Super-capacity energy storage frequency modulation instruction prediction method, system, medium and equipment

Abstract

The invention discloses a super-capacity energy storage frequency modulation instruction prediction method and system, a medium and equipment, and belongs to the technical field of power grid frequency modulation, and the method comprises the steps: obtaining a historical frequency modulation instruction sequence and a current frequency modulation instruction; an error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence, and an error corresponding to the current frequency modulation instruction is a current error; based on the current error, the iteration weight of the prediction model is adjusted according to the historical error sequence until the current error is smaller than a preset value, and the prediction model after weight adjustment and a new current error are obtained; performing error trend analysis on the new current error and the historical error sequence to obtain an error acceleration; setting a fine tuning coefficient for the frequency modulation instruction sequence according to the magnitude of the error acceleration; inputting the frequency modulation instruction sequence with the fine tuning coefficient into the prediction model after weight adjustment to obtain a predicted frequency modulation instruction at a future moment; the prediction precision of the frequency modulation instruction is effectively improved, and the operation efficiency and stability of the energy storage system are guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid frequency regulation, and relates to a method, system, medium and equipment for predicting frequency regulation instructions of ultra-capacity energy storage. Background Art

[0002] With the transformation of the global energy structure and the continuous development of the power market, the power system has an increasing demand for frequency regulation technology. The traditional power system frequency regulation task is mainly undertaken by thermal power units and other generators, but this method has problems such as slow response speed and low control accuracy. Especially in the modern power system with large-scale access to new energy and frequent random load fluctuations, these problems are becoming more and more prominent.

[0003] In order to solve these problems, energy storage technology, as an effective means of frequency regulation, has gradually attracted widespread attention in the industry. Among them, lithium battery energy storage technology has been widely used in power systems due to its advantages such as high energy density and long cycle life. However, lithium battery energy storage technology also faces challenges such as poor safety and relatively slow response speed, especially under high-frequency and large-fluctuation frequency regulation requirements, its performance may be affected.

[0004] In order to solve the problems of lithium battery energy storage technology in the field of frequency modulation, researchers proposed a frequency modulation command prediction method for supercapacitor energy storage. This method realizes a new type of energy storage system by combining the high power density and fast response characteristics of supercapacitors. This energy storage system can not only provide sufficient energy support, but also respond quickly to frequency modulation commands, improving the frequency modulation performance of the power system.

[0005] In terms of frequency modulation command prediction, traditional prediction methods are often based on historical data and empirical formulas. This method may have a certain degree of accuracy when the power system operating conditions are relatively stable. However, in modern power systems, due to the access of new energy sources and random fluctuations in loads, the operating conditions of the power system have become complex and changeable, and traditional prediction methods are difficult to meet the needs of accurate prediction of frequency modulation commands. Summary of the invention

[0006] The present invention provides a method, system, medium and device for predicting frequency modulation instructions of ultra-capacity energy storage to solve the problem that the operating conditions of modern power systems become complex and changeable, and the traditional prediction methods are not accurate enough to meet the modern power system's demand for high-precision frequency modulation instruction prediction.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions: On the one hand, the present invention discloses a method for predicting frequency modulation instructions of super-capacity energy storage, comprising the following steps: S1, obtain a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a current frequency modulation instruction and a historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into a prediction model to obtain a predicted current frequency modulation instruction; the current error is obtained based on the predicted current frequency modulation instruction; the prediction model of the present invention can select an LSTM (Long Short-Term Memory) model or a BP (Back Propagation) model.

[0008] S2, according to the size of the current error and based on the historical error sequence, adjust the iterative weight of the prediction model until the current error is less than the preset value, and obtain the prediction model after iterative weight adjustment and the new current error; S3, based on the new current error and historical error sequence, error trend analysis is performed to obtain error acceleration; S4, according to the size of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the historical frequency modulation instruction sequence at the next moment with the set fine-tuning coefficient is input into the prediction model after iterative weight adjustment to obtain the predicted frequency modulation instruction at the next moment; the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment.

[0009] Furthermore, the historical frequency modulation instruction sequence is the frequency modulation instructions at several moments before the current moment.

