A molten salt energy storage frequency modulation prediction method and system with dynamic attenuation feature enhancement

Abstract

The application discloses a molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement, obtains an original frequency modulation instruction signal sequence; obtains a time domain feature weight value, a frequency domain feature weight value and a mutation detection weight value of the original frequency modulation instruction signal sequence; fuses the time domain feature weight value, the frequency domain feature weight value and the mutation detection weight value into a composite weight value; performs enhancement transformation on the original frequency modulation instruction signal sequence by using the composite weight value; performs prediction on the transformed frequency modulation instruction signal sequence by using a prediction network, and applies a prediction result to power regulation of a molten salt energy storage system; a plurality of signal characteristics are obtained by extracting and analyzing the original frequency modulation instruction signal sequence; the fusion weight is generated based on the plurality of signal characteristics; and the prediction accuracy is improved after the original frequency modulation instruction signal sequence is enhanced and transformed.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of power grid frequency modulation, and in particular to a molten salt energy storage frequency modulation prediction method and system with dynamic attenuation feature enhancement. BACKGROUND

[0002] Molten salt energy storage frequency modulation instruction prediction refers to establishing a prediction model through analysis of historical frequency modulation instructions and related data to predict future frequency modulation instruction requirements, thereby optimizing the control strategy of the hybrid energy storage system and improving its frequency modulation performance and efficiency. Its characteristics include the following points:

[0003] First, improve system performance. Fast response: can quickly adjust output power in a short time, quickly respond to changes in grid frequency, effectively balance the supply-demand difference in the power system, and maintain the stability of the grid frequency. Improve adjustment accuracy: can more accurately track the frequency modulation instruction, reduce adjustment error, and improve the power supply quality and stability of the grid. Enhance system stability: by reasonably allocating the power and energy of different energy storage devices, the overall risk of the system can be reduced, the reliability and stability of the system can be improved, and the risk of system failure due to the failure or performance degradation of a single energy storage device can be reduced. Second, extend the service life of the device. Optimize the charging and discharging strategy: based on the prediction results, a reasonable charging and discharging strategy can be developed to avoid excessive charging and discharging or frequent charging and discharging of the energy storage device, thereby extending its service life. Balance device usage: reasonably allocate the usage frequency and load of different energy storage devices to make the aging degree of each device relatively balanced, reducing the maintenance cost and replacement frequency of the device. Third, reduce costs. Improve economic efficiency: by improving the frequency modulation performance and efficiency, the hybrid energy storage system can generate more frequency modulation income, while reducing the cost of power loss and equipment damage due to unstable grid frequency. Optimize investment costs: based on the prediction results, the capacity and quantity of different energy storage devices can be reasonably configured to avoid excessive investment and resource waste, and improve the return on investment.

[0004] Traditional prediction methods often use VMD decomposition or CEEMD decomposition to decompose the original frequency modulation signal into a series of sub-sequences IMF1, IMF2, IMF3, …, IMF K Then put it into the neural network for prediction, and then superimpose the final prediction result to get the final result. However, the existing data decomposition techniques (VMD, CEEMD) have the following inherent defects: Real-time processing limitations: neither of them is suitable for real-time applications due to long computation time. Complex signal processing challenges: the signal decomposition effect may not be good for high-frequency noise, sudden fluctuations or highly nonlinear signals. Experience dependence: parameter adjustment depends on user experience, with limited automation. SUMMARY

[0005] Aiming at the deficiencies of the prior art, in order to solve this problem, the present application provides a molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement, which can greatly improve the prediction accuracy.

[0006] A molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement, characterized in that it comprises:

[0007] S101, acquiring an original frequency modulation instruction signal sequence;

[0008] S102, acquiring time domain feature weight values, frequency domain feature weight values and mutation detection weight values of the original frequency modulation instruction signal sequence;

[0009] S103, fusing the time domain feature weight values, the frequency domain feature weight values and the mutation detection weight values into composite weight values;

[0010] S104, using the composite weight values to perform enhancement transformation on the original frequency modulation instruction signal sequence;

[0011] S105, using the transformed frequency modulation instruction signal sequence to perform prediction by a prediction network, and applying the prediction result to power regulation of the molten salt energy storage system.

