Method and system for controlling rising rate of steam turbine through valve opening

Abstract

The invention belongs to the technical field of thermal power generation, and particularly relates to a method and system for controlling the rising rate of a steam turbine through the valve opening degree, and the method comprises the steps: training a preset frequency situation prediction model based on a power grid frequency scene through the current power grid frequency related parameters, and enabling the frequency situation prediction model to output a frequency prediction curve; the frequency prediction curve is used as a constraint condition, and a steam turbine set frequency response rising rate is determined by using a steam turbine rising rate aggregation model; and based on the whale algorithm and the turboset frequency response rising rate, the valve opening degree is obtained, and control is conducted based on the valve opening degree. According to the method, the frequency prediction curve is output through the frequency situation prediction model, so that the power grid frequency is accurately predicted, the frequency response rising rate of the turboset is determined, and the primary frequency modulation success rate of the turbine is guaranteed by accurately controlling the valve opening. The defect that an existing method cannot adapt to power grid frequency changes and cannot predict and adjust the valve opening is overcome, and the primary frequency modulation success rate is increased.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of thermal power generation, and particularly relates to a method and system for controlling the speed-up rate of a steam turbine through valve opening degree. BACKGROUND

[0002] With the increasing proportion of new energy access, the frequency fluctuation of the power grid is becoming more and more severe, which brings new challenges to primary frequency modulation and unit speed-up control. As an important part of conventional units, the speed-up capability of steam turbine units is directly related to the stability of system frequency and the quality of frequency modulation response. In the traditional control method, the speed-up process of the steam turbine is mainly realized by setting the valve opening degree curve, which often relies on artificial experience or static strategy, and is difficult to adapt to the dynamically changing power grid frequency environment, resulting in response lag, poor regulation accuracy, and even possible secondary disturbance.

[0003] The change of valve opening degree in the speed-up process of the steam turbine is a complex and delicate process, which is mainly controlled by DEH (digital electro-hydraulic control system). The DEH system is a decentralized control system with computer control technology as the core, which adjusts the steam flow by changing the opening degree of the speed regulating valve, so as to control the speed and load of the steam turbine. In operation, the DEH system will adjust the valve opening degree according to the deviation between the target speed and the actual speed, in order to achieve accurate control. SUMMARY

[0004] (I) Invention purpose

[0005] The purpose of the present application is to provide a method and system for controlling the speed-up rate of a steam turbine through valve opening degree, which can improve the frequency prediction accuracy, optimize the valve opening degree control, enhance the adaptability and accuracy of the steam turbine speed-up response, and improve the efficiency and stability of the power grid frequency regulation.

[0006] (II) Technical solution

[0007] To solve the above problems, the present application provides a method for controlling the speed-up rate of a steam turbine through valve opening degree, comprising:

[0008] Obtaining day-ahead power grid frequency related parameters, and clustering the day-ahead power grid frequency related parameters into multiple power grid frequency scenarios through a clustering algorithm;

[0009] Based on the power grid frequency scenario, a pre-set frequency trend prediction model is trained using the day-ahead power grid frequency related parameters, and the frequency trend prediction model outputs a frequency prediction curve;

[0010] A steam turbine speed-up aggregation model is constructed, the input of the steam turbine speed-up aggregation model is the day-ahead power grid frequency related parameters, and the output of the steam turbine speed-up aggregation model is the frequency response speed-up rate of the steam turbine unit;

[0011] Constraining the frequency prediction curve, using the turbine speed-up aggregation model, determine the frequency response speed-up rate of the turbine unit;

[0012] Based on the whale algorithm and the frequency response speed-up rate of the turbine unit, obtain the valve opening, control based on the valve opening.