[0010] Further, the weight of the prediction model is adjusted according to the size of the current error and based on the historical error sequence until the current error is less than a preset value, thereby obtaining the prediction model after iterative weight adjustment and the new current error. The specific method is as follows: Determine whether the current error is less than the preset value; when the current error is greater than or equal to the preset value, adjust the iteration weight of the prediction model based on the current error and the historical error sequence; Input the historical frequency modulation instruction sequence into the prediction model after iterative weight adjustment to obtain a new current frequency modulation instruction; obtain a new current error based on the new current frequency modulation instruction; The above process is iterated until the current error is less than the preset value.

[0011] Furthermore, the iterative weight of the prediction model after the iterative weight adjustment is:

[0012] in, c is the number of iterations for iterative weight adjustment, For the c The iteration weight of the iteration; is a fitted nonlinear function; Pan-error set; is the summation symbol.

[0013] Further, the

[0014] in, is the length of the historical error sequence, for t The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; for ti The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; is the Euler number.

[0015] Further, the error trend analysis method is:

[0016] in, is the length of the historical error sequence, is the error trend; for ti The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; for t-i+ The root mean square error between the predicted frequency modulation command at time 1 and the actual frequency modulation command; rand () is a random function; express Frequency modulation instructions at all times, The historical error sequence i The time difference factor corresponding to the location.

[0017] Furthermore, the error acceleration is ; The fine-tuning coefficient is: ; The method for setting the fine-tuning coefficient is to add the fine-tuning coefficient to each item of the historical frequency modulation instruction sequence at the next moment; Indicates the error trend About Time The derivative of ; sin() represents the sine function.

[0018] On the other hand, the present invention discloses a storage medium storing a computer program, which executes the steps included in the above method when called and run by a computer.

[0019] On the other hand, the present invention discloses a device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0020] In the last aspect, the present invention discloses a super-capacity energy storage frequency modulation instruction prediction system, comprising: Data acquisition module: used to acquire a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a current frequency modulation instruction and its historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into a prediction model to obtain a predicted current frequency modulation instruction; the current error is obtained based on the predicted current frequency modulation instruction; Weight adjustment module: According to the size of the current error and based on the historical error sequence, the iterative weight of the prediction model is adjusted until the current error is less than the preset value, and the prediction model after iterative weight adjustment and the new current error are obtained; Error trend analysis module: Based on the new current error and historical error sequence, error trend analysis is performed to obtain error acceleration; Input fine-tuning module: according to the size of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the historical frequency modulation instruction sequence at the next moment with the set fine-tuning coefficient is input into the prediction model after iterative weight adjustment to obtain the predicted frequency modulation instruction at the next moment; the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a method for predicting frequency modulation instructions for ultra-capacity energy storage. By combining the error analysis of historical frequency modulation instruction sequences and current frequency modulation instructions, the prediction model of frequency modulation instructions is optimized; the real-time dynamic adjustment of the prediction model weights can predict the dynamic update of the model, ensuring that the prediction model always remains in the best state, thereby reducing the prediction error. If the current error is less than the preset value, it means that the prediction model already has a high prediction accuracy and can meet the needs of practical applications; by performing error trend analysis on the new current error and historical error sequence, the error acceleration is obtained, which can accurately capture the error change trend and provide a basis for the subsequent fine-tuning coefficient setting; a fine-tuning coefficient is set for the frequency modulation instruction sequence according to the error acceleration, which can achieve fine adjustment of the frequency modulation instruction, further improving the accuracy and effectiveness of the frequency modulation instruction prediction.

[0022] Furthermore, the ultra-capacity energy storage frequency modulation instruction prediction system of the present invention provides accurate and timely frequency modulation instruction prediction for the energy storage system, significantly improving the operating efficiency and stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 This is a flow chart of the method for predicting frequency modulation instructions of super-capacity energy storage of the present invention; Figure 2 This is a simplified flow chart of the method for predicting frequency modulation instructions of super-capacity energy storage of the present invention; Figure 3 A simplified diagram of the process for fine-tuning the forecast; Figure 4 This is a block diagram of the super-capacity energy storage frequency modulation instruction prediction system of the present invention.

[0025] Among them, 401-data acquisition module; 402-weight adjustment module; 403-error trend analysis module; 404-input fine-tuning module. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention described and marked in the drawings here can be arranged and designed in various different configurations.

[0027] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0029] Terminology explanation: 1. Root mean square error, the English name is Root Mean Square Error, abbreviated as RMSE.

[0030] 2. Solid State Drive, the English name is Solid State Disk, abbreviated as SSD.

[0031] 3. Flash drive, the English name is Universal Serial Bus, abbreviated as USB.

[0032] 4. Compact Disc, abbreviated as CD.

