Long-period new energy power-risk prediction method and system based on EMD-GRU
Through the EMD-GRU method, time-frequency decomposition and GRU network prediction of new energy output data, combined with risk indicators, the randomness and periodic characteristics problems in new energy output prediction are solved, the prediction accuracy and training speed are improved, and the prediction is achieved. Power supply and demand balance.
Patent Information
- Application Number
- CN202411309436.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-19
- Publication Date
- 2024-12-27
AI Technical Summary
New energy output prediction faces the characteristics of randomness, volatility and periodicity, and the data is a non-stationary nonlinear signal. The existing technology is difficult to effectively process high-dimensional data, resulting in difficulty in prediction.
Using an EMD-GRU-based method, the historical data of new energy power generation output is time-frequency decomposed by empirical modal decomposition (EMD), the sub-sequence characteristics are extracted, and the sub-sequences are predicted using the GRU network, combining risk as a power quality indicator , simplify power planning.
It improves the accuracy and training speed of new energy output prediction, simplifies the training process, can effectively extract time-frequency characteristics in timing, reduce prediction errors, and ensure the supply and demand balance between power supply and load.
Smart Images
Figure CN119204679A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of long-term renewable energy output prediction, and in particular to a long-term renewable energy power-risk prediction method and system based on EMD-GRU. Background Art
[0002] At present, there are endless methods and models for forecasting the output of new energy sources, which are mainly divided into statistical methods, artificial intelligence methods and combined methods. For wind power generation, environmental factors such as wind speed, wind direction, temperature, and humidity have an important impact on wind power output; for photovoltaic power generation, its output is strongly linearly related to solar irradiance, and the strength of solar irradiance is affected by many factors, including solar altitude, sunshine time, geographical location, and various atmospheric indicators. Among them, wind speed, temperature, atmospheric indicators, etc. are affected by weather conditions and have strong randomness; while solar altitude and sunshine time are determined by the rotation and revolution of the earth and have strong periodicity. Therefore, the influencing factors of new energy output prediction show the characteristics of randomness, volatility, and periodicity, and are also from multi-dimensional data. In addition, the time series of new energy output is a non-stationary and nonlinear signal, and it is very difficult to directly predict its power generation output curve in time series. It is also difficult to process high-dimensional data, especially such nonlinear and non-stationary data, with the help of statistical methods. Summary of the invention
[0003] In order to solve the above technical problems, the present invention proposes a long-term new energy power-risk prediction method and system based on EMD-GRU, firstly performs time-frequency decomposition on the historical data signal of new energy power generation output, respectively performs time series analysis and prediction on the decomposed subsequences, and then superimposes the subsequence prediction results to obtain the final prediction value. The decomposed subsequences still have certain non-stationary and nonlinear characteristics. In order to improve the prediction accuracy, considering the research progress of the model and algorithm of deep learning, the GRU network algorithm is selected to predict the subsequences. The combination method used by the present invention is innovative in this field. While ensuring high prediction accuracy, it can also improve the training speed and simplify the training process.
[0004] EMD (Empirical Mode Decomposition) is a very practical signal time-frequency analysis technology. Compared with traditional Fourier decomposition and wavelet decomposition, Fourier decomposition needs to extract all the time domain information of the signal, and wavelet decomposition requires the information to be stable in the frequency domain. When using the two to decompose non-stationary time series and nonlinear signals, it will be difficult to extract all the time-frequency characteristics of the signal, which will produce large errors. When decomposing the signal, EMD can reflect the time domain and frequency domain characteristics of the signal itself, and does not require the pre-selection of basis functions. It is an adaptive signal decomposition method.
[0005] Risk degree and power value are introduced to characterize the time series fluctuation characteristics of renewable energy output. Risk degree is used as an indicator to characterize power quality to simplify the description of renewable energy operation characteristics. The power value corresponding to risk degree is used as an indicator to characterize power flow to describe the scale of power supply and demand. On the basis of considering the fluctuation characteristics of renewable energy, the power supply planning results and the scale of flexibility resource demand are given simply, intuitively and quickly.
