Method and device for controlling frequency modulation of thermal power generating unit assisted by fused salt energy storage

Abstract

The embodiment of the invention relates to the technical field of thermal power generating unit frequency modulation, and provides a control method and device for assisting thermal power generating unit frequency modulation through fused salt energy storage. The electric heater inlet fused salt real-time temperature, the set electric heater outlet fused salt temperature and the set electric heater power are input into the trained fused salt flow neural network model, and the needed target fused salt flow is obtained; and according to the target fused salt flow, controlling the fused salt flow for assisting frequency modulation of the thermal power generating unit. According to the embodiment of the invention, fast and accurate control of the fused salt flow in the frequency modulation process of the auxiliary thermal power generating unit can be effectively realized, great temperature change and even over-temperature decomposition of the fused salt at the outlet of the electric heater caused by delayed fused salt flow control or slow control response are avoided, and the actual application requirements are met.

Description

Technical Field

[0001] The present disclosure relates to the technical field of frequency regulation of thermal power units, and in particular to a control method and device for frequency regulation of thermal power units assisted by molten salt energy storage. Background Art

[0002] In recent years, as an efficient way to store thermal energy, molten salt energy storage technology has attracted widespread attention in the field of renewable energy. Its core principle is to use molten salt as a thermal energy carrier to store and convert energy by absorbing and releasing heat. This technology was first used in solar thermal power generation systems to solve the problem of solar power generation being restricted by weather and time, and to ensure the stability and continuity of power supply.

[0003] The working process of molten salt energy storage mainly includes two stages: energy storage and heat release. In the heat storage stage, solar energy or other heat sources or electric heaters heat the molten salt to a high temperature state, and the heat energy in the molten salt is stored. In the heat release stage, when the heat energy is needed, the molten salt transfers the heat energy to the working medium such as water vapor through a heat exchanger to use the heat energy for power generation or industrial process heating.

[0004] Molten salt energy storage has obvious advantages. First, molten salt has high specific heat capacity and thermal stability, can store a large amount of heat energy at a higher temperature, and is not easy to decompose. Secondly, the molten salt energy storage system has high efficiency, which can reach more than 95%, and the energy loss is small. Thirdly, the molten salt energy storage system has low operating costs, simple maintenance, and a long service life of up to decades. Finally, molten salt energy storage can effectively smooth the intermittent output of renewable energy, improve the stability and reliability of the power grid, and is of great significance for building a new power system with new energy as the main body.

[0005] However, molten salt energy storage technology also faces some challenges, such as the selection and optimization of molten salt materials, the durability of high-temperature equipment, and the complexity of system integration and control technology. In the future, with the continuous advancement of materials science, thermal engineering and control system technology, molten salt energy storage is expected to be applied in a wider range of fields and become an important force in promoting energy transformation and sustainable development.

[0006] The typical way of applying the molten salt energy storage system to thermal power units is as follows: power is taken from the generator outlet or booster station of the generator unit, and the electrical energy is converted into thermal energy and stored through an electric heater. In the heat release stage, the heat in the molten salt is transferred to water vapor through a heat exchanger to supply steam to the outside or return the water vapor to the unit's thermal system. In the energy storage stage, the power delivered to the power grid by the generator unit can be changed by adjusting the heating power of the electric heater, thereby assisting the generator unit to participate in the power grid frequency regulation service to obtain considerable frequency regulation benefits.

[0007] The key equipment of the molten salt energy storage system in the energy storage stage is the molten salt electric heater that converts electrical energy into thermal energy. The molten salt flows into the electric heater from the inlet and flows out from the electric heater outlet after being heated. This requires that the molten salt temperature at the outlet of the electric heater needs to remain stable to achieve the maximum heat storage capacity, and the molten salt temperature should not exceed the decomposition temperature to achieve the safety and stability of the molten salt energy storage system.