[0013] Another aspect of the application, preferably, the acquisition of day-ahead grid frequency related parameters, and the day-ahead grid frequency related parameters are clustered into a plurality of grid frequency scenarios by clustering algorithm, comprising:

[0014] Obtain the day-ahead grid frequency related parameters, identify the grid frequency transformation coefficient in the day-ahead grid frequency related parameters;

[0015] The non-parametric kernel density estimation method is used to fit the grid frequency transformation coefficient, the corresponding valve operating parameter and the frequency modulation output parameter, and the probability estimation value of the grid frequency transformation coefficient, the valve operating parameter and the frequency modulation output parameter is obtained;

[0016] According to the probability estimation value and the K-means clustering algorithm, a plurality of clustering centers are generated, and the clustering centers represent the grid frequency scenarios;

[0017] Based on the cosine similarity method, the similarity between the probability estimation value and each clustering center is calculated, and according to the similarity, the day-ahead grid frequency related parameters corresponding to the probability estimation value are assigned to the nearest clustering center.

[0018] Another aspect of the application, preferably, the probability estimation value is calculated by the following formula:

[0019]

[0020] Where f(x) represents the probability estimation value of the grid frequency transformation coefficient, the valve operating parameter and the frequency modulation output parameter at point x, K(·) represents the kernel density function of the grid frequency discrete scene related frequency transformation output, h is the kernel function window width, X i is the input representation of the grid frequency transformation coefficient, the valve operating parameter and the frequency modulation output parameter.

[0021] Another aspect of the application, preferably, the similarity is calculated by the following formula:

[0022]

[0023] Where Sim(i,j) represents the similarity between the probability estimation value and the clustering center, a is the similarity parameter, sc i,j , represents the interaction degree and the average interaction degree of the clustering center associated with the probability estimation value, KPi , KP j respectively are probability estimation values, a set of clustering centers, and i1 is a power grid frequency transformation coefficient corresponding to the probability estimation value.

[0024] In another aspect of the present application, preferably, the preset frequency trend prediction model is based on a Bi-LSTM network model.

[0025] The frequency trend prediction model comprises an input gate, a forgetting gate, a memory cell and an output gate, an equivalent compensation strategy is introduced in the input gate, and the forgetting gate is improved in combination with an analytic hierarchy process algorithm.

[0026] In another aspect of the present application, preferably, the improved forgetting gate vector is represented as:

[0027]

[0028] The output gate output vector of the frequency trend prediction model is represented as:

[0029]

[0030] wherein H j,t represents the output gate output vector, h j,t,f , h j,t,b are forward and backward output vectors at the current moment, f j.t is a forgetting gate vector representation, sigma is an activation function of the forgetting gate, x t represents an input vector at the current moment, h t-1 represents a hidden layer vector at the previous moment, W f j , b f j respectively represent a weight matrix and a bias term of the forgetting gate, q j is a weight coefficient of the input vector at the current moment based on the analytic hierarchy process.

[0031] In another aspect of the present application, preferably, the steam turbine speed-up aggregation model takes the frequency trend prediction model as an initial model, and constructs an objective function with the minimum steam turbine unit frequency response speed-up rate as a target;

[0032] The valve operating parameters in the day-ahead power grid frequency correlation parameters are extracted, a hybrid enhancement algorithm is used to enhance the valve operating parameters, and the enhanced valve operating parameters are used to train the steam turbine speed-up aggregation model.

[0033] In the training process, the model weight parameters are iteratively updated by using a stochastic gradient descent optimizer with the objective function as an optimization target, so as to obtain optimized steam turbine speed-up aggregation model parameters.

[0034] In another aspect of the present application, preferably, the objective function is:

[0035]

[0036] wherein min(v W , P W ) represents the frequency response speed-up rate of the steam turbine unit, T is the frequency response period of the steam turbine unit, m is the number of steam turbine units, P W represents the output of the steam turbine unit at time t, η f is the opening of the valve in the steam turbine unit at time t, Δf represents the frequency fluctuation value, F max , and F min are the maximum and minimum frequencies of the power grid at time t, respectively.

[0037] In another aspect of the present application, preferably, the frequency response speed-up rate of the steam turbine unit is determined by using the steam turbine speed-up aggregation model with the frequency prediction curve as a constraint condition, comprising:

[0038] identifying the steam turbine unit frequency modulation output parameter in the frequency prediction curve;

[0039] determining the frequency response speed-up rate of the steam turbine unit by using the steam turbine speed-up aggregation model with the steam turbine unit frequency modulation output parameter as a constraint condition and considering the primary frequency modulation success rate.