[0006] Among the numerous AI model algorithms, the LSTM (Long Short-Term Memory Network) algorithm is highly favored in the field of long-term renewable energy output prediction because of its long-term memory function, its ability to effectively utilize the long-term dependencies of limited data samples, its ability to effectively characterize the complex time correlation of renewable energy output time series, and its low risk of gradient explosion or disappearance. GRU (Gated Recurrent Unit), as a variant of LSTM, has fewer parameters and faster training speed than LSTM networks, which can simplify the algorithm flow during the prediction process.
[0007] To achieve the above object, the present invention adopts the following technical solution: A long-term new energy power-risk prediction method based on EMD-GRU, comprising: In the historical data processing stage, the historical data of renewable energy power generation output for 24 hours a day for 365 days in a year is collected; Standardizing the acquired historical data to obtain a sequence of standardized historical data; Perform EMD decomposition on the sequence of standardized historical data, and determine the number of decomposed IMF subsequences according to the specific situation of the historical data; analyze the characteristics of the IMF subsequences, remove the IMF subsequences with the highest and lowest frequencies, and retain the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; The decomposed IMF subsequences are segmented into output curves, and the risk level is proposed as an indicator to describe the power quality in power pinch point analysis. In the output prediction stage, the GRU model is used to predict each IMF subsequence separately. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. By superimposing the forecast values of each IMF sub-sequence, the final result of the long-term renewable energy power generation output forecast is obtained.
[0008] Furthermore, the historical data are from the same altitude and the same geographical location coordinates, and the resolution of all data is unified to 1 hour.
[0009] Furthermore, the normalization process is performed according to the following formula: (1) in, For each historical data sample, is the sample mean, is the standard deviation of the sample.
[0010] Further, the EMD decomposition is performed on the sequence of the historical data after the standardization process, and the number of the decomposed IMF subsequences is determined according to the specific situation of the historical data; the characteristics of the IMF subsequences are analyzed, the IMF subsequences with the highest and lowest frequencies are removed, and the remaining IMF subsequences and the residual sequence are retained, including: First find the original signal All the maximum points of , and the maximum envelope is fitted by cubic spline function interpolation method ; Find the original signal All the minimum points of the minimum value are fitted by the cubic spline function interpolation method to obtain the minimum envelope curve ; t is any specific time, calculate the average value of the upper and lower envelopes : (2) Subtract the original signal sequence from You get a new signal without the low frequency. : (3) Repeat the above process, assuming that after k times, If the definition of IMF is satisfied, then the original signal The first-order IMF component of is: (4) With the original signal minus , and get a new signal without high-frequency components : (5) Based on the repetition The second IMF component is obtained by , and so on, until the nth order IMF component or residual Less than the preset value; or when the residual When is a monotonic function or a constant, the EMD decomposition process stops; Finally, the original signal is decomposed by EMD and expressed as: (6) Among them, i is the loop order number, and n is the final number of loops.
[0011] Furthermore, the output curve of the decomposed IMF subsequences is segmented, and the risk degree is proposed as an index for describing power quality in the analysis of electrical pinch points, including: Assume that the wind power output in the system is , and the photovoltaic output is , and the new energy output is , and its upper and lower limits are respectively and ; The new energy output curve is divided into N intervals at intervals of . Starting from the power lower limit , the risk degree of the nth interval is expressed as follows: (7) Wherein, represents the letter t and its subscript, representing 1 time interval; n is the number of intervals, and the number of intervals n < N. The maximum value that the number of intervals n can take is : (8) The intermediate value within the interval is used as the power value corresponding to the risk degree. The power value corresponding to the risk degree of the nth interval is as follows: (9).
