Application of synchronous motor power generation in power grid system
By using real-time grid frequency and historical data to predict grid frequency, the problem of difficulty in predicting grid frequency changes when synchronous generators are connected to the grid is solved, and the grid stability and energy efficiency are improved.
Patent Information
- Application Number
- CN202411861771.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-17
- Publication Date
- 2025-03-21
AI Technical Summary
The existing synchronous generator grid connection method is difficult to effectively predict the frequency changes of the power grid, resulting in shock current or other instability during grid connection, affecting the stability of the power grid.
By collecting the current grid frequency in real time and recalling the historical frequency data of the upper and previous moments from the historical database, input it into the preset grid frequency prediction model, predict the grid frequency at the next moment, and formulate an adjustment plan based on the prediction results.
It realizes accurate prediction of the frequency change trend of the power grid, adjusts the working parameters of the synchronous generator in advance, ensures matching with the power grid frequency, reduces the risk of shock current, and improves grid stability and energy utilization efficiency.
Smart Images
Figure CN119675126A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of synchronous power generation, and in particular to an application of synchronous motor power generation in a power grid system. Background Art
[0002] Synchronous motors, also known as synchronous generators, are one of the important power generation equipment in power systems. They can efficiently convert mechanical energy into electrical energy and connect to the power grid to meet power demand. However, during the grid connection process of synchronous generators, it is necessary to ensure that their output frequency is strictly synchronized with the grid frequency, which places high demands on the stability of the system and the quality of power. Especially in modern power systems, with the large-scale access to renewable energy and the increase in load fluctuations, how to achieve safe and stable grid connection of synchronous generators has become particularly important.
[0003] The traditional method of connecting synchronous generators to the grid mainly relies on fixed control strategies, such as fixed excitation current settings and simple feedback control systems. Although this method can ensure the safe connection of synchronous generators to the grid to a certain extent, it lacks effective prediction of the frequency change trend of the power grid and is difficult to make timely adjustments in advance, resulting in surge current or other unstable phenomena at the moment of grid connection.
[0004] Based on this, there is an urgent need for an application of synchronous motor power generation in the power grid system, which can predict the grid frequency of the power grid system when the synchronous generator is connected to the grid, thereby realizing timely adjustment of the synchronous generator and improving the grid stability when the synchronous generator is connected to the grid. Summary of the invention
[0005] One of the purposes of the present invention is to provide an application of synchronous motor power generation in a power grid system, which can predict the grid frequency of the power grid system when the synchronous generator is connected to the grid, thereby realizing timely adjustment of the synchronous generator and improving the grid stability when the synchronous generator is connected to the grid.
[0006] In order to achieve the above object, an application of synchronous motor power generation in a power grid system is provided, comprising the following steps:
[0007] S1. Obtain the current grid frequency corresponding to the current moment in the grid system corresponding to the synchronous motor to be connected to the grid;
[0008] S2. Retrieve the historical power grid frequency corresponding to the last moment and the last moment of the corresponding power grid system from the historical database;
[0009] S3, taking the current power grid frequency corresponding to the current moment, and the historical power grid frequencies corresponding to the previous moment and the previous moment as input data, inputting them into a preset power grid frequency prediction model, and outputting the power grid prediction frequency corresponding to the next moment;
[0010] S4. According to the predicted frequency of the power grid corresponding to the next moment of the output, an adjustment plan corresponding to the synchronous motor to be connected to the grid at the next moment is formulated.
[0011] Technical principle and effect of this solution: In this solution, the grid frequency of the grid system where the synchronous motor to be connected to the grid is located is collected in real time from the grid monitoring system. This step ensures the latest understanding of the grid status and provides accurate basic data for subsequent predictions. Then, the grid frequency records of the same grid system at the past two time points (the previous moment and the moment before that) are retrieved from the historical database. The historical data provides trend information on frequency changes, which helps to improve the accuracy of the prediction model.
[0012] The grid frequencies at the current moment, the previous moment, and the moment before that are then used as input features and fed into a pre-trained grid frequency prediction model. The model outputs the predicted grid frequency at the next moment. This step uses the trends and patterns of historical data to predict future frequency changes.
[0013] Finally, according to the predicted grid frequency at the next moment, combined with the characteristics of the synchronous motor and the requirements of the grid, a detailed adjustment plan is formulated.
[0014] In this solution, by accurately predicting the changing trend of the grid frequency, the working parameters of the synchronous motor can be adjusted in advance to ensure that it fully matches the grid frequency at the moment of grid connection and reduce the risk of impact current. That is, the grid frequency of the power grid system can be predicted when the synchronous generator is connected to the grid, so as to achieve timely adjustment of the synchronous generator and improve the grid stability when the synchronous generator is connected to the grid.
[0015] More accurate frequency forecasts and timely adjustment measures can help maintain the stability of grid frequency, especially in the face of rapidly changing load demand or emergencies, enabling quick responses to prevent large frequency fluctuations.
[0016] The intelligent adjustment scheme can be dynamically optimized according to the prediction results, so that the synchronous motor always operates near its optimal efficiency point, thereby improving energy utilization efficiency and reducing operating costs.
[0017] Furthermore, the S3 further includes:
[0018] S23. Construct and train a grid frequency prediction model based on the historical grid frequencies corresponding to each moment in the historical database.
[0019] Beneficial effects: By using a large amount of historical grid frequency data for training, the model can better capture the changing patterns and trends of grid frequency, thereby providing more accurate prediction results.
[0020] Further, the S23 includes:
[0021] S230, retrieving the historical power grid frequency corresponding to each moment from the historical database;
[0022] S231, according to the retrieved historical power grid frequencies corresponding to each moment, the historical power grid frequencies corresponding to three adjacent moments are taken as one data group to generate multiple data groups;
[0023] S232, marking each data group, wherein the marking content is the historical power grid frequency corresponding to the data group at the next moment;
[0024] S233, dividing each data group with annotated content into a training set and a test set according to a preset ratio;
[0025] S234, constructing a power grid frequency prediction model;
[0026] S235. Randomly generate a number of model parameter groups according to the constructed power grid frequency prediction model;
[0027] S236, using each data group with annotated content in the training set as input data, inputting it into the power grid frequency prediction model corresponding to each model parameter group, and inputting the corresponding training prediction result;
[0028] S237, according to the training prediction results corresponding to each data group corresponding to each model parameter group, and the annotation content corresponding to each data group, based on the preset prediction accuracy calculation formula, the prediction accuracy of the power grid frequency prediction model corresponding to each model parameter group is calculated;
[0029] S237, selecting each model parameter group based on a preset parameter screening strategy according to the prediction accuracy corresponding to the power grid frequency prediction model corresponding to each model parameter group;
[0030] S238, according to each selected model parameter group, based on a preset model parameter replacement strategy, replace the model parameters of each selected model parameter group to form each new model parameter group, and re-execute S236 until the preset number of iterations is met;
[0031] S239. When the number of iterations meets the preset number of iterations, the model parameter group with the highest prediction accuracy corresponding to this time is output as the optimal model parameter group of the power grid frequency prediction model.
[0032] Beneficial effects: In this solution, historical grid frequency data is first retrieved from the historical database to ensure that there is enough historical data for model training, covering grid frequency changes in different time periods and operating conditions. Rich historical data improves the diversity and robustness of the model and can better capture the changing patterns of grid frequency.
[0033] Then the historical grid frequencies corresponding to three adjacent moments are taken as a data group to generate multiple data groups. By combining the data at adjacent moments, time series features are formed, which helps to capture the trend and pattern of frequency changes. This enhances the model's understanding of time dependence and improves prediction accuracy.
[0034] In order to make the model learn more effectively, a clear target value (i.e. the frequency at the next moment) is provided for each data group to facilitate supervised learning.
[0035] During the iteration process, through multiple rounds of iteration and optimization, the model can more accurately predict the grid frequency and reduce the impact current and other adverse effects at the moment of grid connection. The model performance is enhanced by calculating the prediction accuracy of the grid frequency prediction model corresponding to each model parameter group, providing a strong basis for subsequent screening.
[0036] Furthermore, the prediction accuracy calculation formula is:
[0037]
[0038] In the formula, MSE is the prediction accuracy, n is the total number of data groups corresponding to a certain model parameter group, and y i is the annotation content corresponding to the i-th data group, is the training prediction result corresponding to the i-th data group, and Y is the preset error threshold.
[0039] Beneficial effects: In this solution, the preset error threshold is set to ensure that the MSE is only affected when the prediction error exceeds a certain limit. This design highlights the importance of larger errors, making the model more focused on reducing significant deviations rather than minor errors. It helps to improve the performance of the model at critical moments and ensure that the model can provide reliable predictions in important situations (such as large fluctuations in grid frequency).
[0040] For errors below the preset threshold, the formula treats them as zero, that is, they are not included in the MSE calculation. This avoids the excessive impact of small errors on the overall evaluation results. It improves the robustness of model evaluation, prevents the model from being disturbed by unnecessary details, and focuses on capturing larger trends and patterns.
[0041] Furthermore, the preset parameter screening strategy is:
[0042] The prediction accuracy corresponding to each calculated model parameter group is sorted from large to small to form a prediction accuracy ranking table;
[0043] According to each prediction accuracy in the prediction accuracy arrangement table, the model parameter group corresponding to the prediction accuracy greater than the preset accuracy threshold is selected, and the other model parameter groups that have not been selected are placed in the spare library; the size of the preset accuracy threshold is related to the total number of prediction accuracy arrangement tables corresponding to this iteration process and the corresponding prediction accuracy average value;
[0044] In each iteration, the top three model parameter groups in terms of prediction accuracy in the reserve library are selected and integrated with the corresponding model parameter group selected in this iteration to form the final selected model parameter group set corresponding to this iteration.
[0045] Beneficial effects: In this solution, the preset accuracy threshold is adjusted dynamically to ensure that each iteration can select the model parameter group that best suits the current data set, thereby continuously improving the prediction accuracy of the model, and being able to flexibly respond to data characteristics and model performance at different stages, ensuring that the screening criteria are always reasonable and effective.
[0046] By taking the excellent model parameter groups in the backup library back into consideration, the diversity of model parameter selection is increased, preventing premature convergence to the local optimal solution and enhancing the robustness of the model.
[0047] The existence of the backup library not only saves model parameter groups that do not meet the standards but still have potential, but also provides additional options for subsequent iterations, improves resource utilization, and reduces unnecessary repeated training. Continuously introducing new and high-performing model parameter groups ensures that the model remains highly competitive throughout the optimization process and adapts to changing application needs.
[0048] Furthermore, the preset model parameter replacement strategy is:
[0049] Based on the selected model parameter groups, the model parameter groups are combined in groups of three to form corresponding model groups;
[0050] According to each model group, a parameter segment that needs to be replaced in the corresponding model group is randomly determined, and based on the determined parameter segment, model parameters corresponding to the corresponding parameter segment are selected from three model parameter groups in the model group;
[0051] The model parameters corresponding to the three model parameter groups in the model group are paired in pairs, and a random number is randomly generated from the paired model parameters. If the random number is greater than the preset random threshold, the paired model parameters are replaced, otherwise they are not replaced. After completing the replacement of each paired model parameter in the model group, the corresponding replaced model parameter group is output.
[0052] Beneficial effects: In this scheme, based on the selected model parameter groups, these model parameter groups are combined in groups of three to form multiple model groups. By combining different well-performing model parameter groups, the diversity of parameter selection is increased, which helps to explore a wider solution space. The combination between different model parameter groups promotes cross-learning between parameters, and may discover new and better parameter configurations. Randomly selecting parameter fragments and model parameters increases the possibility of exploring new solutions and avoids premature convergence to local optimal solutions. Maintaining diversity, by selecting parameters from multiple model parameter groups, ensures that the newly generated model parameter groups are diverse and innovative.
[0053] A random number is randomly generated for each pair of model parameters. If the random number is greater than the preset random threshold, the model parameters of the pair are replaced; otherwise, they are not replaced. Through the random number judgment mechanism, the proportion and degree of each replacement can be controlled, which not only ensures sufficient changes to explore new solutions, but also avoids excessive changes that lead to performance degradation. The preset random threshold helps to find a balance between exploration (trying new parameters) and utilization (retaining existing excellent parameters), ensuring that the model performance is steadily improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 The present invention is a flowchart of a synchronous motor for generating electricity in a power grid system according to a first embodiment of the present invention. DETAILED DESCRIPTION
[0055] The following is further described in detail through specific implementation methods:
[0056] Embodiment 1
[0057] A synchronous motor generator is used in a power grid system. Figure 1 As shown, the following steps are included:
[0058] S1. Obtain the current grid frequency corresponding to the current moment in the grid system corresponding to the synchronous motor to be connected to the grid;
[0059] S2. Retrieve the historical power grid frequency corresponding to the last moment and the last moment of the corresponding power grid system from the historical database;
[0060] S3, taking the current power grid frequency corresponding to the current moment, and the historical power grid frequencies corresponding to the previous moment and the previous moment as input data, inputting them into a preset power grid frequency prediction model, and outputting the power grid prediction frequency corresponding to the next moment;
[0061] The S3 further includes:
[0062] S23. Construct and train a grid frequency prediction model based on the historical grid frequencies corresponding to each moment in the historical database.
[0063] The S23 includes:
[0064] S230, retrieving the historical power grid frequency corresponding to each moment from the historical database;
[0065] S231, according to the retrieved historical power grid frequencies corresponding to each moment, the historical power grid frequencies corresponding to three adjacent moments are taken as one data group to generate multiple data groups;
[0066] S232, marking each data group, wherein the marking content is the historical power grid frequency corresponding to the data group at the next moment;
[0067] S233, dividing each data group with annotated content into a training set and a test set according to a preset ratio;
[0068] S234, constructing a power grid frequency prediction model; in this embodiment, the corresponding power grid frequency prediction model adopts BP neural network technology, specifically, first constructing a three-layer BP neural network model, including an input layer, a hidden layer and an output layer. In this embodiment, the historical power grid frequencies corresponding to three adjacent moments in history are used as inputs of the input layer, so the input layer has 3 nodes, and the output is the corresponding power grid frequency at the next moment. The hidden layer has a total of one node. For the hidden layer, this embodiment uses the following formula to determine the number of hidden layer nodes: Where L is the number of hidden layer nodes, A is the number of input layer nodes, B is the number of output layer nodes, C is a number between 1 and 10, and in this embodiment it is 4, and the hidden layer has a total of 6 nodes. BP neural network usually uses Sigmoid differentiable function and linear function as the network's excitation function. This paper selects the S-type tangent function tansig as the excitation function of the hidden layer neurons. The prediction model selects the S-type logarithmic function tansig as the excitation function of the output layer neurons.
[0069] S235. Randomly generate a number of model parameter groups according to the constructed power grid frequency prediction model;
[0070] S236, using each data group with annotated content in the training set as input data, inputting it into the power grid frequency prediction model corresponding to each model parameter group, and inputting the corresponding training prediction result;
[0071] S237, according to the training prediction results corresponding to each data group corresponding to each model parameter group, and the annotation content corresponding to each data group, based on the preset prediction accuracy calculation formula, the prediction accuracy of the power grid frequency prediction model corresponding to each model parameter group is calculated;
[0072] The prediction accuracy calculation formula is:
[0073]
[0074] In the formula, MSE is the prediction accuracy, n is the total number of data groups corresponding to a certain model parameter group, and y i is the annotation content corresponding to the i-th data group, is the training prediction result corresponding to the i-th data group, and Y is the preset error threshold.
[0075] S237, selecting each model parameter group based on a preset parameter screening strategy according to the prediction accuracy corresponding to the power grid frequency prediction model corresponding to each model parameter group;
[0076] The preset parameter screening strategy is:
[0077] The prediction accuracy corresponding to each calculated model parameter group is sorted from large to small to form a prediction accuracy ranking table;
[0078] According to each prediction accuracy in the prediction accuracy arrangement table, the model parameter group corresponding to the prediction accuracy greater than the preset accuracy threshold is selected, and the other model parameter groups that have not been selected are placed in the spare library; the size of the preset accuracy threshold is related to the total number in the prediction accuracy arrangement table corresponding to this iteration process and the corresponding prediction accuracy average value; in this embodiment, the size of the preset accuracy threshold is positively correlated with the total number of the prediction accuracy arrangement table corresponding to this iteration process, and is also positively correlated with the corresponding prediction accuracy average value.
[0079] In each iteration, the top three model parameter groups in terms of prediction accuracy in the reserve library are selected and integrated with the corresponding model parameter group selected in this iteration to form the final selected model parameter group set corresponding to this iteration.
[0080] S238, according to each selected model parameter group, based on a preset model parameter replacement strategy, replace the model parameters of each selected model parameter group to form each new model parameter group, and re-execute S236 until the preset number of iterations is met;
[0081] The preset model parameter replacement strategy is:
[0082] Based on the selected model parameter groups, the model parameter groups are combined in groups of three to form corresponding model groups;
[0083] According to each model group, a parameter segment that needs to be replaced in the corresponding model group is randomly determined, and based on the determined parameter segment, model parameters corresponding to the corresponding parameter segment are selected from three model parameter groups in the model group;
[0084] The model parameters corresponding to the three model parameter groups in the model group are paired in pairs, and a random number is randomly generated in the paired model parameters. If the random number is greater than the preset random threshold, the paired model parameters are replaced, otherwise they are not replaced. After completing the replacement of each paired model parameter in the model group, the corresponding replaced model parameter group is output. For example, model group 1 = (A, B, C), model group 2 = (A, D, E); for each model group, the parameter segment that needs to be replaced is randomly determined. Suppose we decide to change the first parameter, model group 1 needs to change the first parameter fragment: select the first parameter fragment from A, B, C: [a1, b1, c1], model group 2 needs to change the second parameter fragment: select the second parameter fragment from A, D, E: [a2, d2, e2]. When making update judgments between pairings, model group 1: pairing 1: (A, B) -> random number is 0.8> preset threshold 0.5-> replace a1 with b1; pairing 2: (A, C) -> random number is 0.3< preset threshold 0.5-> do not change; pairing 3: (B, C) -> random number is 0.6> preset threshold 0.5-> replace b1 with c1; result: A'=[b1, a2, a3], B'=[c1, b2, b3], C'=[c1, c2, c3].
[0085] S239. When the number of iterations meets the preset number of iterations, the model parameter group with the highest prediction accuracy corresponding to this time is output as the optimal model parameter group of the power grid frequency prediction model.
[0086] S4. According to the predicted frequency of the power grid corresponding to the next moment of the output, an adjustment plan corresponding to the synchronous motor to be connected to the grid at the next moment is formulated.
[0087] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is described too much here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field of the invention before the application date or priority date, can know all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.
Claims
1. An application of synchronous motor power generation in a power grid system, characterized in that: The following steps are involved: S1. Obtain the current grid frequency corresponding to the current moment in the grid system corresponding to the synchronous motor to be connected to the grid; S2. Retrieve the historical power grid frequency corresponding to the last moment and the last moment of the corresponding power grid system from the historical database; S3, taking the current power grid frequency corresponding to the current moment, and the historical power grid frequencies corresponding to the previous moment and the previous moment as input data, inputting them into a preset power grid frequency prediction model, and outputting the power grid prediction frequency corresponding to the next moment; S4. According to the predicted frequency of the power grid corresponding to the next moment of the output, an adjustment plan corresponding to the synchronous motor to be connected to the grid at the next moment is formulated.
2. The application of a synchronous motor power generation in a power grid system according to claim 1 is characterized in that: The S3 further includes: S23. Construct and train a grid frequency prediction model based on the historical grid frequencies corresponding to each moment in the historical database.
3. The application of a synchronous motor power generation in a power grid system according to claim 2 is characterized in that: The S23 includes: S230, retrieving the historical power grid frequency corresponding to each moment from the historical database; S231, according to the retrieved historical power grid frequencies corresponding to each moment, the historical power grid frequencies corresponding to three adjacent moments are taken as one data group to generate multiple data groups; S232, marking each data group, wherein the marking content is the historical power grid frequency corresponding to the data group at the next moment; S233, dividing each data group with annotated content into a training set and a test set according to a preset ratio; S234, constructing a power grid frequency prediction model; S235. Randomly generate a number of model parameter groups according to the constructed power grid frequency prediction model; S236, using each data group with annotated content in the training set as input data, inputting it into the power grid frequency prediction model corresponding to each model parameter group, and inputting the corresponding training prediction result; S237, according to the training prediction results corresponding to each data group corresponding to each model parameter group, and the annotation content corresponding to each data group, based on the preset prediction accuracy calculation formula, the prediction accuracy of the power grid frequency prediction model corresponding to each model parameter group is calculated; S237, selecting each model parameter group based on a preset parameter screening strategy according to the prediction accuracy corresponding to the power grid frequency prediction model corresponding to each model parameter group; S238, according to each selected model parameter group, based on a preset model parameter replacement strategy, replace the model parameters of each selected model parameter group to form each new model parameter group, and re-execute S236 until the preset number of iterations is met; S239. When the number of iterations meets the preset number of iterations, the model parameter group with the highest prediction accuracy corresponding to this time is output as the optimal model parameter group of the power grid frequency prediction model.
4. The application of a synchronous motor power generation in a power grid system according to claim 3 is characterized in that: The prediction accuracy calculation formula is: In the formula, MSE is the prediction accuracy, n is the total number of data groups corresponding to a certain model parameter group, and y i is the annotation content corresponding to the i-th data group, is the training prediction result corresponding to the i-th data group, and Y is the preset error threshold.
5. The application of a synchronous motor power generation in a power grid system according to claim 4 is characterized in that: The preset parameter screening strategy is: The prediction accuracy corresponding to each calculated model parameter group is sorted from large to small to form a prediction accuracy ranking table; According to each prediction accuracy in the prediction accuracy arrangement table, the model parameter group corresponding to the prediction accuracy greater than the preset accuracy threshold is selected, and the other model parameter groups that have not been selected are placed in the spare library; the size of the preset accuracy threshold is related to the total number of prediction accuracy arrangement tables corresponding to this iteration process and the corresponding prediction accuracy average value; In each iteration, the top three model parameter groups in terms of prediction accuracy in the reserve library are selected and integrated with the corresponding model parameter group selected in this iteration to form the final selected model parameter group set corresponding to this iteration.
6. The application of a synchronous motor power generation in a power grid system according to claim 5, characterized in that: The preset model parameter replacement strategy is: Based on the selected model parameter groups, the model parameter groups are combined in groups of three to form corresponding model groups; According to each model group, a parameter segment that needs to be replaced in the corresponding model group is randomly determined, and based on the determined parameter segment, model parameters corresponding to the corresponding parameter segment are selected from three model parameter groups in the model group; The model parameters corresponding to the three model parameter groups in the model group are paired in pairs, and a random number is randomly generated from the paired model parameters. If the random number is greater than the preset random threshold, the paired model parameters are replaced, otherwise they are not replaced. After completing the replacement of each paired model parameter in the model group, the corresponding replaced model parameter group is output.