[0040] In another aspect of the present application, preferably, a system for controlling the speed-up rate of a steam turbine by using the opening of a valve comprises:

[0041] an acquisition module: acquiring day-ahead power grid frequency related parameters and clustering the day-ahead power grid frequency related parameters into multiple power grid frequency scenarios by using a clustering algorithm;

[0042] a training module: training a preset frequency trend prediction model based on the power grid frequency scenarios and using the day-ahead power grid frequency related parameters, wherein the frequency trend prediction model outputs a frequency prediction curve;

[0043] a construction module: constructing a steam turbine speed-up aggregation model, wherein the input of the steam turbine speed-up aggregation model is the day-ahead power grid frequency related parameters, and the output of the steam turbine speed-up aggregation model is the frequency response speed-up rate of the steam turbine unit;

[0044] a determination module: determining the frequency response speed-up rate of the steam turbine unit by using the steam turbine speed-up aggregation model with the frequency prediction curve as a constraint condition;

[0045] a control module: obtaining the opening of the valve based on the whale optimization algorithm and the frequency response speed-up rate of the steam turbine unit and controlling based on the opening of the valve.

[0046] (III) Beneficial Effects

[0047] The above technical solutions of the present application have the following beneficial technical effects:

[0048] The present application outputs a frequency prediction curve through a frequency trend prediction model, thereby realizing accurate prediction of the power grid frequency, and further determining the frequency response speed of the steam turbine unit, so as to ensure the success rate of primary frequency modulation of the steam turbine by accurately controlling the valve opening degree. The present application overcomes the problem that the existing method cannot adapt to the change of the power grid frequency and predict and adjust the valve opening degree, and improves the success rate of primary frequency modulation. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 is the overall flowchart of an embodiment of the present application. DETAILED DESCRIPTION

[0050] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application with reference to the specific embodiments and the accompanying drawings. It should be understood that the description is only exemplary and is not intended to limit the scope of the present application. In addition, in the following description, the description of the known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present application.

[0051] Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0052] In addition, the technical features involved in different embodiments of the present application described below can be combined with each other as long as there is no conflict.

[0053] Embodiment One

[0054] A method for controlling the speed-up rate of a steam turbine through valve opening degree, Figure 1 The overall flowchart of an embodiment of the present application is shown in FIG. 1. Figure 1 As shown in FIG. 1, it includes:

[0055] The day-ahead power grid frequency related parameters are acquired, and the day-ahead power grid frequency related parameters are clustered into multiple power grid frequency scenarios through a clustering algorithm. The day-ahead power grid frequency related parameters include but are not limited to power grid system frequency, load change, primary frequency modulation performance parameters, standby capacity parameters, thermal power unit output parameters, climbing rate constraint parameters, planned output parameters, and peak demand balance parameters. In the present embodiment, the acquisition of the day-ahead power grid frequency related parameters and the clustering of the day-ahead power grid frequency related parameters into multiple power grid frequency scenarios through a clustering algorithm includes:

[0056] The day-ahead power grid frequency correlation parameter is acquired, and a power grid frequency transformation coefficient in the day-ahead power grid frequency correlation parameter is identified. The power grid frequency transformation coefficient is a parameter used to describe the relationship between load and power generation. When identifying the power grid frequency transformation coefficient in the day-ahead power grid frequency correlation parameter, the day-ahead power grid frequency correlation parameter can be sampled, and then a generalized second-order integrator is used to extract a positive sequence component. Next, equivalent and operation of trigonometric functions are performed in a two-phase rotating coordinate system, and finally the power grid frequency transformation coefficient is back calculated according to an operation period of a frequency calculation link.

[0057] The power grid frequency transformation coefficient, corresponding valve operating parameters and frequency modulation output parameters are fitted by using a non-parametric kernel density estimation method to obtain probability estimation values of the power grid frequency transformation coefficient, the valve operating parameters and the frequency modulation output parameters. The valve operating parameters represent indexes such as opening degree, response rate and execution delay of a steam turbine regulating valve; and the frequency modulation output parameters include active power variation amount and regulation sensitivity provided by a unit participating in primary frequency modulation or AGC automatic power generation control.