[0033] 5. Secure Digital Memory Card, the English name is Secure Digital Memory Card, abbreviated as SD.

[0034] 6. Digital Video Disc, abbreviated as DVD.

[0035] The present invention is further described in detail below in conjunction with the accompanying drawings: Embodiment 1: See also Figure 1 The present invention provides a method for predicting frequency modulation instructions of super-capacity energy storage, comprising the following steps: S1, obtain a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes the current frequency modulation instruction and its historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into the prediction model to obtain the predicted current frequency modulation instruction; the current error is obtained based on the predicted current frequency modulation instruction. The prediction model of the present invention can select an LSTM model or a BP model. In this embodiment, the frequency modulation instruction sequence of the past month is obtained from the historical database, including the frequency modulation instruction values ​​every day and every hour. These data are arranged in chronological order to form a historical frequency modulation instruction sequence. At the same time, the actual frequency modulation requirements corresponding to these frequency modulation instructions are obtained, and the prediction error at each moment (that is, the difference between the predicted frequency modulation instruction and the actual frequency modulation requirement) is calculated to form a historical error sequence. The first 90% of the historical frequency modulation instruction sequence is used as input, and the LSTM model is used to predict the frequency modulation instruction at the current moment; based on the prediction result of the LSTM model, the prediction error at the current moment, that is, the current error, is calculated.

[0036] S2, according to the size of the current error and based on the historical error sequence, adjust the iterative weight of the prediction model until the current error is less than the preset value, and obtain the prediction model after iterative weight adjustment and the new current error, specifically including: S201, determining whether the current error is less than a preset value; specifically, checking whether the current error is less than a preset value of 0.05; S202, when the current error is greater than or equal to a preset value, the iteration weight of the prediction model is adjusted according to the current error and the historical error sequence; the adjusted iteration weight is:

[0037] in, c is the number of iterative weight adjustments, For the c The iteration weight of the iteration; is a fitted nonlinear function; Universal error set.

[0038]

[0039] in, n is the length of the historical error sequence, for t The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; for ti The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment.

[0040] S203, inputting the historical frequency modulation instruction sequence into the prediction model after iterative weight adjustment again to obtain a new current frequency modulation instruction; and calculating a new current error; S204, iterate the process from S201 to S203 until the current error is less than a preset value.

[0041] S3, perform error trend analysis based on the new current error and historical error sequence to obtain error acceleration; The error trend analysis method is:

[0042] in, n is the length of the historical error sequence, S is the error trend; for t-i+ The root mean square error between the predicted frequency modulation command at time 1 and the actual frequency modulation command; rand () is a random function; express Frequency modulation instructions at all times; The historical error sequence i The time difference factor corresponding to the location.

[0043] then S About time t The derivative of the error acceleration is ; Indicates the error trend At the moment The differential of .

[0044] S4, according to the error acceleration The size of the frequency modulation instruction sequence at the next moment is set to a fine-tuning coefficient ; Input the historical frequency modulation instruction sequence of the next moment with the fine-tuning coefficient set into the prediction model after the iterative weight adjustment to obtain the predicted frequency modulation instruction of the next moment; the historical frequency modulation instruction sequence of the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction of the next moment; when When , a fine-tuning coefficient is set for the frequency modulation instruction sequence: ; and add the fine-tuning coefficient to each item of the historical frequency modulation instruction sequence at the next moment; sin() represents the sine function.

[0045] In this embodiment, by using the LSTM model (or other neural network models such as the BP model) and combining the historical frequency modulation instruction sequence for training, the model can learn the changing rules and trends of the frequency modulation instructions, thereby improving the prediction accuracy of future frequency modulation instructions. By iteratively adjusting the weight of the model, the model parameters are continuously optimized according to the current error and the historical error sequence, and the prediction error is further reduced, so that the prediction result is closer to the actual frequency modulation demand. The introduction of the error trend analysis and fine-tuning coefficient setting mechanism enables the model to automatically adjust the prediction strategy according to the changing trend of the historical error, and enhances the adaptability to different working conditions and frequency modulation requirements. The calculation of the error acceleration provides a basis for dynamic adjustment of the model, so that the model can respond faster when facing emergencies or changes in frequency modulation requirements. The accurate frequency modulation instruction prediction of the present invention helps the power system to more effectively manage the charging and discharging strategy of energy storage equipment. By predicting the frequency modulation demand in advance, the charging and discharging time and capacity of the energy storage equipment can be reasonably arranged to avoid equipment loss and energy efficiency reduction caused by excessive charging and discharging. The optimized charging and discharging strategy of the energy storage equipment helps to improve the stability and economy of the power system and reduce operating costs and environmental impact.