[0012] A molten salt energy storage frequency modulation prediction system with dynamic attenuation feature enhancement, characterized in that it comprises:

[0013] An instruction acquisition and decomposition module, which acquires an original frequency modulation instruction signal sequence;

[0014] A calculation module, which acquires time domain feature weight values, frequency domain feature weight values and mutation detection weight values of the original frequency modulation instruction signal sequence;

[0015] A fusion module, which fuses the time domain feature weight values, the frequency domain feature weight values and the mutation detection weight values into composite weight values;

[0016] A transformation module, which uses the composite weight values to perform enhancement transformation on the original frequency modulation instruction signal sequence;

[0017] A power regulation module, which uses the fused subsequence to perform prediction by a prediction network, and applies the prediction result to power regulation of the molten salt energy storage system.

[0018] The present application has the beneficial effect that when fluctuations occur in the frequency of the power grid of a power plant, it can respond in time by extracting the original frequency modulation instruction signal sequence to analyze and acquire multiple signal characteristics, generating fusion weights based on the multiple signal characteristics; performing enhancement transformation on the original frequency modulation instruction signal sequence to improve prediction accuracy, predicting the size of the frequency modulation instruction in advance for energy storage regulation, and improving response accuracy and frequency modulation benefits. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1A flowchart of a method step. DETAILED DESCRIPTION

[0020] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the scope of protection of the present application.

[0021] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0022] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application and the appended claims, unless otherwise clear from context, the singular forms "a", "an" and "the" are intended to include plural forms.

[0023] The terms "comprise" and "include" indicate the presence of described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0024] The term "and / or" means any combination of one or more of the associated listed terms and all possible combinations thereof, and includes these combinations.

[0025] The embodiments of the present disclosure provide a molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement, as shown in Figure 1 The method comprises the following steps:

[0026] S101, obtaining an original frequency modulation instruction signal sequence;

[0027] S102, obtaining time domain feature weight values, frequency domain feature weight values, and mutation detection weight values of the original frequency modulation instruction signal sequence;

[0028] S103, fusing the time domain feature weight values, the frequency domain feature weight values, and the mutation detection weight values into a composite weight value;

[0029] S104, performing enhancement transformation on the original frequency modulation instruction signal sequence using the composite weight value;

[0030] S105, predicting the transformed frequency modulation instruction signal sequence by a prediction network, and applying the prediction result to power regulation of a molten salt energy storage system.

[0031] The step S101 of obtaining the original frequency modulation instruction signal sequence comprises the following steps:

[0032] The frequency modulation instruction signal sequence is P t =[X1,X2,X3,…,X i ,…,X N ],(N>1000)。The original frequency modulation instruction signal sequence P t is a function of time t, and the subscript represents the sampling number, and the sampling interval is 1 second.

[0033] The time domain feature weight value, the frequency domain feature weight value, and the mutation detection weight value of the original frequency modulation instruction signal sequence are obtained in S102.

[0034] The dynamic attenuation feature of the original frequency modulation instruction signal sequence includes the time domain feature weight value, the frequency domain feature weight value, and the mutation detection weight value.

[0035] Step one, calculate the time domain feature weight value W t (i),

[0036] ,

[0037] , which represents the mean value of the past 30-second window, that is, the mean value of the first 30 signals, and is used to represent the local baseline level.

[0038] ,

[0039] , which represents the standard deviation of the past 30-second window, that is, the standard deviation of the first 30 signals, and is used to quantify the local fluctuation intensity.

[0040]

[0041] is a scaling factor for controlling the sensitivity of abnormal values, and the value is 2.3.

[0042] tanh() hyperbolic tangent function, output range (-1, 1), used to compress the normalized difference value to a nonlinear interval to avoid gradient explosion.