[0058] The non-parametric kernel density estimation method does not need to make a prior assumption on data distribution, and is suitable for complex sample data of non-normal distribution or multi-peak distribution. Joint probability models of the power grid frequency transformation coefficient, the valve operating parameters and the frequency modulation output parameters are respectively established by using the non-parametric kernel density estimation method to obtain probability estimation values of each group of parameters. The probability estimation values can reflect statistical correlation between the parameters and joint distribution characteristics of the parameters under a specific power grid operating state.

[0059] A plurality of clustering centers are generated according to the probability estimation values and a K-means clustering algorithm, and the clustering centers represent power grid frequency scenarios. An expected number of clusters K is set, and K clustering centers are randomly initialized in a parameter space. Each clustering center represents a typical power grid frequency operating scenario, such as rapid frequency drop, large amplitude fluctuation, slow rise, etc. Then, distances between each group of parameters and the clustering centers are calculated based on the obtained probability estimation values of each group of parameters.

[0060] Similarities of the probability estimation values and the clustering centers are calculated based on a cosine similarity method, and the day-ahead power grid frequency correlation parameters corresponding to the probability estimation values are assigned to the nearest clustering center according to the similarities. The cosine similarity method can effectively reflect similarity of different parameter combinations in a direction, and avoid a problem of invalidity of Euclidean distance in a high-dimensional space. By calculating the cosine similarities of each probability estimation value and the clustering centers, the parameter is divided into a category to which the clustering center with the highest similarity belongs, so that division of the power grid frequency scenarios is completed. Through multiple iterations, the K-means algorithm gradually adjusts positions of the clustering centers, so that samples in each category are more closely and categories are more separated, and finally stable clustering results are obtained.

[0061] In this embodiment, the probability estimate value is calculated using the following formula:

[0062]

[0063] wherein f(x) represents the probability estimate value of the grid frequency transformation coefficient, the valve operating parameter, and the frequency modulation output parameter at point x, K(·) represents the kernel density function of the grid frequency transformation output associated with the grid frequency discrete scene, h is the kernel function window width, X i represents the input of the grid frequency transformation coefficient, the valve operating parameter, and the frequency modulation output parameter. The probability estimate value represents the relative appearance probability of the three parameter combinations in the sample data distribution at a certain point, i.e., a certain set of grid frequency transformation coefficient, valve operating parameter, and frequency modulation output parameter. According to the above formula, the relative density of the parameter combination under a certain set of grid frequency disturbance in the historical data can be estimated, which provides probability feature support for subsequent clustering and classification.

[0064] The similarity is calculated by the following formula:

[0065]

[0066]

[0067] wherein Sim(i,j) represents the similarity between the probability estimate value and the clustering center, a is the similarity parameter, sc i,j , represents the interaction degree and the average interaction degree of the clustering center associated with the probability estimate value, KP i , KP j are the set of probability estimate values and clustering centers, respectively, and i1 is the grid frequency transformation coefficient corresponding to the probability estimate value. Both the consistency of the probability distribution characteristics and the geometric similarity in the parameter space are considered, which enhances the stability and robustness of the clustering distribution.

[0068] When the grid frequency associated parameters are clustered into at least one grid frequency discrete scene based on the fuzzy clustering algorithm, the non-parametric kernel density estimation method is used to fit the grid frequency transformation coefficient, the valve operating parameter, and the frequency modulation output parameter, and the similarity between the kernel density function of the frequency transformation output and the parameter clustering center is calculated based on the cosine similarity, so as to cluster the grid frequency associated parameters into at least one grid frequency discrete scene, thereby dividing a large number of complex grid frequency data into several subsets or categories with similar characteristics, which is helpful to identify different frequency change modes and trends, thereby providing a basis for constructing a more accurate frequency situation prediction model.

