A reactive power optimization control method for a direct current power transmission system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA SOUTHERN POWER GRID EXTRA HIGH VOLTAGE POWER TRANSMISSION CO LIUZHOU BRANCH
- Filing Date
- 2023-12-28
- Publication Date
- 2026-07-24
Smart Images
Figure CN117791631B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of reactive power optimization control technology for DC transmission lines, and in particular to a reactive power optimization control method for DC transmission systems. Background Technology
[0002] With the advancement of the construction of new power systems, ultra-high voltage direct current (UHVDC) transmission technology has been widely applied. The influence of DC lines on the AC power grid is increasing, leading to the gradual prominence of the "strong DC, weak AC" problem. DC transmission line faults pose a huge challenge to the reactive power and voltage support of the power grid. Therefore, it is urgent to study reactive power optimization control technology for DC transmission lines to improve voltage regulation capabilities and thus enhance the safe and stable operation of the power grid.
[0003] Currently, deep learning has found some applications in reactive power optimization due to its strong nonlinear learning capabilities. For example, some experts have proposed reactive power optimization methods based on deep Q-networks to achieve optimal reactive power control under grid imbalance, while others have proposed reactive power scheduling strategies based on deep belief networks to learn the relationship between grid operating status and reactive power control. However, the reactive power control strategies obtained using these methods cannot minimize network losses and voltage deviations in DC transmission systems, thus limiting their application effectiveness on DC transmission lines. Summary of the Invention
[0004] The purpose of this application is to at least address one of the aforementioned technical deficiencies, particularly the technical deficiency that the existing technology uses deep learning to calculate the reactive power control strategy of DC transmission systems, which fails to minimize network losses and voltage deviations in DC transmission systems, thus limiting its application effectiveness on DC transmission lines.
[0005] This application provides a reactive power optimization control method for a DC transmission system, the method comprising:
[0006] Acquire historical power transmission data of the DC transmission system during a preset historical period, wherein the historical power transmission data includes the historical operating status of the DC transmission system;
[0007] Based on the principles of minimizing network loss and voltage deviation, a reactive power optimization model for the DC transmission system is constructed.
[0008] The reactive power optimization model is solved using the historical power transmission data, and the historical optimal reactive power control strategy of the DC transmission system is determined based on the solution results.
[0009] Using the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, a preset temporal convolutional residual network is trained to obtain a reactive power control strategy prediction model.
[0010] The reactive power control strategy prediction model is used to predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy is the number of capacitors switched on and off in the DC transmission system.
[0011] Optionally, the step of training a preset temporal convolutional residual network using the historical operating states of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels to obtain a reactive power control strategy prediction model includes:
[0012] Determine the temporal convolutional residual network;
[0013] The historical operating state of the DC transmission system is used as a training sample and input into the temporal convolutional residual network to obtain the predicted reactive power control strategy output by the temporal convolutional residual network.
[0014] The historical optimal reactive power control strategy is used as the sample label, and the network parameters in the temporal convolutional residual network are updated with the goal of the predicted reactive power control strategy approaching the historical optimal reactive power control strategy.
[0015] When the preset training conditions are met, the trained temporal convolutional residual network is used as the reactive power control strategy prediction model.
[0016] Optionally, the determination of the temporal convolutional residual network includes:
[0017] Determine the number of network layers and neurons in the temporal convolutional residual network to be constructed;
[0018] Construct a temporal convolutional residual network based on the number of network layers and the number of neurons.
[0019] Optionally, constructing a temporal convolutional residual network based on the number of network layers and the number of neurons includes:
[0020] Multiple sets of datasets are formed by acquiring multiple historical operating states of the DC transmission system and the corresponding historical optimal reactive power control strategies.
[0021] Multiple sets of neurons with different numbers are set according to the number of network layers, and multiple temporal convolutional residual networks are constructed using multiple sets of neurons with different numbers.
[0022] For each temporal convolutional residual network, multiple datasets are used for training, and the policy deviation rate between the predicted reactive power control policy output by each temporal convolutional residual network and the corresponding historical optimal reactive power control policy is determined.
[0023] The temporal convolutional residual network with the smallest policy bias rate is selected as the final temporal convolutional residual network for training.
[0024] Optionally, the formula for calculating the strategy deviation rate is:
[0025]
[0026] in, This is the strategy deviation rate. This represents the actual number of capacitors switched on and off at the i-th reactive power optimization node in the historically optimal reactive power control strategy. To predict the number of capacitors switched on and off at the i-th reactive power optimization node in the reactive power control strategy; K is the number of reactive power optimization nodes.
