Dynamic power optimization method and device for medium and low voltage power distribution network

Abstract

The invention relates to the technical field of power distribution network operation analysis, and particularly provides a method and a device for optimizing dynamic power of a medium-low voltage power distribution network, and the method comprises the steps: taking the power grid data of the medium-low voltage power distribution network in the current time period as the input of a pre-trained convolutional neural network, obtaining the node load power of the low-voltage power distribution network in the prediction period output by the pre-trained convolutional neural network; substituting the load power of the low-voltage power distribution network in the prediction period into a pre-constructed power optimization model, and solving by adopting a simulated annealing algorithm to obtain an optimization result; and obtaining a power optimization scheme of the low-voltage power distribution network in the prediction period based on the optimization result. According to the technical scheme provided by the invention, complex data characteristics can be quickly processed in an extremely short time, the mapping relation between input and output can be reflected more efficiently, and the method can be applied to operation optimization of the medium and low voltage power distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network operation analysis, and in particular to a method and device for dynamic power optimization of medium and low voltage distribution networks. Background Art

[0002] At present, a new round of global energy revolution and technological revolution is evolving deeply and in the ascendant. Vigorously developing renewable energy has become a major strategic direction for global energy transformation and responding to global warming. Completing the flexible, intelligent and digital transformation of the distribution network, realizing multi-level coordination of the main distribution microgrid, massive resource aggregation and interaction, and plug-and-play for multiple users, and effectively promoting the integration and development of distributed smart grids and large power grids has become the primary goal. It can be seen that the construction of a new power system with new energy as the main body is developing rapidly. High-penetration renewable energy has become an important feature of the future development of medium and low voltage distribution networks. More and more distributed power sources are continuously connected to medium and low voltage distribution networks. Their typical randomness and strong uncertainty characteristics affect the power flow distribution of medium and low voltage distribution networks, leading to serious overvoltage problems, and a series of adverse effects such as reduced operating reliability of medium and low voltage distribution networks, deterioration of power quality, complex fault handling and recovery, and reduced operating economy. Therefore, how to coordinate reactive power compensation devices to improve reactive power distribution and improve power supply quality is of great significance.

[0003] Reactive power optimization technology mainly achieves the goals of reducing network losses and improving power quality by optimizing reactive power flow. The construction method based on mathematical models is a type of method widely used in distribution network operation optimization. This type of method can be summarized as taking one or several goals in operation as the objective function, and the physical characteristics and safety characteristics in operation as requirements and constraints, constructing a mathematical model, and obtaining control decisions by solving the mathematical model. The advantage of this type of method is that the control strategy is obtained by solving the mathematical model, and the optimality of the solution can be guaranteed, but it has disadvantages such as low solution efficiency and difficulty in ensuring the optimality and stability of the solution.

[0004] Compared with the above methods, intelligent algorithms have strong computing power and are used to solve distribution network operation optimization problems due to their simple algorithms and easy programming. However, such intelligent algorithms generally have the problem of obtaining local optimal solutions. Considering the uncertainty and bidirectional power flow brought by distributed power sources, traditional optimization methods are no longer applicable, so the need to seek excellent and efficient solution algorithms for solving operation optimization problems has attracted more and more attention. Summary of the invention

[0005] In order to overcome the above-mentioned defects, the present invention proposes a method and device for dynamic power optimization of medium and low voltage distribution networks.

[0006] In a first aspect, a method for dynamic power optimization of a medium and low voltage distribution network is provided, the method comprising:

[0007] Using the grid data of the low-voltage distribution network in the current period as the input of the pre-trained convolutional neural network, and obtaining the node load power of the low-voltage distribution network in the predicted period output by the pre-trained convolutional neural network;

[0008] Substituting the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solving it using a simulated annealing algorithm to obtain an optimization result;

[0009] Based on the optimization results, a power optimization scheme for the low-voltage distribution network in the forecast period is obtained;

[0010] The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

[0011] Preferably, the power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

[0012] Furthermore, the training process of the pre-trained convolutional neural network includes:

[0013] The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period;

[0014] The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

[0015] Preferably, the pre-built power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints for dynamic power optimization configuration of the medium and low voltage distribution network.

[0016] Furthermore, the first objective function is as follows:

[0017]

[0018] In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

[0019] Furthermore, the second objective function is as follows:

[0020]

[0021] In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

[0022] Furthermore, the constraint conditions include at least one of the following: flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation equipment output constraint, and on-load tap-changing transformer constraint.