[0008] The molten salt temperature at the outlet of the electric heater is mainly related to the molten salt temperature at the inlet of the electric heater, the power of the electric heater, and the molten salt flow rate. When the molten salt energy storage system participates in the frequency modulation of the thermal power unit, the automatic generation control (AGC) instruction of the thermal power unit is close to the rectangular wave change. In order to make the online power keep up with this change, the electric power of the molten salt energy storage system also needs to change steeply in a short period of time to improve the AGC response rate and accuracy of the unit. At this time, the existing technology can only ensure the stability of the outlet molten salt temperature by adjusting the rapid change of the molten salt flow rate. However, due to the large specific heat capacity of the molten salt and the slow flow rate, the existing control methods such as proportional integral derivative (PID) control have the problems of control lag and slow control response, which can easily cause a large change in the outlet molten salt temperature or even over-temperature decomposition. Summary of the invention

[0009] The present disclosure aims to solve at least one of the problems existing in the prior art and to provide a control method and device for frequency regulation of a thermal power unit assisted by molten salt energy storage.

[0010] In one aspect of the present disclosure, a control method for frequency regulation of a thermal power unit assisted by molten salt energy storage is provided, the control method comprising:

[0011] Get the real-time temperature of the molten salt at the inlet of the electric heater;

[0012] The real-time temperature of the molten salt at the inlet of the electric heater, the set molten salt temperature at the outlet of the electric heater and the electric heater power are input into the trained molten salt flow neural network model to obtain the required target molten salt flow;

[0013] According to the target molten salt flow rate, the molten salt flow rate for assisting the frequency regulation of the thermal power unit is controlled.

[0014] Optionally, the training process of the molten salt flow neural network model includes:

[0015] The molten salt temperature at the outlet of the electric heater, the molten salt temperature at the inlet of the electric heater, and the power of the electric heater are respectively used as input parameters, and the molten salt flow rate is used as an output parameter to establish the molten salt flow rate neural network model;

[0016] Acquire sample data corresponding to the input parameters and the output parameters under different steady-state operating conditions, and construct a training data set;

[0017] The molten salt flow neural network model is trained using the training data set.

[0018] Optionally, the molten salt flow neural network model includes a radial basis function network model; the radial basis function network model includes an input layer, a hidden layer and an output layer;

[0019] The output data of the hidden layer is expressed as:

[0020]

[0021] Wherein, j represents the node number of the hidden layer and j=1, 2, ..., q, q is the number of nodes in the hidden layer, It means that the jth hidden layer node has the input sample vector Output data when represents the Gaussian function center vector of the jth hidden layer node, represents the width vector of the j-th hidden layer node, T represents transpose, and exp represents the natural exponential function;

[0022] The output data of the output layer is expressed as:

[0023]

[0024] Where y represents the molten salt flow rate, w j Represents the weight coefficient of the jth hidden layer node.

[0025] Optionally, the using the training data set to train the molten salt flow neural network model includes:

[0026] According to the corresponding initial value, using the training data set and the corresponding iterative formula, the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node are iteratively calculated until the root mean square error output by the radial basis function network model is less than the preset accuracy.

[0027] Optionally, the initial value corresponding to the center vector of the Gaussian function is expressed as:

[0028]

[0029] Wherein, i represents the node number of the input layer and i=1,2,...,p, p is the number of nodes in the input layer, c ji represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node, x i.minrepresents the minimum value of the input sample corresponding to the i-th input layer node, x i.max Indicates the maximum value of the input sample corresponding to the i-th input layer node;

[0030] The initial value corresponding to the width vector is expressed as:

[0031] b ji =b f ||x i -c ji ||;

[0032] Among them, b ji represents the width vector of the jth hidden layer node based on the ith input layer node, b f represents the adjustment coefficient and 0 f <1,x i Represents the input sample corresponding to the i-th input layer node;

[0033] The initial value corresponding to the weighting coefficient is expressed as:

[0034]

[0035] Among them, y min Indicates the minimum value of the output sample, y max Indicates the maximum value of the output samples.