[0043] Step two, calculate the frequency domain feature weight value W f (i),

[0044] When i is greater than 60, the frequency domain feature weight value w

[0045] ,

[0046] w represents the fixed window length, w=60, that is, the fixed window length is 60 seconds, which represents the time span of the frequency spectrum analysis, which needs to cover the target frequency period.

[0047] is the original signal sequence set of the past 60 seconds, i.e., the sequence set of the past 60 signals.

[0048] FFT() is a fast Fourier transform, which converts a time-domain signal into a frequency-domain energy distribution.

[0049] k is a key frequency band selection parameter, taking a random value between [0.1, 0.5].

[0050] Step three, calculate the mutation detection weight value W c (i),

[0051] When i > 100,

[0052] The mutation detection weight value formula is as follows:

[0053] ,

[0054] is an adjustment factor, taking a value of 0.5.

[0055] MAD(i) represents the median of the signal difference in the past 100-second window, which is used as a robust noise estimator.

[0056] MAD(i) = median(X i -X i-1 ,X i-1 -X i-2 ,…,X i-99 -X i-100 ), median() is the median.

[0057] The above S103, the time-domain feature weight value, the frequency-domain feature weight value, and the mutation detection weight value are fused into a composite weight value W i ; including:

[0058] When i > 100,

[0059] The composite weight value fusion formula is as follows:

[0060] ,

[0061] , , 、 is a fusion coefficient;

[0062] is used to normalize and compress the composite weight value to the interval (0, 1).

[0063] .

[0064] When 60 < i < 100, the composite weight value fusion formula is as follows:

[0065] ,

[0066] When i < 60, the composite weight value fusion formula is as follows:

[0067] .

[0068] The above S104, using the composite weight value to enhance the transformation of the original frequency modulation command signal sequence; comprising:

[0069] The enhanced transformation formula is as follows:

[0070] ,

[0071] Y i is X i The corresponding enhanced transformation substitute value;

[0072] is the enhancement intensity coefficient, the value can be 0.3, used to control the amplification effect of the weight at the current moment, and enhance the key features.

[0073] is the time sequence correlation coefficient, the value can be 0.1, used to introduce the proportion of the previous time sequence point information, and capture short-term correlation.

[0074] The above S105, the transformed frequency modulation command signal sequence is predicted by the prediction network, and the prediction result is applied to the power regulation of the molten salt energy storage system; comprising:

[0075] According to the transformation method, the frequency modulation command signal sequence P t =[X1,X2,X3,…,X i ,…,X N ] becomes [Y1,Y2,Y3,…,Y i ,…,Y N ].

[0076] [Y1,Y2,Y3,…,Y i ,…,Y N ] is put into the GRU for prediction, and the final prediction result is obtained.

[0077] The prediction result is used for power regulation in the molten salt energy storage system. The power instruction is dynamically allocated by monitoring the frequency deviation of the power grid in real time, and the frequency fluctuation is quickly suppressed.

[0078] According to the method shown in the above Figure 1 , the embodiment of the application also provides a molten salt energy storage frequency modulation prediction system with dynamic attenuation feature enhancement, characterized by comprising:

[0079] An instruction acquisition decomposition module acquires an original frequency modulation instruction signal sequence;

[0080] A calculation module acquires a time domain feature weight value, a frequency domain feature weight value and a mutation detection weight value of the original frequency modulation instruction signal sequence;

[0081] A fusion module fuses the time domain feature weight value, the frequency domain feature weight value and the mutation detection weight value into a composite weight value;

[0082] A transformation module performs enhancement transformation on the original frequency modulation instruction signal sequence by using the composite weight value;

[0083] A power regulation module predicts the fused subsequence by a prediction network and applies a prediction result to power regulation of a molten salt energy storage system.