[0027] Optionally, the historical operating status of the DC transmission system includes the historical active power and historical reactive power of each node in the DC transmission system;
[0028] Before inputting the historical operating status of the DC transmission system as training samples into the temporal convolutional residual network, the method further includes:
[0029] Based on the historical active power and historical reactive power of each node in the DC transmission system, an input matrix is constructed;
[0030] The elements in the input matrix are normalized, and the normalization process is calculated using the following formula:
[0031]
[0032] in, Given the i-th element of the input matrix X, Let X be the minimum value of the i-th element in the input matrix X. This is the maximum value of the i-th element in the input matrix X.
[0033] Optionally, the reactive power optimization model of the DC transmission system, based on the principles of minimizing network losses and voltage deviation, includes:
[0034] Based on the principles of minimizing network loss and voltage deviation, the objective function and constraints are determined. The constraints include at least node voltage constraints, power flow constraints, converter characteristic constraints, control angle constraints, reactive power regulation capability constraints, and switching dead zone constraints.
[0035] Based on the objective function, the node voltage constraint, the power flow constraint, the converter characteristic constraint, the control angle constraint, the reactive power supply regulation capability constraint, and the switching dead zone constraint, a reactive power optimization model for the DC transmission system is constructed.
[0036] This application also provides a reactive power optimization control device for a DC transmission system, comprising:
[0037] The historical data acquisition module is used to acquire historical transmission data of the DC transmission system during a preset historical period, wherein the historical transmission data includes the historical operating status of the DC transmission system;
[0038] The optimization model building module is used to build a reactive power optimization model for the DC transmission system based on the principles of minimizing network loss and voltage deviation.
[0039] The control strategy solving module is used to solve the reactive power optimization model using the historical power transmission data, and to determine the historical optimal reactive power control strategy of the DC transmission system based on the solution results.
[0040] The prediction model training module is used to train a preset temporal convolutional residual network using the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, so as to obtain a reactive power control strategy prediction model.
[0041] The reactive power control strategy prediction model is used to predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy is the number of capacitors switched on and off in the DC transmission system.
[0042] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of a reactive power optimization control method for a DC transmission system as described in any of the above embodiments.
[0043] This application also provides a computer device, including: one or more processors, and memory;
[0044] The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of a reactive power optimization control method for a DC transmission system as described in any of the above embodiments.
[0045] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:
[0046] This application provides a reactive power optimization control method for a DC transmission system. First, historical transmission data of the DC transmission system over a preset historical period is acquired, including the historical operating status of the DC transmission system. Then, a reactive power optimization model of the DC transmission system is constructed based on the principles of minimizing network loss and voltage deviation. This model can then be solved using the historical transmission data, and the historically optimal reactive power control strategy of the DC transmission system is determined based on the solution results. Next, this application trains a preset temporal convolutional residual network using the historical operating status of the DC transmission system as training samples and the historically optimal reactive power control strategy as sample labels, thereby obtaining a reactive power control strategy prediction model. This model can predict the current reactive power of the DC transmission system based on its current operating status. Control strategies, such as the number of capacitors switched on and off in a DC transmission system; in this process, since the application trains the time-series convolutional residual network using the historical optimal reactive power control strategy output by the reactive power optimization model, and the reactive power optimization model of this application is constructed based on the principle of minimizing network loss and voltage deviation, after training the time-series convolutional residual network with the historical optimal reactive power control strategy output by the reactive power optimization model, the time-series convolutional residual network can not only learn the optimal control strategy that minimizes network loss and voltage deviation of the DC transmission system, but also take advantage of its own advantages of low information loss of time-series convolution, wide receptive field and strong deep feature extraction capability of residual network to learn the time series characteristics of DC transmission system, thereby improving the application effect of deep learning algorithm on dynamically changing DC transmission lines. Attached Figure Description
[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0048] Figure 1 A flowchart illustrating a reactive power optimization control method for a DC transmission system provided in this application embodiment;
[0049] Figure 2 A schematic diagram illustrating the relationship between the network loss function and the number of iterations provided in this application embodiment;
[0050] Figure 3 A schematic diagram illustrating the relationship between the policy bias rate and the number of neurons in the temporal convolutional residual network provided in the embodiments of this application;
[0051] Figure 4A geographical wiring diagram of an ultra-high voltage direct current system in a certain region is provided for an embodiment of this application;
[0052] Figure 5 Equivalent circuit diagram of the geographical wiring diagram provided in the embodiments of this application;
[0053] Figure 6 A schematic diagram of network loss rate under different methods obtained by comparing the reactive power control strategy prediction model provided in the embodiments of this application with extreme learning machine and convolutional neural network;
[0054] Figure 7 A schematic diagram of voltage offset rates obtained by comparing the reactive power control strategy prediction model provided in the embodiments of this application with extreme learning machine and convolutional neural network under different methods;
[0055] Figure 8 A schematic diagram of the structure of a reactive power optimization control device for a DC transmission system provided in this application embodiment;
[0056] Figure 9 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation
[0057] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0058] In one embodiment, such as Figure 1 As shown, Figure 1 This application provides a flowchart illustrating a reactive power optimization control method for a DC transmission system. The method may include:
[0059] S110: Obtain historical power transmission data of the DC transmission system during a preset historical period.