[0036] Optionally, the iterative formula corresponding to the weighted coefficient is expressed as:

[0037]

[0038] Where t represents the number of iterations, w j (t+1) represents the weight coefficient of the jth hidden layer node at the t+1th iteration, w j (t) represents the weight coefficient of the jth hidden layer node at the tth iteration, w j (t-1) represents the weight coefficient of the jth hidden layer node at the t-1th iteration, μ represents the learning factor, α represents the iteration coefficient and 0<α<1, and E represents the output deviation evaluation function and is expressed as n represents the sample number and n=1,2,3,...,N, N represents the total number of samples, y n.r Indicates the output sample value corresponding to the nth input sample, y n represents the output value of the radial basis function network model corresponding to the nth input sample;

[0039] The iterative formula corresponding to the center vector of the Gaussian function is expressed as:

[0040]

[0041] ​Among them, c ji (t+1) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the t+1th iteration, c ji (t) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the tth iteration, c ji (t-1) represents the Gaussian function center vector of the j-th hidden layer node based on the i-th input layer node at the t-1-th iteration;

[0042] The iterative formula corresponding to the width vector is expressed as:

[0043]

[0044] Among them, b ji (t+1) represents the width vector of the jth hidden layer node based on the ith input layer node at the t+1th time, b ji (t) represents the width vector of the t-th j-th hidden layer node based on the ith input layer node, b ji (t-1) represents the width vector of the j-th hidden layer node based on the i-th input layer node.

[0045] Optionally, the root mean square error output by the radial basis function network model is recorded as RMS, and the RMS is expressed as:

[0046]

[0047] Optionally, the using the training data set to train the molten salt flow neural network model further includes:

[0048] According to the corresponding initial value, the training data set and the corresponding iterative formula are used to iteratively calculate the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node, until the root mean square error output by the radial basis function network model is less than the preset accuracy, and the input samples in the training data set are normalized.

[0049] Optionally, controlling the molten salt flow rate for assisting the frequency modulation of the thermal power unit according to the target molten salt flow rate includes:

[0050] Using a molten salt flowmeter in conjunction with a variable frequency molten salt pump, the molten salt flow rate is adjusted to the target molten salt flow rate by adjusting the frequency of the variable frequency molten salt pump; or,

[0051] The molten salt flowmeter is used in conjunction with a regulating valve to adjust the molten salt flow rate to the target molten salt flow rate by adjusting the opening of the regulating valve.

[0052] Another aspect of the present disclosure provides a control device for assisting frequency regulation of a thermal power unit with molten salt energy storage, the control device comprising:

[0053] An acquisition module is used to obtain the real-time temperature of the molten salt at the inlet of the electric heater;

[0054] An input module is used to input the real-time molten salt temperature at the electric heater inlet and the set molten salt temperature and electric heater power at the electric heater outlet into the trained molten salt flow neural network model to obtain the required target molten salt flow;

[0055] The control module is used to control the molten salt flow rate of the auxiliary thermal power unit frequency modulation according to the target molten salt flow rate.

[0056] Compared with the prior art, the present invention obtains the required target molten salt flow rate based on the real-time temperature of the molten salt at the inlet of the electric heater and the set molten salt temperature and power at the outlet of the electric heater by using a trained molten salt flow neural network model. The molten salt flow rate for assisting the frequency modulation of the thermal power unit is controlled according to the target molten salt flow rate. This can effectively achieve rapid and accurate control of the molten salt flow rate during the frequency modulation of the auxiliary thermal power unit, avoid large changes in the molten salt temperature at the outlet of the electric heater or even over-temperature decomposition caused by molten salt flow control lag or slow control response, and meet actual application needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0058] Figure 1 A flow chart of a control method for frequency regulation of a thermal power unit assisted by molten salt energy storage provided in one embodiment of the present disclosure;

[0059] Figure 2 A schematic structural diagram of a molten salt energy storage system provided in another embodiment of the present disclosure;

[0060] Figure 3 A schematic diagram of a training process of a molten salt flow neural network model provided in another embodiment of the present disclosure;

[0061] Figure 4 A schematic diagram of a molten salt flow neural network model and a molten salt flow control process provided by another embodiment of the present disclosure;

[0062] Figure 5 A schematic diagram of the structure of a control device for assisting frequency regulation of a thermal power unit with molten salt energy storage provided in another embodiment of the present disclosure. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the embodiments of the present disclosure will be described in detail below in conjunction with the accompanying drawings. However, it can be understood by those skilled in the art that in each embodiment of the present disclosure, many technical details are proposed in order to enable readers to better understand the present disclosure. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed for protection in the present disclosure can also be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of the present disclosure. The various embodiments can be combined and referenced with each other without contradiction.