[0084] The traditional prediction method is not accurate enough because the decomposed subsequence has uncertainty, and the prediction result is not accurate enough when directly put into a neural network for prediction. The present application proposes a new method, which extracts an original frequency modulation instruction signal sequence for analysis to obtain multiple signal characteristics, generates a fusion weight based on the multiple signal characteristics, and performs prediction after enhancement transformation on the original frequency modulation instruction signal sequence to improve prediction accuracy. The method proposed in the present application can further improve prediction accuracy.

[0085] In order to further verify the advantages of the present application, the present application uses the method of the present application and the GRU prediction method to predict the frequency modulation sequence, and the results are shown in Tables 1 and 2.

[0086] The acquisition of the test frequency modulation instruction sequence is from a power plant in Inner Mongolia. The collection time of frequency modulation sequence 1 is from 4:00 to 18:00 on a certain day in November 2024, and a data point is collected every 1 second. The collection time of frequency modulation sequence 2 is from 0:00 to 22:00 on a certain day in February 2025, and a data point is collected every 1 second.

[0087] Table 1:

[0088]

[0089] Table 2: Four evaluation indexes

[0090]

[0091] N represents the sample capacity, and respectively represent the actual value and the predicted value at time n.

[0092] From the experimental results, it can be seen that the four evaluation indexes are reduced, which shows that the proposed model can well improve the prediction accuracy.

[0093] The embodiment of the present application provides a computer device, comprising a processor, a memory and a bus, the memory stores machine readable instructions executable by the processor, when the computer device runs, the processor and the memory communicate through the bus, and the machine readable instructions are executed by the processor to perform the steps provided by any embodiment of the present application.

[0094] The computer device provided by the embodiment of the present application comprises a processor, a memory and a bus. The memory is used for storing execution instructions, including an internal memory and an external memory. The internal memory is also called an internal storage, and is used for temporarily storing operation data in the processor and data exchanged with the external memory such as a hard disk. The processor exchanges data with the external memory through the internal memory. When the electronic device runs, the processor and the memory communicate through the bus, so that the processor executes the following instructions:

[0095] An original frequency modulation instruction signal sequence is acquired.

[0096] Time domain feature weight values, frequency domain feature weight values and mutation detection weight values of the original frequency modulation instruction signal sequence are acquired.

[0097] The time domain feature weight values, the frequency domain feature weight values and the mutation detection weight values are fused into composite weight values.

[0098] The original frequency modulation instruction signal sequence is enhanced and transformed by using the composite weight values.

[0099] The fused subsequence is predicted by a prediction network, and the prediction result is applied to power regulation of the molten salt energy storage system.

[0100] The embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is run by a processor to execute the steps provided by any embodiment of the present application. The storage medium can be a volatile or non-volatile computer readable storage medium.

[0101] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments of the present disclosure can be implemented by hardware, or can be implemented by means of software and necessary general hardware platform. Based on such understanding, the technical solutions of the embodiments of the present disclosure can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (which can be a read-only optical disc, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present disclosure.

[0102] Those skilled in the art can understand that the drawings are only schematic of a preferred embodiment, and the modules or flows in the drawings are not necessarily required for implementing the present disclosure.

[0103] Those skilled in the art can understand that the modules in the device in the embodiments can be distributed in the device in the embodiments according to the description of the embodiments, or can be changed to be located in one or more devices different from the embodiments. The modules in the above embodiments can be combined into one module, or can be further split into multiple sub-modules.

[0104] The sequence numbers of the embodiments of the present disclosure are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0105] Obviously, those skilled in the art can make various modifications and variations to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure belong to the scope of the claims of the present disclosure and the equivalent technologies thereof, the present disclosure also intends to include these modifications and variations.