[0046] Embodiment 2: See also Figure 2 and Figure 3 The present invention provides a method for predicting frequency modulation instructions of super-capacity energy storage, comprising the following steps: S1, obtaining a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a historical frequency modulation instruction sequence and a current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence, and the error corresponding to the current frequency modulation instruction is a current error; The frequency modulation instruction sequence is a time series, which reflects the frequency adjustment requirements of the power system at different time points. The present invention adjusts the prediction model by analyzing the historical errors of this time series, thereby more accurately predicting future frequency modulation instructions.

[0047] Let the current time be t At this moment, the current frequency modulation instruction is t Frequency modulation command at the moment; selectt The 30 sampling sequences 5 minutes before the moment are used as the historical frequency modulation instruction sequence, that is, when t The frequency modulation instruction at this moment is When The corresponding historical frequency modulation instruction sequence is: ;but The corresponding frequency modulation instruction sequence is: .

[0048] Based on the above frequency modulation instruction sequence, the corresponding error sequence is obtained; the frequency modulation instruction sequence includes a historical frequency modulation instruction sequence and a current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is the historical error sequence, and the error corresponding to the current frequency modulation instruction is the current error; then the current error is the root mean square error of the model prediction at the current moment; the historical error sequence is: .

[0049] S2, based on the current error, adjust the iterative weight of the prediction model according to the historical error sequence until the current error is less than the preset value, and obtain the prediction model after iterative weight adjustment and the new current error; According to S1, the current error is ,when When , the iterative weight of the prediction model is adjusted until the current error is less than 8; the iterative weight of the prediction model after the iterative weight adjustment is:

[0050] in, c is the number of iterative weight adjustments, For the c The weight of the sub-cycle;

[0051] In the formula, is a fitted nonlinear function, and Both are activation functions; is the general error set:

[0052] S3, perform error trend analysis on the new current error and historical error sequence to obtain error acceleration; Based on S2, when When , the error trend is analyzed, and the analysis method is:

[0053] in is the frequency modulation instruction sequence, is the error trend, The historical error sequence i The time difference factor corresponding to the location; rand () represents a random function.

[0054] Through the above error trend analysis, the error acceleration ℧ is obtained. , Indicates the error trend At the moment The differential of .

[0055] S4, according to the magnitude of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the frequency modulation instruction sequence to be predicted with the fine-tuning coefficient set is input into the prediction model after the iterative weight adjustment, and the predicted frequency modulation instruction at the next moment is obtained; if , a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment, and the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment; the method of setting a fine-tuning coefficient for the predicted frequency modulation instruction sequence is: adding a fine-tuning coefficient to each item of the predicted frequency modulation instruction sequence, and the fine-tuning coefficient is: , sin() represents the sine function; get the fine-tuned frequency modulation instruction sequence: ; Input the fine-tuned frequency modulation instruction sequence into the prediction model after iterative weight adjustment to predict the frequency modulation instruction at the predicted future moment; if , then the frequency modulation instruction sequence is input into the prediction model after iterative weight adjustment to obtain the predicted frequency modulation instruction at the next moment.

[0056] The present invention uses the historical frequency modulation instruction data of the power system to train the prediction model; first, the ultra-capacity energy storage frequency modulation instruction prediction method of the present invention is used to train the prediction model, that is, the weight of the prediction model is adjusted by the above method to obtain a prediction model with adjusted weights; then, the frequency modulation instruction sequence is input into the prediction model with adjusted weights to perform error trend analysis to obtain a fine-tuning coefficient, and each item of the historical frequency modulation instruction sequence at the next moment is added with the fine-tuning coefficient and input into the prediction model with adjusted weights to obtain a predicted frequency modulation instruction.

[0057] Embodiment 3: The frequency modulation instruction prediction method of the present invention has a high degree of flexibility and portability in its software implementation form. When this method is converted into a computer program, it can be stored in various storage media so that it can be called and run when needed. These storage media include but are not limited to optical disks (such as CDs, DVDs, etc.), flash memories (such as USB flash drives, solid-state hard disks SSDs, etc.), SD cards, and various other portable or non-portable storage devices. The characteristics of these storage media are that they have large capacity, high speed, durability, and portability, and can safely and stably store and transmit computer programs.