[0069] The frequency trend prediction model is trained based on the day-ahead power grid frequency related parameters according to a power grid frequency scenario, and the frequency trend prediction model outputs a frequency prediction curve; the preset frequency trend prediction model is based on a Bi-LSTM network model; the frequency trend prediction model comprises an input gate, a forgetting gate, a memory cell and an output gate, and an equivalent compensation strategy is introduced into the input gate, and the forgetting gate is improved in combination with an analytic hierarchy process.

[0070] The improved forgetting gate vector is represented as:

[0071]

[0072] The output vector of the output gate of the frequency trend prediction model is represented as:

[0073]

[0074] wherein H j,t represents the output vector of the output gate, h j,t,f , h j,t,b are forward and backward output vectors at a current moment, f j.t is a forgetting gate vector representation, σ is an activation function of the forgetting gate, x t represents an input vector at the current moment, h t-1 represents an implicit layer vector at a previous moment, W f j , b f j respectively represent a weight matrix and a bias term of the forgetting gate, and q j is a weight coefficient of the input vector at the current moment based on the analytic hierarchy process. In the embodiment, the frequency trend prediction model is based on the Bi-LSTM network model, introduces the equivalent compensation strategy, improves the forgetting gate in combination with the analytic hierarchy process, and the weight determined by the analytic hierarchy process can be adjusted according to different power grid operation scenarios, so that the improved LSTM network can better adapt to different power grid operation conditions and environmental changes, and the generalization ability and adaptability of the model are improved.

[0075] Further, in the embodiment, the preset frequency trend prediction model is trained based on the day-ahead power grid frequency related parameters according to the power grid frequency scenario, including:

[0076] A first training set is obtained, the preset frequency trend prediction model is iteratively trained based on a preset iteration number, an Adam optimizer is used to optimize initial model hyperparameters during the training, and a converged frequency trend prediction model is output; the first training set is a day-ahead power grid frequency related parameter based on a power grid frequency scenario;

[0077] Load the first test set, input the first test set, execute the converged frequency trend prediction model, output the test result, judge whether the test result meets the preset accuracy threshold based on the preset accuracy threshold, and if the accuracy threshold is met, output the converged frequency trend prediction model.

[0078] A turbine speed-up aggregation model is constructed, an input of the turbine speed-up aggregation model is a day-ahead power grid frequency related parameter, and an output of the turbine speed-up aggregation model is a turbine unit frequency response speed-up rate; in this embodiment, the turbine speed-up aggregation model takes the frequency trend prediction model as an initial model, and a target function is constructed with the minimum turbine unit frequency response speed-up rate as a target; further, the target function is:

[0079]

[0080] wherein, min(v W ,P W ) represents the turbine unit frequency response speed-up rate of the turbine speed-up aggregation model, T is a turbine unit frequency response period, m is the number of turbine units, P W represents the output of the turbine unit at the t period, η f is the opening of the valve in the turbine unit at the t period, Δf represents a frequency fluctuation value, F max , and F min are the maximum and minimum frequencies of the power grid at the t period, respectively.

[0081] In this embodiment, the turbine speed-up aggregation model is provided, the turbine speed-up aggregation model takes the frequency trend prediction model as a basic model, can continuously optimize the prediction result according to the real-time power grid frequency related parameter, and supports adaptive adjustment to cope with frequency changes under different conditions, and maintains high prediction accuracy.

[0082] The valve working parameters in the day-ahead power grid frequency related parameters are extracted, the hybrid enhancement algorithm is used to perform hybrid enhancement on the valve working parameters, and the turbine speed-up aggregation model is trained by using the enhanced valve working parameters;

[0083] In the training process, the target function is used as an optimization target, the model weight parameter is iteratively updated by using a stochastic gradient descent optimizer, and thus the optimized turbine speed-up aggregation model parameter is obtained. Further, in this embodiment, the training includes dividing the valve working parameters into a second training set, a second test set and a validation set; the frequency trend prediction model is iteratively trained by using the second training set, a frequency prediction evaluation loss function is calculated, and the frequency trend prediction model weight parameter is updated and optimized by using an SGD optimizer;

[0084] The validation set is obtained, and the accuracy of the frequency trend prediction model with the optimized weight parameter is verified by using the validation set;

[0085] determining whether the frequency prediction of the frequency trend prediction model is optimal; if the frequency prediction of the frequency trend prediction model is optimal, outputting the steam turbine speed-up aggregation model.