[0023] Furthermore, the power flow constraints are as follows:

[0024]

[0025] P i =P DGi -P Di

[0026] Q i =Q DGi -Q Di +Q Ci

[0027] The voltage constraints are as follows:

[0028] U i,max ≤U i ≤U i,min

[0029] The current constraints are as follows:

[0030] I i,max ≤I i ≤I i,min

[0031] The energy storage device power constraints and distributed power generation power constraints are as follows:

[0032]

[0033] The output constraints of the reactive power compensation equipment are as follows:

[0034] Q cimin Ci cimax

[0035] The constraints of the on-load tap-changing transformer are as follows:​​

[0036] T imin <T i <T imax

[0037] In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the output of the reactive power compensation device at node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.

[0038] Preferably, in the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

[0039] Furthermore, in the process of judging whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows:

[0040]

[0041] In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

[0042] In a second aspect, a medium and low voltage distribution network dynamic power optimization device is provided, the medium and low voltage distribution network dynamic power optimization device comprising:

[0043] The first analysis module is used to use the power grid data of the medium and low voltage distribution network in the current period as the input of the pre-trained convolutional neural network to obtain the node load power of the medium and low voltage distribution network in the predicted period output by the pre-trained convolutional neural network;

[0044] The second analysis module is used to substitute the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solve it using a simulated annealing algorithm to obtain an optimization result;

[0045] A third analysis module is used to obtain a power optimization scheme for the low-voltage distribution network in the forecast period based on the optimization result;

[0046] The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

[0047] Preferably, the power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

[0048] Furthermore, the training process of the pre-trained convolutional neural network includes:

[0049] The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period;

[0050] The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

[0051] Preferably, the pre-built power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints for dynamic power optimization configuration of the medium and low voltage distribution network.

[0052] Furthermore, the first objective function is as follows:

[0053]

[0054] In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

[0055] Furthermore, the second objective function is as follows:

[0056]

[0057] In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

[0058] Furthermore, the constraint conditions include at least one of the following: flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation equipment output constraint, and on-load tap-changing transformer constraint.

[0059] Furthermore, the power flow constraints are as follows:

[0060]

[0061] P i =P DGi -P Di

[0062] Q i =Q DGi -Q Di +Q Ci

[0063] The voltage constraints are as follows:

[0064] U i,max ≤U i ≤U i,min

[0065] The current constraints are as follows:

[0066] I i,max ≤I i ≤I i,min

[0067] The energy storage device power constraints and distributed power generation power constraints are as follows:

[0068]

[0069] The output constraints of the reactive power compensation equipment are as follows:

[0070] Q cimin Ci cimax

[0071] The constraints of the on-load tap-changing transformer are as follows:

[0072] T imin <T i <T imax

[0073] In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the output of the reactive power compensation device at node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax ​​are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.

[0074] Preferably, in the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

[0075] Furthermore, in the process of judging whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows:

[0076]

[0077] In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

[0078] In a third aspect, a computer device is provided, comprising: one or more processors;

[0079] The processor is used to store one or more programs;

[0080] When the one or more programs are executed by the one or more processors, the medium and low voltage distribution network dynamic power optimization method is implemented.

[0081] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed, the dynamic power optimization method of the medium and low voltage distribution network is implemented.

[0082] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0083] The present invention provides a method and device for dynamic power optimization of medium and low voltage distribution networks, including: using the grid data of the medium and low voltage distribution network in the current period as the input of a pre-trained convolutional neural network, obtaining the node load power of the medium and low voltage distribution network in the forecast period output by the pre-trained convolutional neural network; substituting the load power of the medium and low voltage distribution network in the forecast period into a pre-built power optimization model and solving it using a simulated annealing algorithm to obtain an optimization result; based on the optimization result, obtaining a power optimization scheme for the medium and low voltage distribution network in the forecast period; wherein the optimization result includes at least one of the following: the node active power of the medium and low voltage distribution network in the forecast period, and the node reactive power of the medium and low voltage distribution network in the forecast period. The technical solution provided by the present invention uses a convolutional neural network to predict the load development trend of the medium and low voltage distribution network, and the prediction result has high reliability. The voltage is adjusted in advance before the voltage exceeds the limit caused by the predictable load change, and the accuracy of the optimization result is improved. Compared with the traditional optimization method, it can effectively shorten the calculation time, improve the optimization calculation efficiency, reduce the possibility of voltage exceeding the limit, and improve the voltage stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a schematic flow chart of the main steps of the method for dynamic power optimization of medium and low voltage distribution networks according to an embodiment of the present invention;