[0064] One embodiment of the present disclosure relates to a control method for frequency regulation of a thermal power unit assisted by molten salt energy storage, the process of which is as follows: Figure 1 As shown, including:

[0065] Step S110, obtaining the real-time temperature of the molten salt at the inlet of the electric heater.

[0066] Specifically, when the molten salt energy storage system operates normally, step S110 can directly read the real-time temperature of the molten salt of the electric heater in the molten salt energy storage system.

[0067] like Figure 2 As shown, the molten salt energy storage system generally includes a molten salt pipeline, a molten salt pump, a regulating valve, and an electric heater arranged in sequence on the molten salt pipeline. The molten salt pump can be combined with a frequency converter to form a variable frequency molten salt pump. By adjusting the frequency of the frequency converter, the molten salt flow in the molten salt pipeline can be adjusted by the molten salt pump. In addition, the opening of the regulating valve can be adjusted by an electric actuator to adjust the molten salt flow in the molten salt pipeline by adjusting the opening of the regulating valve.

[0068] Step S120, inputting the real-time temperature of the molten salt at the electric heater inlet and the set molten salt temperature and electric heater power at the electric heater outlet into the trained molten salt flow neural network model to obtain the required target molten salt flow.

[0069] Specifically, combined with Figure 2 When the molten salt energy storage system is operating normally, step S120 can obtain the electric heater inlet temperature T in real time from the molten salt energy storage system. in As the real-time temperature of the molten salt at the inlet of the electric heater. The molten salt temperature at the outlet of the electric heater is Figure 2 The outlet temperature of the electric heater is T in And the power of the electric heater can be set according to actual needs.

[0070] For example, Figure 3 As shown in the figure, the training process of the molten salt flow neural network model includes:

[0071] Step S310, taking the molten salt temperature at the outlet of the electric heater, the molten salt temperature at the inlet of the electric heater, and the power of the electric heater as input parameters, and taking the molten salt flow rate as an output parameter, to establish a molten salt flow rate neural network model.

[0072] The molten salt flow neural network model can be a multi-input single-output neural network model. Exemplarily, the molten salt flow neural network model includes a radial basis function (RBF) network model. Figure 4 , the RBF network model includes an input layer, a hidden layer, and an output layer. The input layer can include three nodes, each of which corresponds to an input parameter. For example, Figure 4 As shown, the input parameters corresponding to each node in the input layer are recorded as x1, x2, and x3, respectively, then x1, x2, and x3 can correspond to the molten salt temperature at the outlet of the electric heater, the molten salt temperature at the inlet of the electric heater, and the power of the electric heater, respectively.

[0073] The hidden layer is used to implement nonlinear mapping. Figure 4 , the output data of each node in the hidden layer can be u1, u2, u3, u4, ..., u q , where q is the number of nodes in the hidden layer. Specifically, the output data of the hidden layer is expressed as:

[0074]

[0075] Wherein, j represents the node number of the hidden layer and j=1,2,...,q. It means that the jth hidden layer node has the input sample vector Output data when . Represents the Gaussian function center vector of the j-th hidden layer node. represents the width vector of the jth hidden layer node. T represents transpose. exp represents the natural exponential function. It should be noted that the input sample vector Length, The length and The length of is the number of input parameters and also the number of nodes in the input layer. When the number of input parameters is 3, that is, the input layer includes 3 nodes, the input sample vector Length, The length and The length of is 3.

[0076] The output data of the output layer is expressed as:

[0077]

[0078] Where y represents the molten salt flow rate, w jRepresents the weight coefficient of the jth hidden layer node.