[0060] In this embodiment, when performing reactive power optimization control on a DC transmission system, this application can obtain historical transmission data of the DC transmission system during a preset historical period. In this way, the historical optimal reactive power control strategy of the DC transmission system can be predicted using the historical transmission data, and a preset temporal convolutional residual network can be trained according to the historical optimal reactive power control strategy to obtain a reactive power control strategy prediction model.
[0061] Furthermore, since historical transmission data of DC transmission systems fluctuates due to natural factors such as wind and sunlight, this application proposes setting the preset historical period to a relatively long timeframe, such as one year or two years, to obtain a more accurate reactive power control strategy. This allows for the acquisition of historical transmission data under different wind and sunlight conditions throughout the year, and enables the subsequent model to learn reactive power control strategies under different power grid scenarios during training.
[0062] Furthermore, this application uses historical transmission data to predict the historical optimal reactive power control strategy of a DC transmission system. When training a pre-defined time-series convolutional residual network according to this historical optimal reactive power control strategy, the input of the time-series convolutional residual network can be pre-defined as the operating state of the DC transmission system, and the output as the reactive power control strategy. This allows for direct prediction of the corresponding reactive power control strategy based on the current operating state of the DC transmission system. Based on this, when acquiring historical transmission data of the DC transmission system, this application can also acquire the historical operating state of the DC transmission system, such as the historical active power and historical reactive power of each node in the DC transmission system. This provides sufficient training samples for the subsequent time-series convolutional residual network.
[0063] S120: Construct a reactive power optimization model for DC transmission systems based on the principles of minimizing network losses and voltage deviation.
[0064] In this step, after obtaining the historical transmission data of the DC transmission system in the preset historical period through S110, this application can construct a reactive power optimization model of the DC transmission system. The reactive power optimization model is constructed based on the principle of minimizing network loss and voltage deviation. The reactive power optimization model obtained in this way can output the optimal reactive power control strategy with the minimum network loss and voltage deviation. The optimal reactive power control strategy output by the reactive power optimization model can significantly improve the training effect of the temporal convolutional residual network.
[0065] S130: Solve the reactive power optimization model using historical power transmission data, and determine the historical optimal reactive power control strategy for the DC transmission system based on the solution results.
[0066] In this step, after constructing a reactive power optimization model based on the principles of minimum network loss and minimum voltage deviation through S120, this application can input the historical power transmission data obtained in S110 into the reactive power optimization model and optimize and solve the reactive power optimization model. In this way, the historical optimal reactive power control strategy of the DC power transmission system can be determined based on the solution results. Through the historical optimal reactive power control strategy, a time-series convolutional residual network with good reactive power optimization effect can be trained.
[0067] Furthermore, the process of optimizing the reactive power optimization model in this application can be implemented using existing technologies. Specifically, the appropriate solution algorithm can be selected according to the actual situation, and no restrictions are imposed here.
[0068] S140: Using the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, a preset time-series convolutional residual network is trained to obtain a reactive power control strategy prediction model.
[0069] In this step, after obtaining historical transmission data of the DC transmission system through S110, this application can use the historical transmission data to solve the reactive power optimization model through S130, and determine the historical optimal reactive power control strategy of the DC transmission system based on the solution results. This allows the historical operating states in the historical transmission data to be used as training samples, and the historical optimal reactive power control strategy output by the reactive power optimization model to be used as sample labels. After iteratively training the preset temporal convolutional residual network using the training samples and sample labels, the final reactive power control strategy prediction model can be obtained. This reactive power control strategy prediction model can predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy can be the number of capacitors switched on and off in the DC transmission system. This application improves voltage regulation capability through DC transmission line reactive power optimization control technology by predicting the number of capacitors switched on and off during the DC transmission process and then switching on or off capacitors in the DC transmission system based on the prediction results, thereby enhancing the safe and stable operation level of the power grid.
[0070] In the above embodiments, this application can first obtain historical transmission data of the DC transmission system during a preset historical period, including the historical operating status of the DC transmission system; then, based on the principles of minimizing network loss and voltage deviation, a reactive power optimization model of the DC transmission system is constructed. This allows the historical transmission data to be used to solve the reactive power optimization model, and the historical optimal reactive power control strategy of the DC transmission system can be determined based on the solution results. Next, this application can use the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels to train a preset temporal convolutional residual network, thereby obtaining a reactive power control strategy prediction model. This reactive power control strategy prediction model can predict the current reactive power control strategy of the DC transmission system based on its current operating status, such as the current reactive power control strategy of the DC transmission system. The number of capacitors switched on and off in the DC transmission system; in this process, since the application trains the time-series convolutional residual network using the historical optimal reactive power control strategy output by the reactive power optimization model, and the reactive power optimization model of this application is constructed based on the principle of minimizing network loss and voltage deviation, after the application trains the time-series convolutional residual network using the historical optimal reactive power control strategy output by the reactive power optimization model, the time-series convolutional residual network can not only learn the optimal control strategy that minimizes network loss and voltage deviation of the DC transmission system, but also take advantage of its own advantages of low information loss of time-series convolution, wide receptive field and strong deep feature extraction capability of residual network to learn the time series characteristics of DC transmission system, thereby improving the application effect of deep learning algorithm on dynamically changing DC transmission lines.