[0086] obtaining a second test set, calculating the response time and the response accuracy of the steam turbine speed-up aggregation model in predicting the frequency based on the second test set, and weighting and summing the response time and the response accuracy to obtain a response result;

[0087] determining whether the response result meets a result threshold; if the response result meets the result threshold, outputting the converged steam turbine speed-up aggregation model.

[0088] In this embodiment, the SGD optimizer is used to update and optimize the weight parameters of the frequency trend prediction model. The SGD uses a small batch of data in each iteration to update the model parameters, which can adapt to the changes in data faster, thereby improving the convergence speed of the steam turbine speed-up aggregation model and avoiding overfitting of the steam turbine speed-up aggregation model. Since the SGD uses only part of the data in each iteration, it can to some extent avoid overfitting of the model to the training data, especially on large-scale data sets.

[0089] determining the frequency response speed-up rate of the steam turbine unit using the steam turbine speed-up aggregation model with the frequency prediction curve as a constraint condition, comprising:

[0090] identifying the frequency modulation output parameters of the frequency prediction curve, identifying the frequency modulation response section associated with the steam turbine unit, extracting the frequency modulation output constraint parameters, and inputting the same as an optimization constraint condition to the steam turbine speed-up aggregation model.

[0091] determining the frequency response speed-up rate of the steam turbine unit using the steam turbine speed-up aggregation model based on the adaptive weight combined with the linear acceleration factor with the frequency modulation output parameters of the steam turbine unit as a constraint condition. The steam turbine speed-up aggregation model adopts an adaptive weight mechanism to dynamically adjust the weight of participating in frequency modulation according to the current state of each unit, such as available capacity, response rate, operating state, etc. At the same time, a linear acceleration factor is introduced to guide the speed-up rate optimization path to converge to the global optimal value. The linear acceleration factor takes into account the initial rapid response capability and the stability control in the later period, thereby improving the control ability of the model in the whole period of frequency disturbance.

[0092] Based on the whale algorithm and the frequency response speed-up rate of the steam turbine unit, the valve opening is obtained, and control is performed based on the valve opening. In combination with the whale algorithm, the optimal solution is searched in the frequency response speed-up rate solution space to obtain the optimal speed-up rate that meets the frequency response requirement. The whale algorithm performs global optimization on the model parameters through three behavior mechanisms of surrounding prey, spiral updating position and random search, can effectively jump out of the local optimum, and improves the adaptability of the model to complex frequency disturbance scenarios. Based on the frequency response speed-up rate obtained based on the optimization result, the corresponding steam turbine valve opening control instruction is calculated to form a valve regulation control strategy. The control strategy considers factors such as valve inertia response, dead zone effect and mechanical limit, and corrects the model prediction error in real time to realize fine adjustment of the valve opening, thereby ensuring the frequency regulation efficiency and stability of the entire steam turbine unit during the frequency disturbance response process.

[0093] The embodiment realizes accurate prediction of the power grid frequency by outputting a frequency prediction curve through the frequency situation prediction model, and then determines the frequency response speed-up rate of the steam turbine unit, so as to ensure the success rate of primary frequency regulation of the steam turbine by accurately controlling the valve opening. The existing method cannot adapt to changes in the power grid frequency and predict and adjust the valve opening, and the success rate of primary frequency regulation is improved.

[0094] Embodiment two

[0095] A system for controlling the speed-up rate of a steam turbine through valve opening, comprising:

[0096] An acquisition module: acquiring day-ahead power grid frequency related parameters, and clustering the day-ahead power grid frequency related parameters into multiple power grid frequency scenarios through a clustering algorithm;

[0097] A training module: training a preset frequency situation prediction model based on the day-ahead power grid frequency related parameters based on the power grid frequency scenarios, wherein the frequency situation prediction model outputs a frequency prediction curve;

[0098] A construction module: constructing a steam turbine speed-up aggregation model, wherein the input of the steam turbine speed-up aggregation model is the day-ahead power grid frequency related parameters, and the output of the steam turbine speed-up aggregation model is the frequency response speed-up rate of the steam turbine unit;

[0099] A determination module: determining the frequency response speed-up rate of the steam turbine unit by using the steam turbine speed-up aggregation model with the frequency prediction curve as a constraint condition;

[0100] A control module: obtaining a valve opening based on a whale algorithm and the frequency response speed-up rate of the steam turbine unit, and performing control based on the valve opening.