[0085] Figure 2 is a convolutional neural network training flow chart of an embodiment of the present invention;

[0086] Figure 3 It is a flow chart of solving the simulated annealing algorithm of an embodiment of the present invention. DETAILED DESCRIPTION

[0087] The specific implementation modes of the present invention will be further described in detail below in conjunction with the accompanying drawings.

[0088] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] As disclosed in the background technology, at present, a new round of global energy revolution and scientific and technological revolution is deeply evolving and in full swing. Vigorously developing renewable energy has become a major strategic direction for global energy transformation and coping with global warming. Completing the flexible, intelligent and digital transformation of the distribution network, realizing multi-level coordination of the main distribution microgrid, massive resource aggregation and interaction, and plug-and-play for multiple users, and effectively promoting the integration and development of distributed smart grids and large power grids has become the primary goal. It can be seen that the construction of a new power system with new energy as the main body is developing rapidly, and renewable energy with high penetration rate has become an important feature of the development of medium and low voltage distribution networks in the future. More and more distributed power sources are continuously connected to the medium and low voltage distribution network. Its typical randomness and strong uncertainty characteristics affect the power flow distribution of the medium and low voltage distribution network, resulting in serious overvoltage problems, and a series of adverse effects such as reduced operating reliability of the medium and low voltage distribution network, deterioration of power quality, complex fault handling and recovery, and reduced operating economy. Therefore, how to coordinate reactive power compensation devices to improve reactive power distribution and improve power supply quality is of great significance.

[0090] Reactive power optimization technology mainly achieves the goals of reducing network losses and improving power quality by optimizing reactive power flow. The construction method based on mathematical models is a type of method widely used in distribution network operation optimization. This type of method can be summarized as taking one or several goals in operation as the objective function, and the physical characteristics and safety characteristics in operation as requirements and constraints, constructing a mathematical model, and obtaining control decisions by solving the mathematical model. The advantage of this type of method is that the control strategy is obtained by solving the mathematical model, and the optimality of the solution can be guaranteed, but it has disadvantages such as low solution efficiency and difficulty in ensuring the optimality and stability of the solution.

[0091] Compared with the above methods, intelligent algorithms have strong computing power and are used to solve distribution network operation optimization problems due to their simple algorithms and easy programming. However, such intelligent algorithms generally have the problem of obtaining local optimal solutions. Considering the uncertainty and bidirectional power flow brought by distributed power sources, traditional optimization methods are no longer applicable, so the need to seek excellent and efficient solution algorithms for solving operation optimization problems has attracted more and more attention.

[0092] In order to improve the above problems, the present invention provides a method and device for dynamic power optimization of medium and low voltage distribution networks, including: using the grid data of the medium and low voltage distribution network in the current period as the input of a pre-trained convolutional neural network, obtaining the node load power of the medium and low voltage distribution network in the forecast period output by the pre-trained convolutional neural network; substituting the load power of the medium and low voltage distribution network in the forecast period into a pre-built power optimization model and solving it using a simulated annealing algorithm to obtain an optimization result; based on the optimization result, obtaining a power optimization scheme for the medium and low voltage distribution network in the forecast period; wherein the optimization result includes at least one of the following: the node active power of the medium and low voltage distribution network in the forecast period, and the node reactive power of the medium and low voltage distribution network in the forecast period. The technical solution provided by the present invention uses a convolutional neural network to predict the load development trend of the medium and low voltage distribution network, and the prediction result has high reliability. The voltage is adjusted in advance before the voltage exceeds the limit caused by the predictable load change, and the accuracy of the optimization result is improved. Compared with the traditional optimization method, it can effectively shorten the calculation time, improve the optimization calculation efficiency, reduce the possibility of voltage exceeding the limit, and improve the voltage stability.

[0093] The above scheme is described in detail below.