[0012] Furthermore, in the output prediction stage, the GRU model is used to predict each IMF subsequence respectively. The input variables are the data of each IMF subsequence and the meteorological factors at the corresponding time points, and the output is the prediction result corresponding to each IMF subsequence, including: The update gate determines how much historical information and current information are used to update the hidden state; The update gate at the t-th moment is: (10) Wherein, is the gated update signal, represents the output of the hidden state at the historical t-1 moment, represents the input data at the current t moment, is the sigmoid function, is the weight matrix; The reset gate determines how much historical information to retain. The reset gate at the t-th moment is: (11) Wherein, is the reset signal, is the weight matrix; In the update door and reset gate Under the action of , the candidate hidden state at the current moment, the implicit output state Updated to: (12) Among them, the candidate hidden state is: (13) In the above formula, is a candidate hidden state, represents the hyperbolic tangent function, is the weight matrix.
[0013] The present invention also provides a long-term new energy power-risk prediction system based on EMD-GRU, comprising the following modules: The historical data collection module collects historical data of renewable energy power generation output for 24 hours a day for 365 days in a year during the historical data processing stage; A standardization module performs standardization processing on the acquired historical data to obtain a sequence of the standardized historical data; The decomposition module performs EMD decomposition on the sequence of historical data after standardization, and determines the number of IMF subsequences to be decomposed according to the specific situation of the historical data; analyzes the characteristics of the IMF subsequences, removes the IMF subsequences with the highest and lowest frequencies, and retains the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; The indicator proposal module segments the output curve of the decomposed IMF subsequences and proposes the risk level as an indicator to describe the power quality in the power pinch point analysis; The prediction module uses the GRU model to predict each IMF subsequence in the output prediction stage. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. The result output module obtains the final result of the long-term renewable energy power generation output forecast by superimposing the forecast values of each IMF sub-sequence.
[0014] Furthermore, the historical data in the historical data collection module comes from the same altitude and the same geographical location coordinates, and the resolution of all data is unified to 1 hour.
[0015] Furthermore, the standardization processing of the standardization module is performed according to the following formula: (1) in, For each historical data sample, is the sample mean, is the standard deviation of the sample.
[0016] Furthermore, the decomposition module implements: First find the original signal All the maximum points of , and the maximum envelope is fitted by cubic spline function interpolation method ; Find the original signal All the minimum points of the minimum value are fitted by the cubic spline function interpolation method to obtain the minimum envelope curve ; t is any specific time, calculate the average value of the upper and lower envelopes : (2) Subtract the original signal sequence from We get a new signal without the low frequency. : (3) Repeat the above process, assuming that after k times, If the definition of IMF is satisfied, then the original signal The first-order IMF component of is: (4) With the original signal minus , and get a new signal without high-frequency components : (5) Based on the repetition The second IMF component is obtained by , and so on, until the nth order IMF component or residual Less than the preset value; or when the residual When is a monotonic function or a constant, the EMD decomposition process stops; Finally, the original signal is decomposed by EMD and expressed as: (6) Among them, i is the loop order number, and n is the final number of loops.
[0017] Furthermore, the indicator proposal module includes: Assume that the wind power output in the system is , photovoltaic output is , the new energy output is , whose upper and lower limits are and ; The new energy output curve is The interval is divided into N intervals, with the power lower limit Taking this as the starting point, the risk degree of the nth interval The expression is as follows: (7) where represents the letter t and its subscript, representing 1 time interval; n is the number of intervals, and the number of intervals n < N. The maximum value that the number of intervals n can take is: (8) Taking the intermediate value within the interval as the power value corresponding to the risk degree, the risk degree of the nth interval The corresponding power value is as follows: (9).
[0018] Furthermore, the prediction module realizes: The update gate determines how much historical information and current information are used to update the hidden state; The update gate at the t-th moment is: (10) where is the gated update signal, represents the output of the hidden state at the historical (t - 1) moment, represents the input data at the current t moment, is the sigmoid function, is the weight matrix; The reset gate determines how much historical information to retain. The reset gate at the t-th moment is: (11) where is the reset signal, is the weight matrix; Under the action of the update gate and the reset gate , the candidate hidden state at the current moment, the hidden output state is updated to: (12) where the candidate hidden state: (13) In the above formula, is the candidate hidden state, represents the hyperbolic tangent function, is the weight matrix.