[0071] In one embodiment, in S140, the historical operating states of the DC transmission system are used as training samples, and the historical optimal reactive power control strategy is used as sample labels to train a preset temporal convolutional residual network to obtain a reactive power control strategy prediction model, which may include:
[0072] S141: Determine the temporal convolutional residual network.
[0073] S142: Input the historical operating state of the DC transmission system as training samples into the temporal convolutional residual network to obtain the predicted reactive power control strategy output by the temporal convolutional residual network.
[0074] S143: Using the historical optimal reactive power control strategy as the sample label, and with the goal of the predicted reactive power control strategy approaching the historical optimal reactive power control strategy, the network parameters in the temporal convolutional residual network are updated.
[0075] S144: When the preset training conditions are met, the trained temporal convolutional residual network is used as the reactive power control strategy prediction model.
[0076] In this embodiment, when training the temporal convolutional residual network, a traditional temporal convolutional residual network can be obtained first. Then, the historical operating state of the DC transmission system is used as a training sample and input into the temporal convolutional residual network. By leveraging the advantages of the temporal convolutional residual network itself, such as low loss of temporal convolutional information, wide receptive field, and strong deep feature extraction capability, the mapping relationship between the operating state of the DC transmission system and the reactive power control strategy can be discovered, and the corresponding predicted reactive power control strategy can be output.
[0077] Next, this application uses historically optimal reactive power control strategies as sample labels and updates the network parameters in the temporal convolutional residual network with the goal of predicting reactive power control strategies that approximate historically optimal strategies. When preset training conditions are met, the trained temporal convolutional residual network can be used as the final reactive power control strategy prediction model. This reactive power control strategy prediction module can predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy can be the number of capacitors switched on and off in the DC transmission system.
[0078] Furthermore, the preset training conditions in this application can be either the number of iterations or the loss function value. For example, to verify the training effect of the temporal convolutional residual network, this application makes the following... Figure 2 The relationship between the network loss function and the number of iterations is shown. Figure 2 As can be seen, when the number of iterations reaches 500, the network tends to stabilize and reaches the minimum loss function value of 0.41, at which point the network training effect is good.
[0079] In the above embodiments, by predicting the number of capacitors to be switched on and off during the transmission process of the DC transmission system, and by switching on or off the capacitors in the DC transmission system according to the prediction results, the voltage regulation capability is improved through the reactive power optimization control technology of DC transmission lines, thereby enhancing the safe and stable operation level of the power grid.
[0080] In one embodiment, determining the temporal convolutional residual network in S141 may include:
[0081] S1411: Determine the number of network layers and neurons of the temporal convolutional residual network to be constructed.
[0082] S1412: Construct a temporal convolutional residual network based on the number of network layers and the number of neurons.
[0083] In this embodiment, when determining the temporal convolutional residual network, a traditional temporal convolutional residual network can be obtained, or the number of network layers and the number of neurons in each layer of the traditional temporal convolutional residual network can be adjusted according to the application scenario of this application, and then the temporal convolutional residual network can be constructed based on the adjusted number of network layers and the number of neurons.
[0084] Furthermore, the temporal convolutional residual network of this application can also use the ReLU activation function, which can make some neurons have a value of 0, thus preventing overfitting.
[0085] In one embodiment, constructing a temporal convolutional residual network based on the number of network layers and the number of neurons in S1412 may include:
[0086] S4121: After obtaining multiple historical operating states of the DC transmission system and the corresponding historical optimal reactive power control strategies, multiple datasets are formed.
[0087] S4122: Set up multiple groups of neurons with different numbers of neurons according to the number of network layers, and use multiple groups of neurons with different numbers of neurons to construct multiple temporal convolutional residual networks.
[0088] S4123: For each temporal convolutional residual network, train it using multiple datasets and determine the policy deviation rate between the predicted reactive power control policy output by each temporal convolutional residual network and the corresponding historical best reactive power control policy.
[0089] S4124: Select the temporal convolutional residual network with the smallest policy bias rate as the final trained temporal convolutional residual network.