[0106] Finally, it should be noted that: the above only explains the present application, and is not used to limit the present application, although the present application has been described in detail, for those skilled in the art, it can still modify the above-mentioned technical solutions, or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement, characterized in that, Comprising: S101, acquiring an original frequency modulation instruction signal sequence; S102, acquiring time domain feature weight values, frequency domain feature weight values, and mutation detection weight values of the original frequency modulation instruction signal sequence; S103, fusing the time domain feature weight values, the frequency domain feature weight values, and the mutation detection weight values into a composite weight value; S104, performing enhancement transformation on the original frequency modulation instruction signal sequence by using the composite weight value; S105, predicting the transformed frequency modulation instruction signal sequence by a prediction network, and applying a prediction result to power regulation of a molten salt energy storage system. Step S101, acquiring an original frequency modulation instruction signal sequence, comprising: The frequency modulation instruction signal sequence is P t = [X1, X2, X3, …, X i , …, X N ] The frequency modulation instruction signal sequence P t is a function of time t, and the subscript indicates the sampling number, and the sampling interval is 1 second; In step S102, comprising: Step one, calculate time domain feature weight value W t (i), , represents the mean of the past 30 seconds window, i.e. the mean of the 30 signals before the statistics, used to characterize the local baseline level; ， represents the standard deviation of the past 30 seconds window, i.e. the standard deviation of the previous 30 signals, to quantify the intensity of local fluctuations; ， a scaling factor for controlling the sensitivity to outliers, The tanh() hyperbolic tangent function has an output range of (-1, 1), is used to compress the normalized difference value to a non-linear interval, and avoids gradient explosion; Step two, calculating the frequency domain feature weight value W f (i), When i is greater than 60, the frequency domain feature weight value can be calculated, ， w represents a fixed window length, w = 60, that is, the fixed window length is 60 seconds, which represents the time span of the spectrum analysis, For the original signal sequence set of the previous 60 seconds, that is, the sequence set of the previous 60 signals, FFT() is a fast Fourier transform, k is a key frequency band selection parameter, and a random value between [0.1, 0.5] is taken; Step three, calculating mutation detection weight value W c (i), When i > 100, the mutation detection weight value can be calculated, The mutation detection weight value formula is as follows: ， modulator, MAD(i) represents the median of the signal difference in the past 100-second window, and is used as a robust noise estimator.

2. The molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement according to claim 1, characterized in that, In step S103, comprising: When i > 100, Composite weight values W i The fusion formula is as follows: ， , is a fusion coefficient; for normalizing the composite weight values to the interval (0,1), 。 3. The molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement according to claim 2, characterized in that, In step S103, further comprising: When 60 < i < 100, the composite weight value fusion formula is as follows: ， When i < 60, the composite weight value fusion formula is as follows: 。 4. The molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement according to claim 3, characterized in that, In step S104, comprising: The enhancement transformation formula is as follows: ， Y i is X i corresponding enhanced transform alternative value; To enhance the strength coefficient, for control of the current time weight amplification effect, enhance the key features; T is the time series correlation coefficient, used to introduce the proportion of previous time point information, capturing short-term correlation.

5. The molten salt energy storage frequency modulation prediction method with dynamic attenuation feature enhancement according to claim 4, characterized in that, In step S105, comprising: According to the conversion method, the frequency modulation command signal sequence P t is converted into [Y1, Y2, Y3,..., Y i ,..., Y N ] Put [Y1, Y2, Y3, …, Y i ,…,Y N ] into the GRU for prediction to get the final prediction result, The prediction result is used for power regulation in the molten salt energy storage system.

6. A prediction system of molten salt energy storage frequency modulation prediction method enhanced by the dynamic attenuation feature of claim 1, characterized in that, Comprising: An instruction acquisition decomposition module that acquires an original frequency modulation instruction signal sequence; A calculation module that acquires time domain feature weight values, frequency domain feature weight values, and mutation detection weight values of the original frequency modulation instruction signal sequence; A fusion module that fuses the time domain feature weight values, the frequency domain feature weight values, and the mutation detection weight values into a composite weight value; A transformation module that performs enhancement transformation on the original frequency modulation instruction signal sequence by using the composite weight value; A power regulation module that predicts the fused subsequence by a prediction network, and applies a prediction result to power regulation of a molten salt energy storage system.

Images