[0101] It is to be understood that the above specific embodiments of the present application are merely illustrative of the principles of the present application and are not intended to limit the scope of the present application. Any modification, equivalent substitution, improvement, etc. made without departing from the spirit and scope of the present application should be included in the scope of protection of the present application. In addition, the appended claims of the present application are intended to cover all variations and modifications falling within the scope and boundary of the appended claims or the equivalent forms of such scope and boundary.

[0102] The present application has been described above with reference to the embodiments of the present application. However, these embodiments are merely for illustrative purposes and are not intended to limit the scope of the present application. The scope of the present application is defined by the appended claims and equivalents thereof. Those skilled in the art can make various substitutions and modifications without departing from the scope of the present application, and such substitutions and modifications should fall within the scope of the present application.

[0103] Although the embodiments of the present application have been described in detail, it should be understood that various changes, substitutions and alterations can be made to the embodiments of the present application without departing from the spirit and scope of the present application.

[0104] Obviously, the above-described embodiments are merely for illustrative purposes and are not intended to limit the embodiments. Based on the above description, those skilled in the art can make other different forms of changes or modifications. It is not necessary or possible to exhaust all the embodiments. The obvious changes or modifications derived therefrom are still within the scope of protection of the present application.

Claims

1. A method for controlling the turbine acceleration rate by valve opening, characterized in that, include: Obtain the day-ahead power grid frequency-related parameters, and then use a clustering algorithm to cluster the day-ahead power grid frequency-related parameters into multiple power grid frequency scenarios; Based on the power grid frequency scenario, a preset frequency situation prediction model is trained using the day-ahead power grid frequency-related parameters, and the output of the frequency situation prediction model is a frequency prediction curve. A turbine speed-up convergence model is constructed. The input of the turbine speed-up convergence model is the day-ahead grid frequency-related parameters, and the output of the turbine speed-up convergence model is the turbine unit frequency response rate. Using the frequency prediction curve as a constraint, the turbine speed-up convergence model is used to determine the frequency response rate of the turbine unit. Based on the whale algorithm and the frequency response rate of the turbine unit, the valve opening is obtained, and control is performed based on the valve opening.

2. The method for controlling the turbine acceleration rate by valve opening according to claim 1, characterized in that, The step of obtaining day-ahead grid frequency-related parameters and clustering these parameters into multiple grid frequency scenarios using a clustering algorithm includes: Obtain the day-ahead power grid frequency correlation parameters and identify the power grid frequency transformation coefficients in the day-ahead power grid frequency correlation parameters; The nonparametric kernel density estimation method is used to fit the power grid frequency conversion coefficient, the corresponding valve operating parameters and frequency regulation output parameters to obtain the probability estimates of the power grid frequency conversion coefficient, valve operating parameters and frequency regulation output parameters. Several cluster centers are generated based on the probability estimate and the K-means clustering algorithm, and the cluster centers represent the power grid frequency scenario; The similarity between the probability estimate and each cluster center is calculated based on the cosine similarity method. Based on the similarity, the day-ahead power grid frequency-related parameters corresponding to the probability estimate are assigned to the nearest cluster center.

3. The method for controlling the turbine acceleration rate by valve opening according to claim 2, characterized in that, The probability estimate is calculated using the following formula: Where f(x) represents the probabilistic estimate of the power grid frequency transformation coefficient, valve operating parameters, and frequency regulation output parameters at point x, K(·) represents the kernel density function of the frequency transformation output associated with the discrete scenario of power grid frequency, h is the kernel function window width, and X i This input represents the power grid frequency conversion coefficient, valve operating parameters, and frequency regulation output parameters.