[0090] In this embodiment, when constructing a temporal convolutional residual network based on the number of network layers and neurons, multiple historical operating states of the DC transmission system and corresponding historical optimal reactive power control strategies can be obtained first to form multiple datasets. Then, multiple sets of neurons with different numbers can be set according to the preset number of network layers, and multiple temporal convolutional residual networks can be constructed using these sets of neurons. Each temporal convolutional residual network is then trained using the multiple datasets. After calculating the policy deviation rate between the predicted reactive power control strategy output by each temporal convolutional residual network and the corresponding historical optimal reactive power control strategy, the change in the policy deviation rate of each temporal convolutional residual network is observed. The temporal convolutional residual network with the smallest policy deviation rate is selected as the final trained temporal convolutional residual network.
[0091] For example, this application can set the number of neurons in the temporal convolutional residual network from 100 to 500, with each network layer allocated the same number of neurons; then, this application can use 200 datasets to train the temporal convolutional residual network with different numbers of neurons and observe the policy bias rate. The changes, The minimum number of neurons is determined using the principle of minimum number. Experimental results show that... Figure 3 The relationship between the policy bias rate and the number of neurons in the temporal convolutional residual network is shown; from Figure 3 As can be seen, when the number of neurons is 300, the policy bias rate... Therefore, this application can set the number of neurons in the final temporal convolutional residual network to be 300.
[0092] In the above embodiments, the effectiveness of network training is expressed as the policy bias rate. This indicates the deviation between the predicted reactive power control strategy output by the network and the corresponding historical optimal reactive power control strategy. The smaller the value, the better the reactive power control strategy of the network. Therefore, after selecting the number of neurons that minimizes the network strategy deviation rate through the above method, the training effect of the temporal convolutional residual network constructed with this number of neurons and network layers is better.
[0093] In one embodiment, the formula for calculating the strategy deviation rate is:
[0094]
[0095] in, This is the strategy deviation rate. This represents the actual number of capacitors switched on and off at the i-th reactive power optimization node in the historically optimal reactive power control strategy. To predict the number of capacitors switched on and off at the i-th reactive power optimization node in the reactive power control strategy; K is the number of reactive power optimization nodes.
[0096] In one embodiment, the historical operating status of the DC transmission system includes the historical active power and historical reactive power of each node in the DC transmission system.
[0097] Before inputting the historical operating status of the DC transmission system as training samples into the temporal convolutional residual network in step S142, it may further include:
[0098] S401: Construct an input matrix based on the historical active power and historical reactive power of each node in the DC transmission system.
[0099] S402: Normalize each element in the input matrix. The normalization calculation formula is as follows:
[0100]
[0101] in, Given the i-th element of the input matrix X, Let X be the minimum value of the i-th element in the input matrix X. This is the maximum value of the i-th element in the input matrix X.
[0102] In this embodiment, after using the acquired historical operating states of the DC transmission system as training samples, the input matrix X can be constructed based on the historical active power and historical reactive power of each node in the historical operating states:
[0103]
[0104] Next, this application can normalize each element in the input matrix X using the calculation formula described above during normalization. This can highlight the changes in the data, accelerate the network convergence speed, and improve the network accuracy.
[0105] In one embodiment, the reactive power optimization model of the DC transmission system constructed in S120 based on the principles of minimizing network loss and voltage deviation may include:
[0106] S121: Based on the principles of minimizing network loss and voltage deviation, determine the objective function and constraints, wherein the constraints include at least node voltage constraints, power flow constraints, converter characteristic constraints, control angle constraints, reactive power supply regulation capability constraints, and switching dead zone constraints.
[0107] S122: Based on the objective function, the node voltage constraint, the power flow constraint, the converter characteristic constraint, the control angle constraint, the reactive power supply regulation capability constraint, and the switching dead zone constraint, construct the reactive power optimization model of the DC transmission system.
[0108] In this embodiment, when constructing the reactive power optimization model, this application can determine the objective function and constraints based on the principles of minimizing network loss and minimizing voltage deviation, and then construct the reactive power optimization model of the DC transmission system based on the objective function and constraints.
[0109] Specifically, this application constructs a reactive power optimization model for DC transmission lines based on the principles of minimizing network loss and voltage deviation, as shown below:
[0110]
[0111] In the formula: Let i be the active power of the i-th generator; The active power of the j-th load; Let be the DC power of the k-th converter station; For converter stations, the state coefficient is [value]. For rectifier stations, [value] is [value]. The value is 1; if it is an inverter station, The value is -1; , , , These represent the voltage amplitude, voltage over-limit threshold, upper voltage limit, and lower voltage limit of node i, respectively. This is a penalty factor for voltage exceeding the limit, and its value is usually set to be large. In this application, it can be set to 1000. , , , These are sets of generators, loads, converter stations, and voltage balance nodes.
[0112] Furthermore, the constraints of the reactive power optimization model in this application include, but are not limited to, node voltage constraints, power flow constraints, converter characteristic constraints, control angle constraints, reactive power source regulation capability constraints, and switching dead zone constraints. The specific constraints are as follows:
[0113] (1) Node voltage constraints:
[0114]
[0115] In the formula: , These are the upper and lower limits of the voltage at the i-th node, respectively.