[0019] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the above-mentioned long-period new energy power-risk prediction method based on EMD-GRU are implemented.
[0020] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned long-period new energy power-risk prediction method based on EMD-GRU.
[0021] Beneficial effects: 1. Since the output of renewable energy has strong randomness and volatility, and the original sequence information is complex, the EMD decomposition method can effectively extract the characteristics of each time frequency in the time series, and the prediction error is small.
[0022] 2. During the power pinch point analysis, the power distribution and power level of new energy are described by power value and risk degree respectively, which can not only ensure the energy supply and demand balance between power source and load, but also take into account the impact of new energy operation characteristics on planning results.
[0023] 3. The GRU algorithm has a long-term memory function, can effectively utilize the long-term dependence of limited data samples, can effectively characterize the complex time correlation of new energy output time series, and is not prone to gradient explosion or disappearance problems. At the same time, the GRU network has fewer parameters and faster training speed, which can simplify the algorithm process during the prediction process. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 It is a flow chart of the long-period new energy output power-risk prediction method based on EMD-GRU of the present invention. DETAILED DESCRIPTION
[0025] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.
[0026] like Figure 1 As shown, a long-period new energy power-risk prediction method based on EMD-GRU of the present invention includes a historical data processing stage and an output prediction stage.
[0027] The following steps are implemented in the historical data processing stage: Step 1: In the historical data processing stage, collect the historical data of renewable energy power generation output for 365 days a year, such as wind power, photovoltaic power, etc. Step 2: Standardize the historical data obtained in step 1 to obtain a sequence of standardized historical data; Step 3: Perform EMD decomposition on the sequence of historical data after standardization in step 2, and determine the number of IMF subsequences to be decomposed according to the specific situation of the historical data; analyze the characteristics of the IMF subsequences, remove the IMF subsequences with the highest and lowest frequencies, and retain the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; Step 4: Segment the output curve of the decomposed IMF sequence and propose the risk level as an indicator to describe the power quality in the power pinch analysis.
[0028] The following steps are implemented in the output prediction stage: Step 5: Use the GRU model to predict each IMF subsequence separately. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. Step 6: By superimposing the forecast values of each IMF sub-sequence, the final result of the long-term renewable energy power generation output forecast is obtained.
[0029] Specifically, the historical data in step 1 should be from the same altitude and the same geographical location coordinates, and the resolution of all data should be unified to 1 hour.
[0030] Specifically, the step 2 is standardized according to the following formula: (1) in, For each historical data sample, is the sample mean, is the standard deviation of the sample.
[0031] Specifically, step 3 is decomposed according to the following formula and steps: First find the original signal All the maximum points of , and the maximum envelope is fitted by cubic spline function interpolation method ; Similarly, find the original signal All the minimum points of the minimum value are fitted by the cubic spline function interpolation method to obtain the minimum envelope curve . t is any specific time. Calculate the average value of the upper and lower envelopes : (2) The original signal minus A new signal with low frequencies removed is obtained. : (3) Generally it is not a stationary signal and does not meet the two conditions defined by IMF. Repeat the above process. Assume that after k times (k is generally less than 10) it meets the definition of IMF, then the first-order IMF component of the original signal is: denotes the first-order IMF component of the original signal ; (4) Use the original signal to subtract to obtain a new signal with high-frequency components removed : (5) According to the repeated process of obtaining the second IMF component is obtained. Repeat this process until the nth-order IMF component or the residual component is less than the preset value; or when the residual component is a monotonic function or a constant, the EMD decomposition process stops.