[0094] Example 1

[0095] See attached Figure 1 , Figure 1 FIG. 1 is a flow chart showing the main steps of a method for dynamic power optimization of a medium and low voltage distribution network according to an embodiment of the present invention. Figure 1 As shown, the method for dynamic power optimization of medium and low voltage distribution networks in the embodiment of the present invention mainly includes the following steps:

[0096] Step S101: using the grid data of the medium and low voltage distribution network in the current period as the input of the pre-trained convolutional neural network, and obtaining the node load power of the medium and low voltage distribution network in the predicted period output by the pre-trained convolutional neural network;

[0097] Step S102: Substituting the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solving it using a simulated annealing algorithm to obtain an optimization result;

[0098] Step S103: Based on the optimization result, a power optimization scheme for the low-voltage distribution network in the forecast period is obtained;

[0099] The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

[0100] In this embodiment, the power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

[0101] In a specific implementation, the node historical data includes the three-phase voltage, three-phase current, system active power and reactive power of the medium and low voltage distribution network. The source of the node load historical data is SCADA (Supervisory Control And Data Acquisition), fusion terminal, and user acquisition system; among them, the real-time load data of the low-voltage distribution station node can be converted from the user's electric energy meter; the node load historical data cycle obtained is 1 month.

[0102] In one embodiment, the training process of the pre-trained convolutional neural network includes:

[0103] The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period;

[0104] The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

[0105] In a specific implementation, the convolutional neural network includes four parts: convolution, activation, pooling, and full connection, wherein convolution, activation, and pooling are stacked alternately;

[0106] like Figure 2 As shown in Figure 2, the training process specifically includes:

[0107] 1. Data preprocessing

[0108] (1) Data identification: classify the collected data and construct a feature set of historical load data of medium and low voltage distribution network nodes;

[0109] (2) Screening and identifying data in the feature set of historical data of medium and low voltage distribution network nodes load, and eliminating redundant data and noise data;

[0110] (3) Feature set of historical data of loads at nodes in medium and low voltage distribution networks;

[0111] It is worth noting that when processing the feature set of measured data in medium and low voltage distribution stations, problems such as outliers and missing values ​​will be encountered. The existence of outliers and missing values ​​will affect the prediction accuracy. Therefore, the data needs to be processed, and the appropriate outlier processing method needs to be evaluated and selected based on the specific situation to ensure the accuracy and reliability of the prediction results.

[0112] Preferably, the present invention uses the outlier replacement method to process outliers and uses the average method to process missing values;

[0113] 2. Data Normalization

[0114] Normalization of the historical data feature set of the medium and low voltage distribution network node load;

[0115] It is worth noting that in order to improve the accuracy of the load forecasting model, data normalization is to eliminate the differences in dimensions and magnitudes between different variables, so that data analysis and modeling can be performed more accurately. In practical applications, different variables often have different units and magnitudes, which makes comparison and analysis between them difficult and may even reduce the accuracy of the model.

[0116] In the present invention, the data is normalized, that is, the original data is standardized through linear transformation, and the value range of the data is [0,1]. Since different features of the data represent different meanings and there are large differences between the values, if normalization is not performed, it will have a great impact on the optimization of the model parameters. Therefore, the present invention uses the maximum and minimum values ​​of the data features as the reference values ​​for normalization. The function expression is as follows:

[0117]

[0118] In the formula, x 1 is the normalized data, x is the original data, and x min 、x mamx are the minimum and maximum values ​​in the data sample, respectively.

[0119] 3. Dataset Division

[0120] Divide the data in the feature set of historical data of medium and low voltage distribution network node load into training set and test set;

[0121] Among them, the ratio of training set to test set is 7:3, the first 70% of the data is the training set, and the last 30% of the data is the test set. The main purpose of the training set is to be used for model training. After the model is trained, 30% of the data in the test set is predicted, and the model prediction results are compared with the test set data to verify the model accuracy. In order to make the normalized data set better fit the CNN model and complete the training of the data samples, the data is flattened, that is, converted into 1-dimensional data, and the data is transposed as the input of the CNN model.