[0116] (2) Current constraints:
[0117]
[0118] In the formula: Let be the DC power of the i-th node converter station; Let i be the reactive power source of the i-th node; Let be the reactive power compensation capacity of the i-th node; Let i be the reactive load of the i-th node; Let be the reactive power of the i-th node converter station; For the set of all nodes; , Let be the active power and reactive power of node i, respectively. The calculation formula is as follows:
[0119]
[0120] In the formula: The total number of nodes; Let be the voltage amplitude of the i-th node; This represents the phase angle difference between the i-th node and the j-th node; The real part of the admittance matrix; This is the imaginary part of the admittance matrix.
[0121] Reactive power compensation capacity The calculation formula is as follows:
[0122]
[0123] In the formula: This is the reactive power compensation coefficient. If the reactive power compensation device is a reactor, then this value is 1; if the reactive power compensation device is a capacitor / filter, then this value is -1. The number of compensation devices; This is the rated voltage of the reactive power compensation device; This refers to the rated capacity of the reactive power compensation device. This is the set of all compensation nodes.
[0124] (3) Converter characteristic constraints:
[0125]
[0126] In the formula: , , , , These are the DC voltage, DC current, control angle, transformer ratio, and apparent power of the i-th node converter station, respectively. , These refer to the number of converters and the number of poles, respectively. For commutation reactance; This is the commutation overlap factor, with a value of 0.995.
[0127] Transformer turns ratio The calculation formula is as follows:
[0128]
[0129] In the formula: , These are the rated voltages for the low-voltage and high-voltage sides, respectively. , These represent the transformer tap position and voltage regulation step size of the i-th node converter station, respectively.
[0130] (4) Control angle constraints:
[0131]
[0132] In the formula: , These are the upper and lower limits of the control angle for the i-th node, respectively.
[0133] (5) Reactive power supply regulation capability constraints:
[0134]
[0135] In the formula: , These are the upper and lower limits of the reactive power output of the i-th node, respectively.
[0136] (6) Throwing dead zone constraint:
[0137] The dead-zone constraints for capacitor / filter switching are shown below:
[0138]
[0139] In the formula: Ideal reactive power; To compensate for the dead zone of the equipment.
[0140] Furthermore, after introducing a reactive power optimization control method for a DC transmission system through the above-described multiple embodiments, the reactive power control method of this application will be described in detail below through a specific example:
[0141] like Figure 4 , 5 As shown, Figure 4 This is a geographical wiring diagram of an ultra-high voltage direct current system in a certain region, provided in an embodiment of this application. Figure 5 Equivalent circuit diagram of the geographical wiring diagram provided in the embodiments of this application; Figure 4 The geographical wiring diagram includes substations, power plants, converter stations, and equivalent circuits such as... Figure 5 As shown, the total number of nodes is 85, with 85 nodes for the generator and 18 for the parallel capacitor. The entire DC system has a rated power of 10000MW and a rated voltage of 800kV. The equivalent parameters of each component in the system are shown in Table 1.
[0142]
[0143] Table 1 Energy Storage Device Parameters
[0144] This application uses a pre-trained reactive power control strategy prediction model to predict the number of capacitors to be switched on or off in the aforementioned UHVDC system. The network input consists of operating parameters: active and reactive power at 85 nodes; the network output is the reactive power optimization control strategy: the number of capacitors to be switched on or off. Therefore, the network input data dimension is 85×2, and the output dimension is 18×1.
[0145] Furthermore, to verify the reactive power optimization effect of this application, the reactive power control strategy prediction model of this application is compared with extreme learning machine and convolutional neural network to obtain the network loss rate and voltage deviation rate under different methods:
[0146]
[0147] In the formula: Network loss rate; This represents the actual network loss. To optimize the network loss of the method; Voltage offset rate; The rated voltage of the i-th node; This represents the actual voltage value of the i-th node.
[0148] The network loss rate and voltage offset results for different methods are as follows: Figure 6 , Figure 7 As shown. From Figure 6 , Figure 7 As can be seen from the results, the proposed method has the lowest network loss rate and voltage deviation rate, which verifies that the proposed method has the best effect on reactive power optimization of DC transmission lines.
[0149] Furthermore, to verify that this method performs well in various scenarios, this application can be validated on power grid scenarios with different wind forces and sunlight throughout the four seasons. For each scenario, the training set size is 600 and the test set size is 200. The network loss rate and voltage deviation rate of different methods under different scenarios are shown in Table 2:
[0150]
[0151] Table 2 Comparison of results in different scenarios
[0152] As shown in Table 2, the proposed method exhibits the lowest network loss rate and voltage deviation rate across various scenarios, validating its effectiveness and practicality. This application can effectively and rapidly achieve active voltage regulation of DC transmission lines, improving their voltage regulation capability, reducing network losses, and contributing to the safe and stable operation of the power system.