[0032] Finally, the original signal after EMD decomposition can be expressed as: (6) where i is the loop sequence number and n is the final number of loops.
[0033] Specifically, step 4 includes: Assume that the wind power output in the system is , the photovoltaic power output is , and the new energy output is . The upper and lower limits of its new energy output are and .
[0034] Divide the new energy output curve into N intervals at intervals of . Starting from the lower limit of the new energy output , the risk degree of the nth interval is expressed as follows: (7) where is t and its subscript, representing a time interval of 1 . n is the number of intervals, and the number of intervals n < N. The maximum value it can take for: (8) The middle value in the interval is taken as the power value of the corresponding risk, and the risk of the nth interval is The corresponding power value as follows: (9) Specifically, in step 5, the basic formula of GRU is as follows: The update gate determines how much historical information and current information to use to update the hidden state.
[0035] The update gate at time t is: (10) in, is the gated update signal, The size of determines the degree of memory of the candidate hidden state. represents the implicit state output at the historical time t-1, represents the input data at the current time t, is the sigmoid function, is the weight matrix.
[0036] The reset gate determines how much historical information is retained. The reset gate at time t is: (11) in, The reset signal is a reset signal. The larger the reset signal value is, the more historical information needs to be remembered. is the weight matrix.
[0037] Under the action of the update gate and the reset gate, the candidate hidden state at the current moment and the hidden output state Can be updated to: (12) Among them, the candidate hidden states are: (13) In the above formula, is a candidate hidden state, represents the hyperbolic tangent function, is the weight matrix.
[0038] The candidate hidden state is responsible for fusing the information features of the input data and historical data. This operation is consistent with the reset signal obtained by the reset gate. About. Represents the final unit state at the current moment, which includes two processes: forgetting and remembering. The implicit state of the previous moment The product of represents the forgetting process, The closer it is to 1, the more information from the previous moment will be forgotten. The product of the candidate hidden state represents the memory process, The size determines the memory level of the candidate hidden state, that is, how many previous hidden states are retained. The above process means that as much new memory is added, as much old memory is forgotten.
[0039] The present invention also provides a long-term new energy power-risk prediction system based on EMD-GRU, comprising the following modules: The historical data collection module collects historical data of renewable energy power generation output for 24 hours a day for 365 days in a year during the historical data processing stage, such as wind power, photovoltaic power, etc. A standardization module performs standardization processing on the acquired historical data to obtain a sequence of the standardized historical data; The decomposition module performs EMD decomposition on the sequence of historical data after standardization, and determines the number of IMF subsequences to be decomposed according to the specific situation of the historical data; analyzes the characteristics of the IMF subsequences, removes the IMF subsequences with the highest and lowest frequencies, and retains the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; The indicator proposal module segments the output curve of the decomposed IMF subsequences and proposes the risk level as an indicator to describe the power quality in the power pinch point analysis.
[0040] The prediction module uses the GRU model to predict each IMF subsequence in the output prediction stage. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. The result output module obtains the final result of the long-term renewable energy power generation output forecast by superimposing the forecast values of each IMF sub-sequence.
[0041] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the above-mentioned long-period new energy power-risk prediction method based on EMD-GRU are implemented.
[0042] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned long-period new energy power-risk prediction method based on EMD-GRU.
[0043] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The solutions in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and interpreted scripting language JavaScript, etc.
[0044] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0045] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0046] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0047] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0048] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A long-term new energy power-risk prediction method based on EMD-GRU, characterized in that: include: In the historical data processing stage, the historical data of renewable energy power generation output for 24 hours a day for 365 days in a year is collected; Standardizing the acquired historical data to obtain a sequence of standardized historical data; Perform EMD decomposition on the sequence of standardized historical data, and determine the number of decomposed IMF subsequences according to the specific situation of the historical data; analyze the characteristics of the IMF subsequences, remove the IMF subsequences with the highest and lowest frequencies, and retain the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; The decomposed IMF subsequences are segmented into output curves, and the risk level is proposed as an indicator to describe the power quality in power pinch point analysis. In the output prediction stage, the GRU model is used to predict each IMF subsequence separately. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. By superimposing the forecast values of each IMF sub-sequence, the final result of the long-term renewable energy power generation output forecast is obtained.
2. The long-term new energy power-risk prediction method based on EMD-GRU according to claim 1 is characterized in that: The historical data are from the same altitude and the same geographical location coordinates, and the resolution of all data is unified to 1 hour.
3. The long-term new energy power-risk prediction method based on EMD-GRU according to claim 1 is characterized in that: The standardization process is performed according to the following formula: （1） in, For each historical data sample, is the sample mean, is the standard deviation of the sample.
4. The long-term new energy power-risk prediction method based on EMD-GRU according to claim 1 is characterized in that: The EMD decomposition is performed on the sequence of the standardized historical data, and the number of the decomposed IMF subsequences is determined according to the specific situation of the historical data; the characteristics of the IMF subsequences are analyzed, the IMF subsequences with the highest and lowest frequencies are removed, and the remaining IMF subsequences and the residual sequence are retained, including: First find the original signal All the maximum points of , and the maximum envelope is fitted by cubic spline function interpolation method ; Find the original signal All the minimum points of the minimum value are fitted by the cubic spline function interpolation method to obtain the minimum envelope curve ; t is any specific time, calculate the average value of the upper and lower envelopes : （2） Subtract the original signal sequence from We get a new signal without the low frequency. : （3） Repeat the above process, assuming that after k times, If the definition of IMF is satisfied, then the original signal The first-order IMF component of is: （4） With the original signal minus , and get a new signal without high-frequency components : （5） Based on the repetition The second IMF component is obtained by , and so on, until the nth order IMF component or residual Less than the preset value; or when the residual When is a monotonic function or a constant, the EMD decomposition process stops; Finally, the original signal is decomposed by EMD and expressed as: （6） Among them, i is the loop order number, and n is the final number of loops.
5. The long-term new energy power-risk prediction method based on EMD-GRU according to claim 1 is characterized in that: The output curve segmentation of the decomposed IMF subsequence and the risk degree proposed as an indicator to describe the power quality in the power pinch point analysis include: Assume that the wind power output in the system is , photovoltaic output is , the new energy output is , whose upper and lower limits are and ; The new energy output curve is The interval is divided into N intervals, with the power lower limit is the starting point, the risk of the nth interval The expression is as follows: （7） Among them, represents the letter t and its subscript, representing 1 time interval; n is the number of intervals, the number of intervals n < N, and the maximum value that the number of intervals n can take is: （8） The middle value in the interval is taken as the power value of the corresponding risk, and the risk of the nth interval is The corresponding power value as follows: （9）。 6. The long-term new energy power-risk prediction method based on EMD-GRU according to claim 1 is characterized in that: In the output prediction stage, the GRU model is used to predict each IMF subsequence respectively. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point. The output is the prediction results corresponding to each IMF subsequence, including: The update gate determines how much historical information and current information to use to update the hidden state; The update gate at time t is: （10） in, is the gated update signal, represents the implicit state output at the historical time t-1, represents the input data at the current time t, is the sigmoid function, is the weight matrix; The reset gate determines how much historical information is retained. The reset gate at time t is: （11） in, For the reset signal, is the weight matrix; In the update door and reset gate Under the action of , the candidate hidden state at the current moment, the implicit output state Updated to: （12） Among them, the candidate hidden state is: （13） In the above formula, is a candidate hidden state, represents the hyperbolic tangent function, is the weight matrix.