[0058] When the computer programs in these storage media are called and run by the computer, they will execute the various steps of the above-mentioned frequency modulation instruction prediction method according to the preset logic and algorithm. This includes key steps such as obtaining the frequency modulation instruction sequence and its corresponding error sequence, adjusting the weight of the prediction model based on the historical error sequence, performing error trend analysis, and setting the fine-tuning coefficient. In this way, the method of the present invention can be implemented in various computer systems in the form of software and exert its unique advantages.

[0059] The present invention has high flexibility and diversity in realizing its various functions. It can be realized not only by pure hardware, such as using specific integrated circuits, field programmable gate arrays and other hardware devices to execute the logic and algorithm of frequency modulation instruction prediction, but also by pure software, such as writing the frequency modulation instruction prediction method into a computer program and running it on a general-purpose computer or embedded system.

[0060] In addition, the present invention can also be implemented in a combination of hardware and software, that is, the high performance of the hardware device and the flexibility of the software are used to jointly complete the task of frequency modulation instruction prediction. For example, a high-performance processor and memory can be used to perform complex computing tasks, while software can be used to manage and control the entire prediction process. This combination can give full play to the respective advantages of hardware and software, and improve the accuracy and efficiency of frequency modulation instruction prediction.

[0061] In the process of implementing the present invention, multiple steps or methods can be executed by software or firmware stored in a memory and executed by an appropriate instruction execution system. This approach not only facilitates program writing, debugging and maintenance, but also achieves cross-platform compatibility, so that the method of the present invention can be widely used in different computer systems.

[0062] Embodiment 4: See also Figure 4 The present invention provides a super-capacity energy storage frequency modulation instruction prediction system, comprising: Data acquisition module 401: acquiring a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a current frequency modulation instruction and its historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into a prediction model to obtain a predicted current frequency modulation instruction; and the current error is obtained based on the predicted current frequency modulation instruction; Weight adjustment module 402: according to the size of the current error and based on the historical error sequence, adjust the iterative weight of the prediction model until the current error is less than a preset value, and obtain the prediction model after iterative weight adjustment and the new current error; Error trend analysis module 403: performs error trend analysis based on the new current error and the historical error sequence to obtain error acceleration; Input fine-tuning module 404: according to the size of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the historical frequency modulation instruction sequence at the next moment with the set fine-tuning coefficient is input into the prediction model after iterative weight adjustment to obtain the predicted frequency modulation instruction at the next moment; the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment.

[0063] Embodiment 5: In order to further verify the advantages of the frequency modulation instruction prediction method of the ultra-capacity energy storage of the present invention, the prediction method of the present invention and the traditional model are used to predict the frequency modulation instruction. The prediction model of this embodiment selects the BP model: Dataset: The frequency regulation data of a power plant on June 19 and June 20, 2016 were selected, with 800 data points each day. These data points cover the changes in frequency regulation instructions in different time periods and are widely representative.

[0064] Data division: In order to fairly evaluate the performance of the prediction method, this embodiment divides the daily data into 80% and 20%. Among them, 80% of the data (i.e., 640 data points) is used as the training set for training the prediction model; the remaining 20% ​​of the data (i.e., 160 data points) is used as the test set to verify the accuracy of the prediction model.

[0065] Comparison method: In order to highlight the advantages of the present invention, traditional prediction models are selected as comparison methods. These traditional models have a certain application basis in power systems and are used to predict frequency modulation instructions in this experiment.

[0066] The results are as follows:

[0067] The experimental results show that the prediction method of the present invention shows higher accuracy in frequency modulation instruction prediction. Whether for frequency modulation instruction 1 or frequency modulation instruction 2, the difference between the prediction result of the present invention and the actual frequency modulation instruction is smaller than that of the traditional model, and the prediction method of the present invention has higher prediction accuracy. In addition, the prediction method of the present invention also shows good stability. In the prediction of frequency modulation instructions in different time periods, the prediction results of the present invention can maintain a high accuracy, while the prediction results of the traditional model have certain fluctuations.