[0153] The following describes a reactive power optimization control device for a DC transmission system provided in an embodiment of this application. The reactive power optimization control device for a DC transmission system described below can be referred to in correspondence with the reactive power optimization control method for a DC transmission system described above.
[0154] In one embodiment, such as Figure 8 As shown, Figure 8 This application provides a schematic diagram of a reactive power optimization control device for a DC transmission system, as shown in an embodiment of the present application. The application also provides a reactive power optimization control device for a DC transmission system, including a historical data acquisition module 210, an optimization model construction module 220, a control strategy solution module 230, and a prediction model training module 240, specifically comprising the following:
[0155] The historical data acquisition module 210 is used to acquire historical power transmission data of the DC transmission system during a preset historical period. The historical power transmission data includes the historical operating status of the DC transmission system.
[0156] The optimization model construction module 220 is used to construct the reactive power optimization model of the DC transmission system based on the principles of minimizing network loss and voltage deviation.
[0157] The control strategy solving module 230 is used to solve the reactive power optimization model using the historical power transmission data, and to determine the historical optimal reactive power control strategy of the DC power transmission system based on the solution results.
[0158] The prediction model training module 240 is used to train a preset temporal convolutional residual network using the historical operating state of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, so as to obtain a reactive power control strategy prediction model.
[0159] The reactive power control strategy prediction model is used to predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy is the number of capacitors switched on and off in the DC transmission system.
[0160] In the above embodiments, this application can first obtain historical transmission data of the DC transmission system during a preset historical period, including the historical operating status of the DC transmission system; then, based on the principles of minimizing network loss and voltage deviation, a reactive power optimization model of the DC transmission system is constructed. This allows the historical transmission data to be used to solve the reactive power optimization model, and the historical optimal reactive power control strategy of the DC transmission system can be determined based on the solution results. Next, this application can use the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels to train a preset temporal convolutional residual network, thereby obtaining a reactive power control strategy prediction model. This reactive power control strategy prediction model can predict the current reactive power control strategy of the DC transmission system based on its current operating status, such as the current reactive power control strategy of the DC transmission system. The number of capacitors switched on and off in the DC transmission system; in this process, since the application trains the time-series convolutional residual network using the historical optimal reactive power control strategy output by the reactive power optimization model, and the reactive power optimization model of this application is constructed based on the principle of minimizing network loss and voltage deviation, after the application trains the time-series convolutional residual network using the historical optimal reactive power control strategy output by the reactive power optimization model, the time-series convolutional residual network can not only learn the optimal control strategy that minimizes network loss and voltage deviation of the DC transmission system, but also take advantage of its own advantages of low information loss of time-series convolution, wide receptive field and strong deep feature extraction capability of residual network to learn the time series characteristics of DC transmission system, thereby improving the application effect of deep learning algorithm on dynamically changing DC transmission lines.
[0161] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of a reactive power optimization control method for a DC transmission system as described in any of the above embodiments.
[0162] In one embodiment, this application also provides a computer device, including: one or more processors, and memory.
[0163] The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of a reactive power optimization control method for a DC transmission system as described in any of the above embodiments.
[0164] Indicatively, such as Figure 9 As shown, Figure 9 This is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. (Refer to...) Figure 9The computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions, such as application programs, that can be executed by the processing component 302. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the reactive power optimization control method for the high-voltage direct current transmission system described in any of the above embodiments.
[0165] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.
[0166] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0167] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0168] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.
[0169] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A reactive power optimization control method for a DC transmission system, characterized in that, The method includes: Acquire historical power transmission data of the DC transmission system during a preset historical period, wherein the historical power transmission data includes the historical operating status of the DC transmission system; Based on the principles of minimizing network loss and voltage deviation, a reactive power optimization model for the DC transmission system is constructed. The reactive power optimization model is solved using the historical power transmission data, and the historical optimal reactive power control strategy of the DC transmission system is determined based on the solution results. Using the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, a preset temporal convolutional residual network is trained to obtain a reactive power control strategy prediction model. The reactive power control strategy prediction model is used to predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy is the number of capacitors switched on and off in the DC transmission system. The process involves training a pre-defined temporal convolutional residual network using the historical operating states of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels to obtain a reactive power control strategy prediction model, including: Determine the number of network layers and neurons in the temporal convolutional residual network to be constructed; Multiple sets of datasets are formed by acquiring multiple historical operating states of the DC transmission system and the corresponding historical optimal reactive power control strategies. Multiple sets of neurons with different numbers are set according to the number of network layers, and multiple temporal convolutional residual networks are constructed using multiple sets of neurons with different numbers. For each temporal convolutional residual network, multiple datasets are used for training, and the policy deviation rate between the predicted reactive power control policy output by each temporal convolutional residual network and the corresponding historical optimal reactive power control policy is determined. The temporal convolutional residual network with the smallest policy bias rate is selected as the final trained temporal convolutional residual network; The historical operating state of the DC transmission system is used as a training sample and input into the temporal convolutional residual network to obtain the predicted reactive power control strategy output by the temporal convolutional residual network. The historical optimal reactive power control strategy is used as the sample label, and the network parameters in the temporal convolutional residual network are updated with the goal of the predicted reactive power control strategy approaching the historical optimal reactive power control strategy. When the preset training conditions are met, the trained temporal convolutional residual network is used as the reactive power control strategy prediction model.