[0079] Using the output data corresponding to the hidden layer and the output layer, the mapping between the input parameters of the electric heater outlet molten salt temperature, the electric heater inlet molten salt temperature, and the electric heater power and the output parameter molten salt flow rate is established. Therefore, during the training stage, the RBF network model needs to be trained based on a large number of actual operation data samples including input parameters and output parameters to determine and w j The specific value of .

[0080] Step S320, obtaining sample data corresponding to input parameters and output parameters under different steady-state operating conditions, and constructing a training data set.

[0081] Specifically, since the data of the molten salt energy storage system under steady-state operating conditions can accurately and truly reflect the molten salt energy storage situation, the sample data for constructing the training data set needs to select the data set of the electric heater outlet molten salt temperature, electric heater inlet molten salt temperature, electric heater power and molten salt flow rate of the molten salt energy storage system under different steady-state operating conditions, and the data of the molten salt energy storage system under variable operating conditions should not be selected.

[0082] Step S330, using the training data set to train the molten salt flow neural network model.

[0083] Exemplarily, step S330 includes: according to the corresponding initial value, using the training data set and the corresponding iterative formula, iteratively calculating the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node, until the root mean square error output by the radial basis function network model is less than the preset accuracy.

[0084] Exemplarily, the initial value corresponding to the center vector of the Gaussian function is expressed as:

[0085]

[0086] Where i represents the node number of the input layer and i = 1, 2, ..., p, p is the number of nodes in the input layer, c ji represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node, x i.min represents the minimum value of the input sample corresponding to the i-th input layer node, x i.max Indicates the maximum value of the input sample corresponding to the i-th input layer node.

[0087] The initial value corresponding to the width vector is expressed as:

[0088] b ji =b f ||x i -cji ||.

[0089] Among them, b ji represents the width vector of the jth hidden layer node based on the ith input layer node, b f represents the adjustment coefficient and 0 f <1,x i Represents the input sample corresponding to the i-th input layer node.

[0090] The initial value corresponding to the weighting coefficient is expressed as:

[0091]

[0092] Among them, y min Indicates the minimum value of the output sample, y max Indicates the maximum value of the output samples.

[0093] According to the initial values ​​corresponding to the center vector, width vector, and weighting coefficient of the Gaussian function, during the model training phase, the output data of the output layer can be expressed as: in, Represents the vector corresponding to the nth input sample. n Represents the output value of the RBF network model corresponding to the nth input sample. The element is c ji . The element is b ji i = 1, 2, ..., p. p is the number of nodes in the input layer and p = 3. j = 1, 2, ..., q. q is the number of nodes in the hidden layer.

[0094] Specifically, the iterative formula corresponding to the weighted coefficient is expressed as:

[0095]

[0096] Where t represents the number of iterations. j (t+1) represents the weight coefficient of the jth hidden layer node at the t+1th iteration. j (t) represents the weight coefficient of the jth hidden layer node at the tth iteration. j (t-1) represents the weight coefficient of the jth hidden layer node at the t-1th iteration. μ represents the learning factor. α represents the iteration coefficient and 0<α<1.

[0097] The iterative formula corresponding to the center vector of the Gaussian function is expressed as:

[0098]

[0099] Among them, c ji ​(t+1) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the t+1th iteration, c ji (t) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the tth iteration, c ji (t-1) represents the Gaussian function center vector of the j-th hidden layer node based on the i-th input layer node at the t-1-th iteration.

[0100] The iterative formula corresponding to the width vector is expressed as:

[0101]

[0102] Among them, b ji (t+1) represents the width vector of the jth hidden layer node based on the ith input layer node at the t+1th time, b ji (t) represents the width vector of the t-th j-th hidden layer node based on the ith input layer node, b ji (t-1) represents the width vector of the j-th hidden layer node based on the i-th input layer node.

[0103] In the above iterative formulas, E represents the output deviation evaluation function and is expressed as:

[0104]

[0105] Where n represents the sample number and n = 1, 2, 3, ..., N. N represents the total number of samples. n.r Represents the output sample value corresponding to the nth input sample, that is, the expected output value corresponding to the nth input sample. n Represents the output value of the RBF network model corresponding to the nth input sample.