[0122] 4. CNN Convolutional Neural Network Modeling

[0123] (1) CNN model structure and parameter setting

[0124] In the present invention, the CNN model uses 2 convolution-pooling-normalization layers, the number of convolution kernels is 16, 32 and 64 respectively, the size is 3×3, the convolution step is 1×1, the learning drop factor is 0.1, the activation function is ReLu, and the optimizer is Adam;

[0125] It is worth mentioning that the core idea of ​​the Adam optimizer is to calculate the moving average gradient and the average square gradient in each time step and use them to update the model parameters. Specifically, the Adam optimizer defines two exponentially weighted averages: the first exponentially weighted average is the exponentially weighted average of the gradient, and the second exponentially weighted average is the exponentially weighted average of the square of the gradient. These two weighted averages are used to adjust the learning rate of each parameter, thereby achieving the effect of adaptive learning rate;

[0126] Among them, the present invention adopts convolution-pooling-normalization layer, which has the effect of better extracting eigenvalues. Using a small convolution kernel can reduce the number of parameters, reduce the risk of overfitting, and improve the efficiency of extracting eigenvalues. The activation function selects the ReLu function as the activation function, which can make some neuron values ​​​​0 to prevent overfitting.

[0127] (2) Model training

[0128] The training set and the test set are used as the input of the CNN model and input into the established CNN model for training;

[0129] (3) Data denormalization

[0130] Because the data is normalized before model training, the data needs to be denormalized when outputting to get the correct output value;

[0131] 5. Prediction result output

[0132] The model output is the predicted value of the test set;

[0133] Among them, by comparing the model prediction results with the test set data, if the prediction accuracy is higher, the model is repeatedly trained until the model is optimal; using the trained model, the model data is replaced with real-time data for prediction, and the prediction results are output.

[0134] In this embodiment, the pre-constructed power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints configured for dynamic power optimization of the medium and low voltage distribution network.

[0135] In one embodiment, the first objective function is as follows:

[0136]

[0137] In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

[0138] In one embodiment, the second objective function is as follows:

[0139]

[0140] In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

[0141] In one embodiment, the constraint conditions include at least one of the following: flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation device output constraint, and on-load tap-changing transformer constraint.

[0142] In one embodiment, the power flow constraints are as follows:

[0143]

[0144] P i =P DGi -P Di

[0145] Q i =Q DGi -Q Di +Q Ci

[0146] The voltage constraints are as follows:

[0147] U i,max ≤U i ≤U i,min

[0148] The current constraints are as follows:

[0149] I i,max ≤I i ≤I i,min

[0150] The energy storage device power constraints and distributed power generation power constraints are as follows:

[0151]

[0152] The output constraints of the reactive power compensation equipment are as follows:

[0153] Q cimin Ci cimax

[0154] The constraints of the on-load tap-changing transformer are as follows:

[0155] T imin <T i <T imax

[0156] In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the reactive power compensation device output of node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.​​

[0157] In this embodiment, in the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

[0158] In one embodiment, in the process of determining whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows:

[0159]

[0160] In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

[0161] In a specific embodiment, Figure 3 As shown in the figure, the voltage-reactive power optimization model with the goal of minimizing voltage offset and network loss is solved by the simulated annealing algorithm. The simulated annealing algorithm will not stick to the local optimal solution and can accept a better solution with a certain probability, thereby increasing the ability of global search. The algorithm is essentially divided into two layers of cycles. The first layer is the number of cycles at each temperature, and the second layer is the cycle of the cooling process. The model calculation process is as follows:

[0162] 1) Initialize the simulated annealing algorithm parameters, set the initial temperature, generate the initial variables, and calculate the corresponding solution;

[0163] 2) Randomly perturb the initial variables to generate new variables within their domain, calculate the change in the objective function value to obtain a new solution, and then decide whether the new solution is accepted based on the Metropolis criterion; Metropolis criterion: if the new solution is better than the old solution, then the new solution must be accepted; if the new solution is worse than the old solution, then the new solution is accepted with a certain probability.

[0164] 3) After reaching the set number of cycles at this temperature, the temperature is lowered;

[0165] 4) Optimize result output.

[0166] Example 2

[0167] Based on the same inventive concept, the present invention also provides a medium and low voltage distribution network dynamic power optimization device, the medium and low voltage distribution network dynamic power optimization device comprising:

[0168] The first analysis module is used to use the power grid data of the medium and low voltage distribution network in the current period as the input of the pre-trained convolutional neural network to obtain the node load power of the medium and low voltage distribution network in the predicted period output by the pre-trained convolutional neural network;

[0169] The second analysis module is used to substitute the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solve it using a simulated annealing algorithm to obtain an optimization result;

[0170] A third analysis module is used to obtain a power optimization scheme for the low-voltage distribution network in the forecast period based on the optimization result;

[0171] The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

[0172] Preferably, the power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

[0173] Furthermore, the training process of the pre-trained convolutional neural network includes:

[0174] The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period;

[0175] The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

[0176] Preferably, the pre-built power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints for dynamic power optimization configuration of the medium and low voltage distribution network.

