electrical distribution network
WO2019190036A1 (en) * 2018-03-27 2019-10-03 경상대학교산학협력단 Method and apparatus for transmission power control based on neural network in wireless communication system
CN110535146A (en) * 2019-08-27 2019-12-03 哈尔滨工业大学 The Method for Reactive Power Optimization in Power of Policy-Gradient Reinforcement Learning is determined based on depth
CN111756049A (en) * 2020-06-18 2020-10-09 国网浙江省电力有限公司电力科学研究院 Data-driven reactive power optimization method considering loss of real-time measurement information of power distribution network
US20210112073A1 (en) * 2020-09-25 2021-04-15 Intel Corporation Broadcast remote sealing for scalable trusted execution environment provisioning
CN113078641A (en) * 2021-04-29 2021-07-06 国网山东省电力公司经济技术研究院 Power distribution network reactive power optimization method and device based on evaluator and reinforcement learning

Non-Patent Citations (5)

* Cited by examiner, † Cited by third party
Title
吴昊: "电力系统无功优化数学模型的建立综述", 科技风, pages 187 - 188 *
吴秀华等: "人工神经网络在电力系统无功电压优化中的应用", 沈阳农业大学学报, vol. 39, no. 6, pages 713 - 717 *
李竹星 等: "计及双馈风机接入的配电网无功优化", 电器与能效管理技术, pages 86 - 91 *
赵宏: "《人工智能技术丛书 深度学习基础教程》", 机械工业出版社, pages: 69 - 71 *
邵美阳等: "基于数据驱动和深度置信网络的配电网无功优化", 电网技术, vol. 43, no. 6, pages 1874 - 1883 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4254705A1 (en) * 2022-03-30 2023-10-04 Fundación Tecnalia Research & Innovation Computer-implemented method or assessment of a power distribution grid
CN115017735A (en) * 2022-07-05 2022-09-06 重庆大学 Multi-agent probability voltage stability calculation method for high-dimensional system
CN115017735B (en) * 2022-07-05 2023-08-15 重庆大学 Multi-agent probability voltage stability calculation method for high-dimensional system
CN115632406A (en) * 2022-12-21 2023-01-20 国网天津市电力公司城东供电分公司 Reactive voltage control method and system based on digital-mechanism fusion drive modeling
CN117526443A (en) * 2023-11-07 2024-02-06 北京清电科技有限公司 Novel power system-based power distribution network optimization regulation and control method and system
CN117526443B (en) * 2023-11-07 2024-04-26 北京清电科技有限公司 Power system-based power distribution network optimization regulation and control method and system

Similar Documents

Publication Publication Date Title
CN113972667A (en) Power distribution network reactive power optimization method based on one-dimensional convolutional neural network
CN110059878B (en) Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof
Tang et al. Short‐term power load forecasting based on multi‐layer bidirectional recurrent neural network
CN110729765B (en) SOP-considered power distribution network flexibility evaluation index system and optimal scheduling method
CN109768573B (en) Power distribution network reactive power optimization method based on multi-target differential gray wolf algorithm
Li et al. Many-objective distribution network reconfiguration via deep reinforcement learning assisted optimization algorithm
CN104636985A (en) Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network
CN110212551B (en) Micro-grid reactive power automatic control method based on convolutional neural network
CN114217524A (en) Power grid real-time self-adaptive decision-making method based on deep reinforcement learning
CN110837915B (en) Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning
CN113378100B (en) Power distribution network source network load storage collaborative optimization scheduling model and method considering carbon emission
CN112861992B (en) Wind power plant ultra-short term power prediction method based on independent sparse stacking self-encoder
CN112418496A (en) Power distribution station energy storage configuration method based on deep learning
CN111625998A (en) Method for optimizing structure of laminated solar cell
Xu et al. Short-term wind power forecasting based on Elman neural network with particle swarm optimization
CN114725944A (en) Power electronic distribution network source and network load optimization operation control method and system
Xu et al. Short-term electricity consumption forecasting method for residential users based on cluster classification and backpropagation neural network
CN117172097A (en) Power distribution network dispatching operation method based on cloud edge cooperation and multi-agent deep learning
Fan et al. Ultra-short-term bus load forecasting method based on multi-source data and hybrid neural network
CN114069650B (en) Power distribution network closed loop current regulation and control method and device, computer equipment and storage medium
CN114583696A (en) Power distribution network reactive power optimization method and system based on BP neural network and scene matching
Xia et al. SCG and LM improved BP neural network load forecasting and programming network parameter settings and data preprocessing
CN114362151A (en) Trend convergence adjusting method based on deep reinforcement learning and cascade graph neural network
CN113887794A (en) Method and device for reactive power optimization of power distribution network
CN115545256A (en) Small-scale photovoltaic power prediction method based on multi-dimensional data feature learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination