Frequency modulation capacity demand prediction method considering fluctuation mode clustering
By combining K-means clustering and bidirectional LSTM networks with random forest feature selection, the complexity and uncertainty of power system frequency regulation capacity demand forecasting are solved, enabling accurate positioning and efficient forecasting of power grid frequency regulation capacity demand.
Patent Information
- Application Number
- CN202511030569.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-25
- Publication Date
- 2025-12-02
AI Technical Summary
Existing methods for forecasting frequency regulation capacity demand in power systems rely on human experience or massive amounts of data, resulting in high computational complexity, low efficiency, and failure to consider the uncertainty of forecast results, making it difficult to cope with the random fluctuations of new energy sources and loads.
A clustering method that considers fluctuation patterns is adopted. Through K-means clustering, bidirectional long short-term memory neural network and random forest feature selection, combined with local neighborhood information and variational mode decomposition, frequency modulation capacity demand is predicted.
This has improved the speed and security of power grid frequency response, enhanced the accuracy and economy of frequency regulation capacity demand forecasting, and provided a more reliable basis for operation.
Smart Images
Figure CN121052425A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of frequency regulation capacity demand forecasting technology for power systems, and provides a frequency regulation capacity demand forecasting method that considers fluctuation pattern clustering. Background Technology
[0002] With the continuous construction and development of new power systems, the penetration rate of wind and solar power generation in the power grid is constantly increasing. The randomness of wind and solar power output poses a huge challenge to the frequency control of the power grid, and the demand for frequency regulation capacity in the power system is also gradually increasing. Frequency regulation capacity demand is the extra power reserved by the power grid to balance random fluctuations in source load. The ideal frequency regulation capacity demand is to enable the reserved extra power to cover the power deviation of source load fluctuations.
[0003] Currently, the main methods for predicting frequency regulation capacity demand in power systems are based on dispatching operation experience and data-driven methods. The dispatching operation experience-based method relies heavily on dispatching experience, directly setting the frequency regulation capacity demand as a percentage of the system load. While simple and effective, this method carries the economic costs of overestimating capacity and the safety risks of underestimating it. The data-driven method, on the other hand, determines the system's frequency regulation demand by analyzing the impact of historical renewable energy and load fluctuations on frequency regulation capacity. This method relies on massive amounts of training data, considering the random fluctuations in renewable energy and load to determine the demand. However, due to the complex and variable operation of power systems, mining and calculating massive amounts of data is computationally complex and inefficient. Furthermore, most existing frequency regulation capacity demand prediction methods are deterministic, failing to consider the impact of uncertainty in the prediction results. Summary of the Invention
[0004] This invention aims to at least solve one of the technical problems existing in related technologies. To this end, this invention provides a frequency regulation capacity demand forecasting method that considers fluctuation pattern clustering, enabling rapid frequency response of the power grid, improving system economy, and enhancing operational safety.
[0005] This invention provides a frequency modulation capacity demand forecasting method considering fluctuation pattern clustering, comprising: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
[0006] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S1 includes: S11: Collect historical operating data of the power grid in the area to be predicted, including source measurement data and load levels of wind power output, photovoltaic power output, thermal power unit output, load-side data of load, and frequency regulation capacity demand; S12: Construct a power grid historical operation data matrix based on the historical operation data; S13: Calculate the system net load data based on the aforementioned power grid historical operation data matrix. ; in, For time series data of load, For time series data of wind power, This is time series data for photovoltaics.
[0007] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S2 includes: S21: Normalize the system net load data to obtain the normalized net load value: in, For the first The normalized net load value at time 10:00. No. Net load value at time, The maximum value in the net load data. This is the minimum value in the net load data; S22: Calculate the local neighborhood similarity based on the normalized net load value; S23: Determine the neighborhood center based on the normalized net load value; S24: Classify the system net load data according to the similarity between the neighborhood center and the local neighborhood to obtain the frequency regulation capacity requirements under different fluctuation modes.
[0008] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S22 includes: S221: with Centered on, select a length of neighborhood subsequence Calculate slope characteristics , among which, the slope at time for: in, For sequence length, For the first The normalized net load value at time 10:00. The time interval between adjacent moments; by Each slope constitutes a slope characteristic; S222: Calculation Standard deviation ,calculate Standard deviation : in, The average value of the neighborhood subsequence. This represents the average value of the slope characteristic; S223: Calculate the adaptive threshold for net load profile : in, for the median; S224: Convert the slope feature into local morphological information feature based on the net load morphology adaptive threshold. : in, for The corresponding local morphological information feature values will The local morphological information feature values are arranged in order to obtain ; S225: Based on the first neighborhood subsequence Second Neighborhood Subsequence Calculate the distance considering local neighborhood morphological information. : in, This is an important parameter for local neighborhood morphological information. for and Euclidean distance, Distance of local morphological information of net load subsequence. for Local morphological information features, for Local morphological information features, for No. A normalized net load value for No. A normalized net load value for No. Local morphological information feature values, for No. A normalized net load value Sort the subsequences by ordinal number. It is an absolute value function.
[0009] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S23 includes calculating the cluster centers: in, The quantity in the class. It serves as the cluster center for its class.
[0010] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S3 includes the following steps: S31: Use variational mode decomposition to decompose the features under different modes into time series components to obtain intrinsic mode components and residual components; S32: Based on the intrinsic mode components and residual components, an optimization-based random forest feature selection method is used to obtain the optimal predicted feature subset.
[0011] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S32 includes the following steps: S321: Construct the feature set after variational mode decomposition This includes intrinsic mode components and residual components; S322: Using the feature set as input to the random forest and the frequency modulation capacity requirement as output, train the prediction model to obtain the information gain value of each feature. And the accuracy of current frequency regulation capacity demand forecast , As a basic threshold; S323: According to Sort the feature set by value from largest to smallest, and select the top... Each feature is added to the pre-selected feature set, which is initially an empty set. The middle composition ,in, For containing A feature set of features The number of features , The total number of features; S324: Use After retraining the random forest, test the corresponding prediction accuracy. ,when At that time, continue to Add The features constitute a set After retraining the random forest, test the corresponding prediction accuracy. ,like Then proceed to step S325; when Then stop going. Add features, using the current feature set as pre-selected features, and proceed to step S325. For containing A feature set of features; S325: According to The pre-selected features are removed in ascending order of value until the set of predicted features becomes empty. The prediction model is retrained for each feature removed to obtain the corresponding frequency modulation capacity demand prediction accuracy. The optimal subset of predicted features is selected based on the prediction accuracy and the number of features.
[0012] According to the frequency modulation capacity demand prediction method considering fluctuation pattern clustering provided by the present invention, step S4 includes the following steps: S41: The probabilistic prediction model of the bidirectional long short-term memory neural network is: in, FM capacity demand forecast No. quantiles, For the transfer function of a bidirectional LSTM network, For quantiles, This is a quantile regression function; The transfer function of the bidirectional LSTM network is as follows: in, For a moment The forward LSTM output, Forward LSTM activation function, The input layer forward propagation weight matrix, For a moment Variable characteristics, For forward propagation of its own weight matrix, For a moment The forward LSTM output, time The output of the backward LSTM For backward LSTM activation function, The input layer backpropagation weight matrix, The weight matrix is used for backpropagation. For a moment The output of the backward LSTM For a moment The predicted value of the frequency modulation capacity requirement of the final output gate. The activation function is a concatenation of forward and backward propagation. The weight matrix is used for forward propagation to the output layer. This is the weight matrix for backpropagation to the output layer; S42: Train a regression model for a bidirectional long short-term memory neural network using the aforementioned optimal subset of predicted features, and calculate the loss function. : in, For the first Frequency modulation capacity requirements at any given time For predicted values, These are the quantile regression coefficients. It is an absolute value function. When the loss function is minimized, a bidirectional long short-term memory neural network model is obtained; S43: Input the data to be predicted into a bidirectional long short-term memory neural network model to obtain the predicted frequency modulation capacity requirement.
[0013] The present invention also provides a frequency modulation capacity demand forecasting system that considers fluctuation pattern clustering, comprising: Data collection module: used to collect historical operating data of the power grid in the area to be predicted, and obtain system net load data; K-means clustering module: used to perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. Optimal Predictive Feature Subset Selection Module: Used to select the optimal predictive feature subset based on the frequency modulation capacity requirement using an optimized random forest feature selection method; Data prediction module: used to predict frequency modulation capacity demand using a bidirectional long short-term memory neural network probabilistic model based on the optimal prediction feature subset.
[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of a frequency modulation capacity demand forecasting method considering fluctuation pattern clustering as described above.
[0015] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: (1) In view of the complex and ever-changing operation mode of the power system and considering the significant differences in frequency regulation capacity demand under different fluctuation modes, this invention uses the K-means method that considers local neighborhood information to cluster the frequency regulation capacity scenario mode of the power system, so as to achieve accurate positioning of the frequency regulation capacity demand scenario.
[0016] (2) In view of the problem that the frequency regulation capacity demand of the system is affected by the random fluctuation of new energy and load, this invention proposes a similarity measurement method that considers the local neighborhood morphological information. The neighborhood slope of the subsequence is used to describe the fluctuation characteristics of the net load, and the feature expression of the net load morphological information is realized through adaptive threshold. Combined with Euclidean distance as a similarity measurement method for the division of frequency regulation capacity demand scenarios in K-means clustering, the net load fluctuation pattern can be mined.
[0017] (3) This invention uses a bidirectional long short-term memory neural network and a quantile regression model to build a frequency regulation capacity demand probability prediction model. Compared with the traditional LSTM, the bidirectional long short-term memory neural network can be trained not only from input to output, but also from output to input, which effectively improves the accuracy of frequency regulation capacity demand prediction; combined with the quantile regression model, the uncertainty of the frequency regulation capacity demand prediction results is quantified, providing a more reliable basis for power grid operation.
[0018] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in this invention 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 some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0020] Figure 1 This is a flowchart illustrating a frequency modulation capacity demand forecasting method that considers fluctuation pattern clustering provided by the present invention.
[0021] Figure 2This is a schematic diagram of a frequency modulation capacity demand prediction device that considers fluctuation pattern clustering provided by the present invention.
[0022] Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention.
[0023] Figure label: 101. Data collection module; 102. K-means clustering module; 103. Optimal prediction feature subset selection module; 104. Data prediction module; 810. Processor; 820. Communication interface; 830. Memory; 840. Communication bus. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but cannot be used to limit the scope of this invention.
[0025] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.
[0026] The following is combined with Figures 1 to 3 This invention is described.
[0027] Example like Figure 1 As shown, Figure 1 A flowchart of a frequency modulation capacity demand forecasting method that considers fluctuation pattern clustering includes: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
[0028] Specifically, step S1 includes: S11: Collect historical operating data of the power grid in the area to be predicted, including source measurement data and load levels of wind power output, photovoltaic power output, thermal power unit output, load-side data of tie line load, and frequency regulation capacity demand; S12: Construct a power grid historical operation data matrix based on the historical operation data; Assuming a power grid historical operation data matrix .
[0029] in, For wind power time series data, For photovoltaic time series data, For time series data of thermal power, For time series data of load, This is time-series data of tie line load.
[0030] S13: Calculate the system net load data based on the aforementioned power grid historical operation data matrix. ; in, For time series data of load, For time series data of wind power, This is time series data for photovoltaics.
[0031] Since the frequency regulation capacity demand of the system is closely related to the random fluctuations of new energy sources and loads, and considering the significant differences in frequency regulation capacity demand under different fluctuation modes, this invention determines the current system's frequency regulation demand mode based on the net load fluctuation characteristics. Traditional K-means methods typically use Euclidean distance to measure similarity, but Euclidean distance only focuses on straight-line distances, ignoring the morphological information of curves. This invention proposes a similarity measurement method that considers local neighborhood morphological information. It uses the neighborhood slope of subsequences to describe the fluctuation characteristics of net load, and uses an adaptive threshold to realize the feature expression of net load morphological information. Combined with Euclidean distance as a similarity measurement method for frequency regulation capacity demand scenario segmentation in K-means clustering, this enables the discovery of net load fluctuation patterns and further segmentation of frequency regulation capacity scenario modes.
[0032] Specifically, step S2 includes: S21: Normalize the system net load data to obtain the normalized net load value: in, For the first The normalized net load value at time 10:00. No. Net load value at time, The maximum value in the net load data. This is the minimum value in the net load data; S22: Calculate the local neighborhood similarity based on the normalized net load value; Step S22 includes: S221: with Centered on, select a length of neighborhood subsequence Calculate slope characteristics , among which, the slope at time for: in, For sequence length, For the first The normalized net load value at time 10:00. The time interval between adjacent moments; by Each slope constitutes a slope characteristic; S222: Calculation Standard deviation ,calculate Standard deviation : in, The average value of the neighborhood subsequence. This represents the average value of the slope characteristic; S223: Calculate the adaptive threshold for net load profile : in, for the median; S224: Convert the slope feature into local morphological information feature based on the net load morphology adaptive threshold. : in, for The corresponding local morphological information feature values will The local morphological information feature values are arranged in order to obtain ; S225: Based on the first neighborhood subsequence Second Neighborhood Subsequence Calculate the distance considering local neighborhood morphological information. : in, This is an important parameter for local neighborhood morphological information. for and Euclidean distance, Distance of local morphological information of net load subsequence. for Local morphological information features, for Local morphological information features, for No. A normalized net load value for No. A normalized net load value for No. Local morphological information feature values, for No. A normalized net load value Sort the subsequences by ordinal number. It is an absolute value function.
[0033] S23: Determine the neighborhood centers based on the normalized net load value, and calculate the cluster centers: in, The quantity in the class. The cluster center of its class; S24: Classify the system net load data according to the similarity between the neighborhood center and the local neighborhood to obtain the frequency regulation capacity requirements under different fluctuation modes.
[0034] Specifically, step S3 includes the following steps: S31: Use variational mode decomposition to decompose the time series of wind, solar and load under different modes to obtain intrinsic mode components and residual components; Considering the different temporal characteristics of the data, variational mode decomposition is used to decompose the input features into a time series, obtaining intrinsic mode components and residual components of different frequencies. For example: in, For the first The eigenmode components at each frequency, , The total number of frequencies is represented by the residual components, which are obtained by classifying the components that are not within the frequencies of the intrinsic mode components according to their frequencies.
[0035] S32: Based on the intrinsic mode components and residual components, an optimization-based random forest feature selection method is used to obtain the optimal predicted feature subset.
[0036] Step S32 includes the following steps: S321: Construct the feature set after variational mode decomposition This includes intrinsic mode components and residual components; S322: Using the feature set as input to the random forest and the frequency modulation capacity requirement as output, train the prediction model to obtain the information gain value of each feature. And the accuracy of current frequency regulation capacity demand forecast ,make As a basic threshold; in, The calculation formula is: In the formula, Let be the number of trees in the random forest. For the first One characteristic, For the first The first of the trees Information gain of each feature For characteristic ordinal numbers, The tree ordinal number, This is a function for calculating information gain.
[0037] S323: According to Sort the feature set by value from largest to smallest, and select the top... Each feature is added to the pre-selected feature set, which is initially an empty set. The middle composition ,in, For containing A feature set of features The number of features , The total number of features; S324: Use After retraining the random forest, test the corresponding prediction accuracy. ,when At that time, continue to Add The features constitute a set After retraining the random forest, test the corresponding prediction accuracy. ,like Then proceed to step S325; when Then stop going. Add features, using the current feature set as pre-selected features, and proceed to step S325. For containing A feature set of features; S325: According to The pre-selected features are removed in ascending order of value until the set of predicted features becomes empty. The prediction model is retrained for each feature removed to obtain the corresponding frequency modulation capacity demand prediction accuracy. The optimal subset of predicted features is selected based on the prediction accuracy and the number of features.
[0038] Specifically, step S4 includes the following steps: S41: The probabilistic prediction model of the bidirectional long short-term memory neural network is: in, FM capacity demand forecast No. quantiles, For the transfer function of a bidirectional LSTM network, For quantiles, This is a quantile regression function; The transfer function of the bidirectional LSTM network is as follows: in, For a moment The forward LSTM output, Forward LSTM activation function, The input layer forward propagation weight matrix, For a moment Variable characteristics, For forward propagation of its own weight matrix, For a moment The forward LSTM output, time The output of the backward LSTM For backward LSTM activation function, The input layer backpropagation weight matrix, The weight matrix is used for backpropagation. For a moment The output of the backward LSTM For a moment The predicted value of the frequency modulation capacity requirement of the final output gate. The activation function is a concatenation of forward and backward propagation. The weight matrix is used for forward propagation to the output layer. This is the weight matrix for backpropagation to the output layer; S42: Calculate the loss function : in, For the first Frequency modulation capacity requirements at any given time For predicted values, These are the quantile regression coefficients. It is an absolute value function. When the loss function is minimized, a bidirectional long short-term memory neural network model is obtained; S43: Input the data to be predicted into a bidirectional long short-term memory neural network model to obtain the predicted frequency modulation capacity requirement.