[0177] Furthermore, the first objective function is as follows:

[0178]

[0179] In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

[0180] Furthermore, the second objective function is as follows:

[0181]

[0182] In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

[0183] Furthermore, the constraint conditions include at least one of the following: flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation equipment output constraint, and on-load tap-changing transformer constraint.

[0184] Furthermore, the power flow constraints are as follows:

[0185]

[0186] P i =P DGi -P Di

[0187] Q i =Q DGi -Q Di +Q Ci

[0188] The voltage constraints are as follows:

[0189] U i,max ≤U i ≤U i,min

[0190] The current constraints are as follows:

[0191] I i,max ≤I i ≤I i,min

[0192] The energy storage device power constraints and distributed power generation power constraints are as follows:

[0193]

[0194] The output constraints of the reactive power compensation equipment are as follows:

[0195] Q cimin Ci cimax

[0196] The constraints of the on-load tap-changing transformer are as follows:

[0197] T imin <T​​i <T imax

[0198] In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the reactive power compensation device output of node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.

[0199] Preferably, in the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

[0200] Furthermore, in the process of judging whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows:

[0201]

[0202] In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

[0203] Example 3

[0204] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding function, so as to implement the steps of a dynamic power optimization method for a medium and low voltage distribution network in the above embodiment.

[0205] Example 4

[0206] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It can be understood that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of a dynamic power optimization method for a medium and low voltage distribution network in the above embodiment.

[0207] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0208] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0209] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0210] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0211] Finally, 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 above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for dynamic power optimization of medium and low voltage distribution networks, characterized in that: The method comprises: Using the grid data of the low-voltage distribution network in the current period as the input of the pre-trained convolutional neural network, and obtaining the node load power of the low-voltage distribution network in the predicted period output by the pre-trained convolutional neural network; Substituting the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solving it using a simulated annealing algorithm to obtain an optimization result; Based on the optimization results, a power optimization scheme for the low-voltage distribution network in the forecast period is obtained; The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

2. The method according to claim 1, characterized in that The power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

3. The method according to claim 2, characterized in that The training process of the pre-trained convolutional neural network includes: The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period; The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

4. The method according to claim 1, characterized in that The pre-built power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints configured for dynamic power optimization of the medium and low voltage distribution network.

5. The method according to claim 4, characterized in that The first objective function is as follows: In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

6. The method according to claim 5, characterized in that The second objective function is as follows: In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

7. The method according to claim 6, characterized in that The constraint conditions include at least one of the following: power flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation device output constraint, and on-load tap-changing transformer constraint.

8. The method according to claim 7, characterized in that The power flow constraints are as follows: P i =P DGi -P Di Q i =Q DGi -Q Di +Q Ci The voltage constraints are as follows: IN i,max ≤U i ≤U i,min The current constraints are as follows: I i,max ≤I i ≤I i,min The energy storage device power constraints and distributed power generation power constraints are as follows: The output constraints of the reactive power compensation equipment are as follows: Q cimin <Q Ci <Q cimax The constraints of the on-load tap-changing transformer are as follows: T imin <T i <T imax In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the output of the reactive power compensation device at node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.

9. The method according to claim 1, characterized in that In the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

10. The method according to claim 9, characterized in that In the process of judging whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows: In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

11. A dynamic power optimization device for medium and low voltage distribution networks, characterized in that: The device comprises: The first analysis module is used to use the power grid data of the medium and low voltage distribution network in the current period as the input of the pre-trained convolutional neural network to obtain the node load power of the medium and low voltage distribution network in the predicted period output by the pre-trained convolutional neural network; The second analysis module is used to substitute the load power of the low-voltage distribution network in the forecast period into a pre-built power optimization model and solve it using a simulated annealing algorithm to obtain an optimization result; A third analysis module is used to obtain a power optimization scheme for the low-voltage distribution network in the forecast period based on the optimization result; The optimization result includes at least one of the following: active power of nodes in the low-voltage distribution network during the forecast period, and reactive power of nodes in the low-voltage distribution network during the forecast period.

12. The device according to claim 11, characterized in that The power grid data includes: static data and dynamic data; the static data includes at least one of the following: grid structure ledger, branch information, node information; the dynamic data includes: node load power data.