4. The method for controlling the turbine acceleration rate by valve opening according to claim 3, characterized in that, The similarity is calculated using the following formula: Where Sim(i,j) represents the similarity between the probability estimate and the cluster center, α is the similarity parameter, and sc i,j , KP represents the interaction degree and average interaction degree of cluster centers associated with the probability estimate. i KP j Let i1 be the set of probability estimates and cluster centers, respectively, and let i1 be the power grid frequency conversion coefficient corresponding to the probability estimates.

5. The method for controlling the turbine acceleration rate by valve opening according to claim 1, characterized in that, The preset frequency situation prediction model is based on a Bi-LSTM network model; The frequency situation prediction model includes an input gate, a forget gate, a memory cell, and an output gate. An equivalent compensation strategy is introduced into the input gate, and the forget gate is improved by combining the hierarchical analysis algorithm.

6. The method for controlling the turbine acceleration rate by valve opening according to claim 5, characterized in that, The improved forget gate vector is represented as follows: The output gate output vector of the frequency situation prediction model is represented as: Among them, H j,t h represents the output vector of the output gate. j,t,f ,h j,t,b These are the forward and backward output vectors at the current time step, f. j.t Let x be the vector representation of the forget gate, σ be the activation function of the forget gate, and x be the vector representation of the forget gate. t h represents the input vector at the current time step. t-1 W represents the hidden layer vector at the previous time step. f j b f j Let q represent the weight matrix and bias term of the forget gate, respectively. j The weighting coefficients of the input vector at the current time are based on the hierarchical analysis method.

7. The method for controlling the turbine acceleration rate by valve opening according to claim 1, characterized in that, The turbine speed-up convergence model uses the frequency situation prediction model as the initial model and constructs an objective function with the goal of minimizing the frequency response rate of the turbine unit. Valve operating parameters are extracted from the day-ahead power grid frequency correlation parameters. A hybrid enhancement algorithm is used to enhance the valve operating parameters. The enhanced valve operating parameters are then used to train the turbine speed-up aggregation model. During training, the objective function is used as the optimization target, and a stochastic gradient descent optimizer is used to iteratively update the model weight parameters to obtain the optimized turbine acceleration aggregation model parameters.

8. The method for controlling the turbine acceleration rate by valve opening according to claim 7, characterized in that, The objective function is: Wherein, min(v W ,P W ) represents the turbine unit frequency response rate of the turbine acceleration convergence model, T is the turbine unit frequency response period, m is the number of turbine units, and P W η represents the output of the steam turbine unit during time period t. f Let F represent the valve opening degree in the turbine unit during time period t, Δf represent the frequency fluctuation value, and F max F min These represent the maximum and minimum frequencies of the power grid during time period t, respectively.

9. The method for controlling the turbine acceleration rate by valve opening according to claim 1, characterized in that, Using the frequency prediction curve as a constraint, and employing the turbine speed-up convergence model, the frequency response rate of the turbine unit is determined, including: Identify the frequency prediction curve that includes the turbine unit frequency regulation output parameters; Using the frequency regulation output parameters of the steam turbine unit as constraints and considering the success rate of primary frequency regulation, the frequency response rate of the steam turbine unit is determined using the aforementioned steam turbine speed-up aggregation model.

10. A system for controlling the rate of increase of a steam turbine by valve opening, characterized in that, include: Acquisition module: Acquires day-ahead grid frequency-related parameters and clusters these parameters into multiple grid frequency scenarios using a clustering algorithm; Training module: Based on the power grid frequency scenario, using the day-ahead power grid frequency-related parameters, a preset frequency situation prediction model is trained, and the output of the frequency situation prediction model is a frequency prediction curve; Construction Module: Constructs a turbine speed-up convergence model. The input of the turbine speed-up convergence model is the day-ahead grid frequency-related parameters, and the output of the turbine speed-up convergence model is the turbine unit frequency response rate of increase. Determination module: Using the frequency prediction curve as a constraint, the turbine speed-up convergence model is used to determine the frequency response rate of the turbine unit; Control module: Based on the whale algorithm and the frequency response rate of the turbine unit, the valve opening is obtained, and control is performed based on the valve opening.

Images