[0106] Exemplarily, the root mean square error output by the RBF network model is recorded as RMS, and the RMS is expressed as:

[0107]

[0108] Specifically, the preset accuracy is recorded as ε. When the RMS output by the RBF network model satisfies RMS<ε, the training is stopped. At this time, the RBF network model can meet the accuracy requirement as a molten salt flow neural network model.

[0109] After the RBF network model training is completed to obtain the trained molten salt flow neural network model, step S120 can use the real-time molten salt temperature at the electric heater inlet and the set molten salt temperature and electric heater power at the electric heater outlet as the input parameters of the molten salt flow neural network model, namely, the corresponding input values ​​of the molten salt temperature at the electric heater inlet, the molten salt temperature at the electric heater outlet, and the electric heater power, and the required target molten salt flow can be obtained through the trained molten salt flow neural network model.

[0110] In particular, step S330 also includes: according to the corresponding initial value, using the training data set and the corresponding iterative formula, iteratively calculating the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node, until the root mean square error output by the radial basis function network model is less than the preset accuracy, and the input samples in the training data set are normalized.

[0111] Specifically, since the input samples corresponding to the various input parameters usually correspond to different units and have obvious differences in magnitude, it is necessary to unify the format of the input samples by normalizing the input samples to eliminate differences in units and magnitudes.

[0112] Assume that the actual running value of the input sample corresponding to the i-th input layer node is x i.real , x i.max and x i.min x i.real The corresponding maximum and minimum actual operating values ​​are i.real After normalization, we can get x i.real The corresponding normalized value x i , x i It can be expressed as

[0113] Step S130, controlling the molten salt flow rate for assisting the frequency modulation of the thermal power unit according to the target molten salt flow rate.

[0114] Specifically, step S130 can be controlled in a variety of ways.

[0115] For example, step S130 may include: using a molten salt flow meter in conjunction with a variable frequency molten salt pump, and adjusting the frequency of the variable frequency molten salt pump to adjust the molten salt flow rate to a target molten salt flow rate. Figure 4 Step S130 can use a molten salt flow meter in conjunction with a variable frequency molten salt pump (i.e. Figure 4 The frequency converter in the variable frequency molten salt pump is used to adjust the molten salt flow rate. Specifically, a flow signal can be given by PID control. Based on the flow signal, the molten salt flow rate is adjusted to the required target molten salt flow rate by adjusting the frequency of the frequency converter in the variable frequency molten salt pump.

[0116] For example, step S130 may also include: using a molten salt flow meter in conjunction with a regulating valve to adjust the molten salt flow rate to a target molten salt flow rate by adjusting the opening of the regulating valve. Figure 4 In step S130, the molten salt flow rate can be adjusted by using a molten salt flow meter in conjunction with a regulating valve. Specifically, a flow signal can be given by a PID control method, and the parameter setting of the electric actuator can be adjusted based on the flow signal. The molten salt flow rate can be adjusted to the required target molten salt flow rate by adjusting the valve opening of the regulating valve using the set electric actuator.

[0117] Since the molten salt flow rate responds quickly to the valve opening of the variable frequency molten salt pump or the regulating valve, and the frequency of the variable frequency molten salt pump can be quickly adjusted through the adjustment rate of the corresponding frequency converter, and the valve opening of the regulating valve can be quickly adjusted through the parameter setting of the electric actuator, the molten salt flow rate can be quickly and accurately controlled through the conventional PID control method, thereby meeting the actual application needs.

[0118] Compared with the prior art, the control method for molten salt energy storage to assist the frequency modulation of thermal power units provided in the embodiments of the present invention obtains the required target molten salt flow rate based on the real-time temperature of the molten salt at the inlet of the electric heater and the set molten salt temperature and power of the electric heater at the outlet, and uses a trained molten salt flow neural network model to obtain the required target molten salt flow rate. The molten salt flow rate for assisting the frequency modulation of the thermal power units is controlled according to the target molten salt flow rate. This can effectively achieve rapid and accurate control of the molten salt flow rate during the frequency modulation of the auxiliary thermal power units, avoid large changes in the molten salt temperature at the outlet of the electric heater or even over-temperature decomposition caused by molten salt flow rate control lag or slow control response, and meet actual application needs.