[0068] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting frequency modulation instructions of supercapacitor energy storage, characterized in that: The following steps are involved: S1, obtaining a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a current frequency modulation instruction and its historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into a prediction model to obtain a predicted current frequency modulation instruction; the current error is obtained based on the predicted current frequency modulation instruction; wherein the prediction model is an LSTM model or a BP model; S2, according to the size of the current error and based on the historical error sequence, adjust the iterative weight of the prediction model until the current error is less than the preset value, and obtain the prediction model after iterative weight adjustment and the new current error; S3, based on the new current error and historical error sequence, error trend analysis is performed to obtain error acceleration; S4, according to the size of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the historical frequency modulation instruction sequence at the next moment with the set fine-tuning coefficient is input into the prediction model after iterative weight adjustment to obtain the predicted frequency modulation instruction at the next moment; the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment.

2. The method for predicting frequency modulation instructions of super-capacity energy storage according to claim 1, characterized in that: The historical frequency modulation instruction sequence is the frequency modulation instructions at several moments before the current moment.

3. The method for predicting frequency modulation instructions of supercapacitor energy storage according to claim 1, characterized in that: According to the size of the current error, based on the historical error sequence, the weight of the prediction model is adjusted until the current error is less than the preset value, and the prediction model after iterative weight adjustment and the new current error are obtained. The specific method is: Determine whether the current error is less than the preset value; when the current error is greater than or equal to the preset value, adjust the iteration weight of the prediction model based on the current error and the historical error sequence; Input the historical frequency modulation instruction sequence into the prediction model after iterative weight adjustment to obtain a new current frequency modulation instruction; obtain a new current error based on the new current frequency modulation instruction; The above process is iterated until the current error is less than the preset value.

4. The method for predicting frequency modulation instructions of supercapacitor energy storage according to claim 3 is characterized in that: The iterative weight of the prediction model after the iterative weight adjustment is: in, c is the number of iterations for iterative weight adjustment, For the c The iteration weight of the iteration; To fit the nonlinear function; Pan-error set; is the summation symbol.

5. The method for predicting frequency modulation instructions of super-capacity energy storage according to claim 4 is characterized in that: The universal error set The calculation formula is: in, n is the length of the historical error sequence, for t The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; for ti The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; e is the Euler number.

6. The method for predicting frequency modulation instructions of supercapacitor energy storage according to claim 1, characterized in that: The error trend analysis method is: in, is the length of the historical error sequence, is the error trend; for t - i The root mean square error between the predicted frequency modulation command and the actual frequency modulation command at each moment; for t - i The root mean square error between the predicted frequency modulation command at time +1 and the actual frequency modulation command; rand () is a random function; express Frequency modulation instructions at all times; The historical error sequence i The time difference factor corresponding to the location.

7. The method for predicting frequency modulation instructions of super-capacity energy storage according to claim 6 is characterized in that: The error acceleration is ; The fine-tuning coefficient is: ; The method for setting the fine-tuning coefficient is to add the fine-tuning coefficient to each item of the historical frequency modulation instruction sequence at the next moment; Indicates the error trend At the moment The derivative of ; sin() represents the sine function.

8. A storage medium, characterized in that: The computer program stores a computer program, which, when called and run by a computer, executes the steps of the method described in any one of claims 1 to 7.

9. A device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A supercapacitor energy storage frequency modulation instruction prediction system, based on the supercapacitor energy storage frequency modulation instruction prediction method according to claim 1, characterized in that: include: Data acquisition module: used to acquire a frequency modulation instruction sequence and its corresponding error sequence; the frequency modulation instruction sequence includes a current frequency modulation instruction and its historical frequency modulation instruction sequence; the historical frequency modulation instruction sequence is used to predict the current frequency modulation instruction; the error sequence corresponding to the historical frequency modulation instruction sequence is a historical error sequence; the historical frequency modulation instruction sequence is input into a prediction model to obtain a predicted current frequency modulation instruction; the current error is obtained based on the predicted current frequency modulation instruction; Weight adjustment module: According to the size of the current error and based on the historical error sequence, the iterative weight of the prediction model is adjusted until the current error is less than the preset value, and the prediction model after iterative weight adjustment and the new current error are obtained; Error trend analysis module: Based on the new current error and historical error sequence, error trend analysis is performed to obtain error acceleration; Input fine-tuning module: according to the size of the error acceleration, a fine-tuning coefficient is set for the historical frequency modulation instruction sequence at the next moment; the historical frequency modulation instruction sequence at the next moment with the set fine-tuning coefficient is input into the prediction model after weight adjustment to obtain the predicted frequency modulation instruction at the next moment; the historical frequency modulation instruction sequence at the next moment is the historical frequency modulation instruction sequence corresponding to the predicted frequency modulation instruction at the next moment.

Images