[0039] like Figure 2 As shown, the following describes a frequency modulation capacity demand prediction device considering fluctuation pattern clustering provided by the present invention. The frequency modulation capacity demand prediction device considering fluctuation pattern clustering described below can be referred to in correspondence with the frequency modulation capacity demand prediction method considering fluctuation pattern clustering described above. The device includes: Data collection module 101: used to collect historical operating data of the power grid in the area to be predicted, and obtain system net load data; K-means clustering module 102: used to perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency modulation capacity requirements under different fluctuation modes. Optimal Predictive Feature Subset Selection Module 103: Used to select the optimal predictive feature subset based on the frequency modulation capacity requirement using an optimized random forest feature selection method; Data prediction module 104: used to predict frequency modulation capacity demand using a bidirectional long short-term memory neural network probability model based on the optimal prediction feature subset.
[0040] Figure 3An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 3 As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute a frequency modulation capacity demand forecasting method considering fluctuation pattern clustering, the method including: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
[0041] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0042] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein when the program instructions are executed by a computer, the computer is able to execute a frequency modulation capacity demand forecasting method considering fluctuation pattern clustering provided by the above methods, the method comprising: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
[0043] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the frequency modulation capacity demand forecasting method considering fluctuation pattern clustering provided above, the method comprising: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
[0044] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0045] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
[0047] It should be noted that the embodiments of this disclosure can be implemented using hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a programmable memory or a data carrier such as an optical or electronic signal carrier.
[0048] Furthermore, although the operation of the methods of this disclosure is described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowcharts may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps. It should also be noted that the features and functions of two or more devices according to this disclosure may be embodied in one device. Conversely, the features and functions of one device described above may be further divided and embodied by multiple devices.
[0049] While this disclosure has been described with reference to several specific embodiments, it should be understood that this disclosure is not limited to the specific embodiments disclosed. This disclosure is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
Claims
1. A frequency modulation capacity demand forecasting method considering fluctuation pattern clustering, characterized in that, include: S1: Collect historical operating data of the power grid in the area to be predicted to obtain the system net load data; S2: Perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. S3: Based on the frequency modulation capacity requirement, use an optimization-based random forest feature selection method to select the optimal subset of predicted features; S4: Based on the optimal predicted feature subset, use a bidirectional long short-term memory neural network probabilistic model to predict frequency modulation capacity demand.
2. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 1, characterized in that, Step S1 includes: S11: Collect historical operating data of the power grid in the area to be predicted, including source measurement data and load levels of wind power output, photovoltaic power output, thermal power unit output, load-side data of tie line load, and frequency regulation capacity demand; S12: Construct a power grid historical operation data matrix based on the historical operation data; S13: Calculate the system net load data based on the aforementioned power grid historical operation data matrix. ; in, For time series data of load, For time series data of wind power, This is time series data for photovoltaics.
3. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 1, characterized in that, Step S2 includes: S21: Normalize the system net load data to obtain the normalized net load value: in, For the first The normalized net load value at time 10:
00. No. Net load value at time, The maximum value in the net load data. This is the minimum value in the net load data; S22: Calculate the local neighborhood similarity based on the normalized net load value; S23: Determine the neighborhood center based on the normalized net load value; S24: Classify the system net load data according to the similarity between the neighborhood center and the local neighborhood to obtain the frequency regulation capacity requirements under different fluctuation modes.
4. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 3, characterized in that, Step S22 includes: S221: with Centered on, select a length of neighborhood subsequence Calculate slope characteristics , among which, the slope at time for: in, For sequence length, For the first The normalized net load value at time 10:
00. The time interval between adjacent moments; by Each slope constitutes a slope characteristic; S222: Calculation Standard deviation ,calculate Standard deviation : in, The average value of the neighborhood subsequence. This represents the average value of the slope characteristic; S223: Calculate the adaptive threshold for net load profile : in, for the median; S224: Convert the slope feature into local morphological information feature based on the net load morphology adaptive threshold. : in, for The corresponding local morphological information feature values will The local morphological information feature values are arranged in order to obtain ; S225: Based on the first neighborhood subsequence Second Neighborhood Subsequence Calculate the distance considering local neighborhood morphological information. : in, This is an important parameter for understanding the morphological information of the local neighborhood. for and Euclidean distance, Distance of local morphological information of net load subsequence. for Local morphological information features, for Local morphological information features, for No. A normalized net load value for No. A normalized net load value for No. Local morphological information feature values, for No. A normalized net load value The ordinal number of the subsequence. It is an absolute value function.
5. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 4, characterized in that, Step S23 includes calculating cluster centers: in, The quantity in the class. It serves as the cluster center for its class.
6. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 1, characterized in that, Step S3 includes the following steps: S31: Use variational mode decomposition to decompose the time series of wind, solar and load under different modes to obtain intrinsic mode components and residual components; S32: Based on the intrinsic mode components and residual components, an optimization-based random forest feature selection method is used to obtain the optimal predicted feature subset.
7. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 6, characterized in that, Step S32 includes the following steps: S321: Construct the feature set after variational mode decomposition This includes intrinsic mode components and residual components; S322: Using the feature set as input to the random forest and the frequency modulation capacity requirement as output, train the prediction model to obtain the information gain value of each feature. And the accuracy of current frequency regulation capacity demand forecast , As a basic threshold; S323: According to Sort the feature set by value from largest to smallest, and select the top... Each feature is added to the pre-selected feature set, which is initially an empty set. The middle composition ,in, For containing A feature set of features The number of features , The total number of features; S324: Use After retraining the random forest, test the corresponding prediction accuracy. ,when At that time, continue to Add The features constitute a set After retraining the random forest, test the corresponding prediction accuracy. ,like Then proceed to step S325; when Then stop going. Add features, using the current feature set as pre-selected features, and proceed to step S325. For containing A feature set of features; S325: According to The pre-selected features are removed in ascending order of value until the set of predicted features becomes empty. The prediction model is retrained for each feature removed to obtain the corresponding frequency modulation capacity demand prediction accuracy. The optimal subset of predicted features is selected based on the prediction accuracy and the number of features.
8. The frequency modulation capacity demand forecasting method considering fluctuation pattern clustering according to claim 4, characterized in that, Step S4 includes the following steps: S41: The probabilistic prediction model of the bidirectional long short-term memory neural network is: in, The first value of the predicted frequency regulation capacity demand quantiles, For the transfer function of a bidirectional LSTM network, For quantiles, This is a quantile regression function; The transfer function of the bidirectional LSTM network is as follows: in, For a moment The forward LSTM output, Forward LSTM activation function, The input layer forward propagation weight matrix, For a moment Variable characteristics, For forward propagation of its own weight matrix, For a moment The forward LSTM output, time The output of the backward LSTM For backward LSTM activation function, The input layer backpropagation weight matrix, The weight matrix is used for backpropagation. For a moment The output of the backward LSTM For a moment The predicted value of the frequency modulation capacity requirement of the final output gate. The activation function is a concatenation of forward and backward propagation. The weight matrix is used for forward propagation to the output layer. This is the weight matrix for backpropagation to the output layer; S42: Train a regression model for a bidirectional long short-term memory neural network using the aforementioned optimal subset of predicted features, and calculate the loss function. : in, For the first Frequency modulation capacity requirements at any given time For predicted values, These are the quantile regression coefficients. It is an absolute value function. When the loss function is minimized, a bidirectional long short-term memory neural network model is obtained; S43: Input the data to be predicted into a bidirectional long short-term memory neural network model to obtain the predicted frequency modulation capacity requirement.
9. A frequency modulation capacity demand forecasting system considering fluctuation pattern clustering, for executing the frequency modulation capacity demand forecasting method considering fluctuation pattern clustering as described in any one of claims 1 to 8, characterized in that, include: Data collection module: used to collect historical operating data of the power grid in the area to be predicted, and obtain system net load data; K-means clustering module: used to perform K-means clustering on the net load data of the system, taking into account local neighborhood information, calculate the local neighborhood similarity and neighborhood center of the net load neighborhood, and obtain the frequency regulation capacity requirement under different fluctuation modes. Optimal Predictive Feature Subset Selection Module: Used to select the optimal predictive feature subset based on the frequency modulation capacity requirement using an optimized random forest feature selection method; Data prediction module: used to predict frequency modulation capacity demand using a bidirectional long short-term memory neural network probabilistic model based on the optimal prediction feature subset.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the frequency modulation capacity demand forecasting method that considers fluctuation pattern clustering as described in any one of claims 1 to 8.