[0119] Another embodiment of the present disclosure relates to a control device for assisting a thermal power unit in frequency regulation by molten salt energy storage, such as Figure 5 As shown, including:

[0120] The acquisition module 510 is used to obtain the real-time temperature of the molten salt at the inlet of the electric heater.

[0121] The input module 520 is used to input the real-time molten salt temperature at the electric heater inlet and the set molten salt temperature and electric heater power at the electric heater outlet into the trained molten salt flow neural network model to obtain the required target molten salt flow.

[0122] The control module 530 is used to control the molten salt flow rate for assisting the frequency regulation of the thermal power unit according to the target molten salt flow rate.

[0123] The specific implementation method of the control device for assisting the frequency regulation of a thermal power unit with molten salt energy storage provided in the embodiment of the present disclosure can be referred to the control method for assisting the frequency regulation of a thermal power unit with molten salt energy storage provided in the embodiment of the present disclosure, which will not be repeated here.

[0124] Compared with the prior art, the control device for assisting the frequency regulation of thermal power units with molten salt energy storage provided in the embodiment of the present invention obtains the required target molten salt flow rate based on the real-time temperature of the molten salt at the inlet of the electric heater and the set molten salt temperature and power of the electric heater at the outlet, and uses a trained molten salt flow neural network model to control the molten salt flow rate for assisting the frequency regulation of the thermal power units according to the target molten salt flow rate. This can effectively achieve rapid and accurate control of the molten salt flow rate during the frequency regulation of the auxiliary thermal power units, avoid large changes in the molten salt temperature at the outlet of the electric heater or even over-temperature decomposition caused by molten salt flow rate control lag or slow control response, and meet actual application needs.

[0125] Those skilled in the art will appreciate that the above-mentioned embodiments are specific embodiments for implementing the present disclosure, and in actual applications, various changes may be made thereto in form and detail without departing from the spirit and scope of the present disclosure.

Claims

1. A control method for frequency modulation of a thermal power unit assisted by molten salt energy storage, characterized in that: The control method comprises: Get the real-time temperature of the molten salt at the inlet of the electric heater; The real-time temperature of the molten salt at the inlet of the electric heater, the set molten salt temperature at the outlet of the electric heater and the electric heater power are input into the trained molten salt flow neural network model to obtain the required target molten salt flow; According to the target molten salt flow rate, the molten salt flow rate for assisting the frequency regulation of the thermal power unit is controlled.

2. The control method according to claim 1, characterized in that: The training process of the molten salt flow neural network model includes: The molten salt temperature at the outlet of the electric heater, the molten salt temperature at the inlet of the electric heater, and the power of the electric heater are respectively used as input parameters, and the molten salt flow rate is used as an output parameter to establish the molten salt flow rate neural network model; Acquire sample data corresponding to the input parameters and the output parameters under different steady-state operating conditions, and construct a training data set; The molten salt flow neural network model is trained using the training data set.

3. The control method according to claim 2, characterized in that: The molten salt flow neural network model includes a radial basis function network model; the radial basis function network model includes an input layer, a hidden layer and an output layer; The output data of the hidden layer is expressed as: Wherein, j represents the node number of the hidden layer and j=1, 2, ..., q, q is the number of nodes in the hidden layer, It means that the jth hidden layer node has the input sample vector Output data when represents the Gaussian function center vector of the jth hidden layer node, represents the width vector of the j-th hidden layer node, T represents transpose, and exp represents the natural exponential function; The output data of the output layer is expressed as: Where y represents the molten salt flow rate, w j Represents the weight coefficient of the jth hidden layer node.

4. The control method according to claim 3, characterized in that: The step of training the molten salt flow neural network model by using the training data set includes: According to the corresponding initial value, using the training data set and the corresponding iterative formula, the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node are iteratively calculated until the root mean square error output by the radial basis function network model is less than the preset accuracy.