7. A long-term new energy power-risk prediction system based on EMD-GRU, characterized in that: Includes the following modules: The historical data collection module collects historical data of renewable energy power generation output for 24 hours a day for 365 days in a year during the historical data processing stage; A standardization module performs standardization processing on the acquired historical data to obtain a sequence of the standardized historical data; The decomposition module performs EMD decomposition on the sequence of historical data after standardization, and determines the number of IMF subsequences to be decomposed according to the specific situation of the historical data; analyzes the characteristics of the IMF subsequences, removes the IMF subsequences with the highest and lowest frequencies, and retains the remaining IMF subsequences and residual sequences; IMF represents the intrinsic mode function; The indicator proposal module segments the output curve of the decomposed IMF subsequences and proposes the risk level as an indicator to describe the power quality in the power pinch point analysis; The prediction module uses the GRU model to predict each IMF subsequence in the output prediction stage. The input variables are the meteorological factor data of each IMF subsequence and the corresponding time point, and the output is the prediction result corresponding to each IMF subsequence. The result output module obtains the final result of the long-term renewable energy power generation output forecast by superimposing the forecast values of each IMF sub-sequence.
8. The long-term new energy power-risk prediction system based on EMD-GRU according to claim 7 is characterized in that: The historical data in the historical data collection module are from the same altitude and the same geographical location coordinates, and the resolution of all data is unified to 1 hour.
9. The long-term new energy power-risk prediction system based on EMD-GRU according to claim 7 is characterized in that: The standardization processing of the standardization module is performed according to the following formula: （1） in, For each historical data sample, is the sample mean, is the standard deviation of the sample.
10. The long-term new energy power-risk prediction system based on EMD-GRU according to claim 7, characterized in that: The decomposition module implements: First find the original signal All the maximum points of , and the maximum envelope is fitted by cubic spline function interpolation method ; Find the original signal All the minimum points of the minimum value are fitted by the cubic spline function interpolation method to obtain the minimum envelope curve ; t is any specific time, calculate the average value of the upper and lower envelopes : （2） Subtract the original signal sequence from We get a new signal without the low frequency. : （3） Repeat the above process, assuming that after k times, If the definition of IMF is satisfied, then the original signal The first-order IMF component of is: （4） With the original signal minus , and get a new signal without high-frequency components : （5） Based on the repetition The second IMF component is obtained by , and so on, until the nth order IMF component or residual Less than the preset value; or when the residual When is a monotonic function or a constant, the EMD decomposition process stops; Finally, the original signal is decomposed by EMD and expressed as: （6） Among them, i is the loop order number, and n is the final number of loops.
11. The long-term new energy power-risk prediction system based on EMD-GRU according to claim 7, characterized in that: The indicator proposal module includes: Assume that the wind power output in the system is , photovoltaic output is , the new energy output is , whose upper and lower limits are and ; The new energy output curve is The interval is divided into N intervals, with the power lower limit is the starting point, the risk of the nth interval The expression is as follows: （7） Among them, represents the letter t and its subscript, representing 1 time interval; n is the number of intervals, the number of intervals n < N, and the maximum value that the number of intervals n can take is: （8） The middle value in the interval is taken as the power value of the corresponding risk, and the risk of the nth interval is The corresponding power value as follows: （9）。 12. The long-term new energy power-risk prediction system based on EMD-GRU according to claim 7, characterized in that: The prediction module implements: The update gate determines how much historical information and current information to use to update the hidden state; The update gate at time t is: （10） in, is the gated update signal, represents the implicit state output at the historical time t-1, represents the input data at the current time t, is the sigmoid function, is the weight matrix; The reset gate determines how much historical information is retained. The reset gate at time t is: （11） in, For the reset signal, is the weight matrix; In the update door and reset gate Under the action of , the candidate hidden state at the current moment, the implicit output state Updated to: （12） Among them, the candidate hidden state is: （13） In the above formula, is a candidate hidden state, represents the hyperbolic tangent function, is the weight matrix.
13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of a long-period new energy power-risk prediction method based on EMD-GRU as described in any one of claims 1-6 are implemented.
14. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a long-period new energy power-risk prediction method based on EMD-GRU as described in any one of claims 1 to 6 are implemented.