2. The reactive power optimization control method for a DC transmission system according to claim 1, characterized in that, The formula for calculating the strategy deviation rate is as follows: in, This is the strategy deviation rate. This represents the actual number of capacitors switched on and off at the i-th reactive power optimization node in the historically optimal reactive power control strategy. To predict the number of capacitors switched on and off at the i-th reactive power optimization node in the reactive power control strategy; K is the number of reactive power optimization nodes.
3. The reactive power optimization control method for a DC transmission system according to claim 1, characterized in that, The historical operating status of the DC transmission system includes the historical active power and historical reactive power of each node in the DC transmission system; Before inputting the historical operating status of the DC transmission system as training samples into the temporal convolutional residual network, the method further includes: Based on the historical active power and historical reactive power of each node in the DC transmission system, an input matrix is constructed; The elements in the input matrix are normalized, and the normalization process is calculated using the following formula: in, Given the i-th element of the input matrix X, Let X be the minimum value of the i-th element in the input matrix X. This is the maximum value of the i-th element in the input matrix X.
4. A reactive power optimization control method for a DC transmission system according to any one of claims 1-3, characterized in that, The reactive power optimization model for the DC transmission system, based on the principles of minimizing network loss and voltage deviation, includes: Based on the principles of minimizing network loss and voltage deviation, the objective function and constraints are determined. The constraints include at least node voltage constraints, power flow constraints, converter characteristic constraints, control angle constraints, reactive power regulation capability constraints, and switching dead zone constraints. Based on the objective function, the node voltage constraint, the power flow constraint, the converter characteristic constraint, the control angle constraint, the reactive power supply regulation capability constraint, and the switching dead zone constraint, a reactive power optimization model for the DC transmission system is constructed.
5. A reactive power optimization control device for a DC transmission system, characterized in that, include: The historical data acquisition module is used to acquire historical transmission data of the DC transmission system during a preset historical period, wherein the historical transmission data includes the historical operating status of the DC transmission system; The optimization model building module is used to build a reactive power optimization model for the DC transmission system based on the principles of minimizing network loss and voltage deviation. The control strategy solving module is used to solve the reactive power optimization model using the historical power transmission data, and to determine the historical optimal reactive power control strategy of the DC transmission system based on the solution results. The prediction model training module is used to train a preset temporal convolutional residual network using the historical operating status of the DC transmission system as training samples and the historical optimal reactive power control strategy as sample labels, so as to obtain a reactive power control strategy prediction model. The reactive power control strategy prediction model is used to predict the current reactive power control strategy of the DC transmission system based on the current operating state of the DC transmission system. The current reactive power control strategy is the number of capacitors switched on and off in the DC transmission system. The prediction model training module includes: Determine the number of network layers and neurons in the temporal convolutional residual network to be constructed; Multiple sets of datasets are formed by acquiring multiple historical operating states of the DC transmission system and the corresponding historical optimal reactive power control strategies. Multiple sets of neurons with different numbers are set according to the number of network layers, and multiple temporal convolutional residual networks are constructed using multiple sets of neurons with different numbers. For each temporal convolutional residual network, multiple datasets are used for training, and the policy deviation rate between the predicted reactive power control policy output by each temporal convolutional residual network and the corresponding historical optimal reactive power control policy is determined. The temporal convolutional residual network with the smallest policy bias rate is selected as the final trained temporal convolutional residual network; The historical operating state of the DC transmission system is used as a training sample and input into the temporal convolutional residual network to obtain the predicted reactive power control strategy output by the temporal convolutional residual network. The historical optimal reactive power control strategy is used as the sample label, and the network parameters in the temporal convolutional residual network are updated with the goal of the predicted reactive power control strategy approaching the historical optimal reactive power control strategy. When the preset training conditions are met, the trained temporal convolutional residual network is used as the reactive power control strategy prediction model.
6. A storage medium, characterized in that: The storage medium stores computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the reactive power optimization control method for a DC transmission system as described in any one of claims 1 to 4.
7. A computer device, characterized in that, include: One or more processors, and memory; The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of the reactive power optimization control method for a DC transmission system as described in any one of claims 1 to 4.