5. The control method according to claim 4, characterized in that: The initial value corresponding to the center vector of the Gaussian function is expressed as: Wherein, i represents the node number of the input layer and i=1,2,...,p, p is the number of nodes in the input layer, c ji represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node, x i.min represents the minimum value of the input sample corresponding to the i-th input layer node, x i.max Indicates the maximum value of the input sample corresponding to the i-th input layer node; The initial value corresponding to the width vector is expressed as: b ji =b f ||x i -c ji ||; Among them, b ji represents the width vector of the jth hidden layer node based on the ith input layer node, b f represents the adjustment coefficient and 0 f <1,x i Represents the input sample corresponding to the i-th input layer node;​ The initial value corresponding to the weighting coefficient is expressed as: Among them, y min Indicates the minimum value of the output sample, y max Indicates the maximum value of the output samples.

6. The control method according to claim 5, characterized in that: The iterative formula corresponding to the weighted coefficient is expressed as: Among them, t represents the number of iterations, w j (t+1) represents the weight coefficient of the jth hidden layer node at the t+1th iteration, w j (t) represents the weight coefficient of the jth hidden layer node at the tth iteration, w j (t-1) represents the weight coefficient of the jth hidden layer node at the t-1th iteration, μ represents the learning factor, α represents the iteration coefficient and 0<α<1, and E represents the output deviation evaluation function and is expressed as n represents the sample number and n=1,2,3,...,N, N represents the total number of samples, y n.r Indicates the output sample value corresponding to the nth input sample, y n represents the output value of the radial basis function network model corresponding to the nth input sample; The iterative formula corresponding to the center vector of the Gaussian function is expressed as: Among them, c ji (t+1) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the t+1th iteration, c ji (t) represents the Gaussian function center vector of the jth hidden layer node based on the ith input layer node at the tth iteration, c ji (t-1) represents the Gaussian function center vector of the j-th hidden layer node based on the i-th input layer node at the t-1-th iteration; The iterative formula corresponding to the width vector is expressed as: Among them, b ji (t+1) represents the width vector of the jth hidden layer node based on the ith input layer node at the t+1th time, b ji (t) represents the width vector of the t-th j-th hidden layer node based on the ith input layer node, b ji (t-1) represents the width vector of the j-th hidden layer node based on the i-th input layer node.

7. The control method according to claim 6, characterized in that: The root mean square error output by the radial basis function network model is recorded as RMS, and the RMS is expressed as:

8. The control method according to claim 4, characterized in that: The step of training the molten salt flow neural network model using the training data set further includes: According to the corresponding initial value, the training data set and the corresponding iterative formula are used to iteratively calculate the Gaussian function center vector, width vector and weighting coefficient of each hidden layer node, until the root mean square error output by the radial basis function network model is less than the preset accuracy, and the input samples in the training data set are normalized.

9. The control method according to any one of claims 1 to 8, characterized in that: The controlling of the molten salt flow rate for assisting the frequency modulation of the thermal power unit according to the target molten salt flow rate comprises: Using a molten salt flowmeter in conjunction with a variable frequency molten salt pump, the molten salt flow rate is adjusted to the target molten salt flow rate by adjusting the frequency of the variable frequency molten salt pump; or, The molten salt flowmeter is used in conjunction with a regulating valve to adjust the molten salt flow rate to the target molten salt flow rate by adjusting the opening of the regulating valve.

10. A control device for frequency modulation of a thermal power unit assisted by molten salt energy storage, characterized in that: The control device comprises: An acquisition module is used to obtain the real-time temperature of the molten salt at the inlet of the electric heater; An input module is used to input the real-time molten salt temperature at the electric heater inlet and the set molten salt temperature and electric heater power at the electric heater outlet into the trained molten salt flow neural network model to obtain the required target molten salt flow; The control module is used to control the molten salt flow rate for assisting the frequency modulation of the thermal power unit according to the target molten salt flow rate.

Images