13. The device according to claim 12, characterized in that The training process of the pre-trained convolutional neural network includes: The training data is constructed by using the historical grid data of the medium and low voltage distribution network and the node load power of the medium and low voltage distribution network in the corresponding historical forecast period; The initial convolutional neural network is trained using the training data to obtain the pre-trained convolutional neural network.

14. The device according to claim 11, characterized in that The pre-built power optimization model includes: a first objective function with the goal of minimizing the voltage offset of the medium and low voltage distribution network, a second objective function with the goal of minimizing the network loss of the medium and low voltage distribution network, and constraints configured for dynamic power optimization of the medium and low voltage distribution network.

15. The device according to claim 14, characterized in that The first objective function is as follows: In the above formula, is the voltage reference value of node i, and are the upper and lower limits of the voltage at node i, respectively; W is the total number of nodes in the medium and low voltage distribution network; U i is the voltage of node i, and ΔU is the voltage offset of the medium and low voltage distribution network.

16. The device according to claim 15, characterized in that The second objective function is as follows: In the above formula, g ij is the conductance of the branch between nodes i and j, U j is the voltage at node j, P loss is the network loss of the medium and low voltage distribution network, θ ij is the voltage phase angle difference between nodes i and j.

17. The device according to claim 16, characterized in that The constraint conditions include at least one of the following: power flow constraint, voltage constraint, current constraint, energy storage device power output constraint, distributed power source power output constraint, reactive power compensation device output constraint, and on-load tap-changing transformer constraint.

18. The device according to claim 17, characterized in that The power flow constraints are as follows: P i =P DGi -P Di Q i =Q DGi -Q Di +Q Ci The voltage constraints are as follows: IN i,max ≤U i ≤U i,min The current constraints are as follows: I i,max ≤I i ≤I i,min The energy storage device power constraints and distributed power generation power constraints are as follows: The output constraints of the reactive power compensation equipment are as follows: Q cimin <Q Ci <Q cimax The constraints of the on-load tap-changing transformer are as follows: T imin <T i <T imax In the above formula, P i and Q i are the active and reactive power of node i in the prediction period, P Di and Q Di are the load power of node i in the forecast period, P DGi and Q DGi is the active and reactive power of the distributed generation at node i, G ij is the conductance between nodes i and j, Q Ci is the output of the reactive power compensation device at node i, B ij is the susceptance between nodes i and j, U i,max and U i,min are the maximum and minimum voltage values ​​of node i, I i,max and I i,min are the maximum and minimum current flowing through node i, I i is the current flowing through node i, P i,pv is the photovoltaic active output of node i, P i,ev is the active output of the electric vehicle charging pile at node i, SOC is the energy storage capacity, P i,pv,min , P i,ev,min , P i,pv,max , P i,ev,max They are the minimum photovoltaic output power of node i, the minimum active output power of the electric vehicle charging pile, the maximum photovoltaic output power, the maximum active output power of the electric vehicle charging pile, and the SOC min is the minimum energy storage capacity, SOC max is the maximum energy storage capacity, N DG is the total number of distributed generation, Q cimin , Q cimax are the minimum and maximum output values ​​of the reactive power compensation equipment at node i, T imin , T imax are the minimum and maximum values ​​of the on-load tap-changing transformer at node i, T i It is the gear position of the on-load tap-changing transformer at node i.

19. The device according to claim 11, characterized in that In the process of substituting the load power of the low-voltage distribution network in the forecast period into the pre-built power optimization model and solving it using the simulated annealing algorithm, it is determined whether to accept the new solution based on the Metropolis criterion.

20. The device according to claim 19, characterized in that In the process of judging whether to accept a new solution based on the Metropolis criterion, the probability of accepting a new solution is as follows: In the above formula, P is the probability of accepting a new solution, E t+1 is the solution obtained by the t+1th iteration, E t is the solution obtained by the t-th iteration calculation, T is the temperature at the t-th iteration calculation, and e is a natural constant.

21. A computer device, characterized in that: include: one or more processors; The processor is configured to execute one or more programs; When the one or more programs are executed by the one or more processors, the medium and low voltage distribution network dynamic power optimization method as described in any one of claims 1 to 10 is implemented.

22. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the dynamic power optimization method for medium and low voltage distribution networks as described in any one of claims 1 to 10 is implemented.

Images