New energy power generation scene generation method and system considering uncertainty

Abstract

The invention discloses a new energy power generation scene generation method and system considering uncertainty, and relates to the technical field of new energy power scene generation, and the method comprises the steps: carrying out the modeling through combining the error fluctuation of Bayesian optimization and kernel density estimation; modeling is carried out based on the error time sequence of the Markov chain; and based on a new energy prediction error modeling process, generating a new energy power sequence scene, and analyzing prediction error characteristics. According to the new energy power generation scene generation method considering uncertainty provided by the invention, the output interval of each prediction box is determined through the Bayesian optimization method, the utilization efficiency of data in the prediction boxes is improved, the prediction error in each prediction box is fitted through the kernel density estimation method, the adaptability of the model is enhanced, and the generation efficiency of the new energy power generation scene is improved. The time sequence correlation of the prediction error is modeled through the Markov chain, the time sequence consistency of the generated scene is enhanced, and the method achieves better effects in the aspects of utilization efficiency, adaptability and time sequence consistency.

Description

Technical Field

[0001] The present invention relates to the technical field of renewable energy power scene generation, and in particular to a renewable energy power generation scene generation method taking uncertainty into consideration. Background Art

[0002] The output of renewable energy is highly volatile and intermittent, and its changing trend is often inversely regulated with the load curve. With the rapid growth of renewable energy power generation, it has a significant impact on the planning, operation, and dispatch of power grids. According to the scenario analysis method in stochastic programming theory, multiple renewable energy power sequence scenarios with a high degree of consistency with the original sequence in time domain characteristics are generated to describe the uncertainty of renewable energy output, which is of great significance to the stochastic planning and day-ahead dispatch of power grids.

[0003] The generation of new energy sequence scenarios usually adopts statistical modeling methods, and the more commonly used ones are time series analysis and probability density function methods. The time series analysis method simulates the new energy power sequence by establishing a model that can describe the relationship between a variable and its past value. The probability density function method models the probability density distribution based on historical data and combines the sampling method to obtain the new energy output sequence. With the rapid development of artificial intelligence technology, the generative network in the deep learning method can effectively fit the distribution of data, providing a new idea for scenario generation.

[0004] In summary, the research on new energy sequence scenario generation technology has achieved certain results, but there are still some problems: the statistical characteristics of new energy prediction errors can reflect the distribution of prediction uncertainty, and its distribution estimation can provide richer information than new energy output, but in statistical methods, most studies model new energy output, and the research on new energy prediction error modeling methods is not in-depth, and the modeling accuracy needs to be improved. Existing deep learning methods cannot effectively explain the relationship between their model structure and new energy uncertainty, resulting in inaccurate modeling. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problems solved by the present invention are: the existing cloud service platform calculation method has insufficient modeling accuracy, and how to effectively explain the optimization problem of the relationship between the model structure and the uncertainty of new energy.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a method for generating new energy power generation scenarios taking into account uncertainty, including modeling the error volatility combining Bayesian optimization and kernel density estimation; modeling based on the error timing of the Markov chain; generating new energy power series scenarios based on the new energy prediction error modeling process, and analyzing the prediction error characteristics.

[0008] As a preferred solution of the method for generating a new energy power generation scenario considering uncertainty described in the present invention, wherein: the modeling of the error volatility combined with Bayesian optimization and kernel density estimation includes fitting the prediction error distribution considering the predicted output. Since there are differences in the probability distribution of prediction errors of different predicted outputs, when fitting the prediction error, the predicted output size is considered, and the predicted output and actual output data at the same time in the historical data are composed of data pairs, and the prediction error of each data pair is calculated. The data pairs are arranged from small to large according to the value of the predicted output, and divided into N groups, each group is a prediction box, and the prediction error data in a single prediction box is fitted to obtain the probability distribution of the prediction error when the predicted output is within any interval. The prediction error calculation is expressed as:

[0009]

[0010] Among them, E i is the prediction error at time i, P Pi is the actual output at time i, P Mi The prediction contribution for time i, C i is the startup capacity at time i. When fitting the prediction error in the prediction box, due to the different probability distributions of prediction errors of different prediction outputs, the prediction errors in different prediction boxes are fitted by the kernel density estimation method. Let e1, e2, ..., e n is a random sample of prediction errors in any prediction box, let the probability density function of the random variable E be f(e), and the kernel density estimate is expressed as:

[0011]

[0012] Among them, n is the sample size, h is the bandwidth, K() is the kernel function, and e k is the kth error sample. In the kernel density estimation, the kernel function and bandwidth are the selected hyperparameters. The Gaussian kernel function is used as the probability density estimation model of the prediction error in the prediction box. The kernel density estimation is rewritten as:

[0013]

[0014] Combined with the optimal bandwidth calculation method, the optimal bandwidth is expressed as:

[0015]

[0016] Among them, σ is the standard deviation of the sample variable.

[0017] As a preferred solution of the method for generating a new energy power generation scenario considering uncertainty described in the present invention, wherein: the modeling of the error volatility combined with Bayesian optimization and kernel density estimation also includes Bayesian optimization parameter adjustment. In each iteration, Bayesian optimization infers the area where the most likely optimal value of the function is located by establishing a posterior probability model of the objective function, samples the area to obtain a new function value, and updates the posterior probability model. The prediction box is divided based on Bayesian optimization. The objective function is the average value of the determination coefficient of all prediction boxes. The hyperparameter that needs to be adjusted is the predicted output range covered by each prediction box, which is expressed as:

[0018] z * = arg max g(z)

[0019] Among them, z is the hyperparameter vector, g(z) is the objective function, and z * It is the optimization result of the hyperparameter. When establishing the prediction box, the coefficient R is determined. 2 As an indicator, the goodness of fit of each prediction box is evaluated, R 2 The closer it is to 1, the higher the goodness of fit. The calculation formula is:

[0020]

[0021] in, is the fitted value of the kth error sample, is the average value of all error samples.

[0022] As a preferred solution of the method for generating a new energy power generation scenario considering uncertainty described in the present invention, wherein: the modeling based on the error temporal property of the Markov chain includes defining a discrete state of the prediction error. If the maximum value in the historical data of the prediction error is E max , the number of states is M, then the definition falls within the range The prediction error state between is j (j = 1, 2, ..., M), and the Markov chain assumes that the probability of state transition at any time depends only on the previous state. If the prediction error sequence state is represented as X1, X2, ..., X i-1 ,X i ,..., then the state X at time i i The state conditional probability depends only on the state X at time i-1 i-1 , expressed as:

[0023] P(X i |X1,X2,...,X i-1 )=P(X i |X i-1 )

[0024] Among them, P(·|·) is the conditional transition probability.

[0025] As a preferred solution of the method for generating a new energy power generation scenario considering uncertainty described in the present invention, wherein: the modeling based on the error temporal property of the Markov chain also includes the Markov chain describing the change law of the prediction error between different moments through the state transfer matrix and the cumulative state transfer matrix, the state transfer matrix describes the probability of the prediction error state transfer at adjacent moments, all the prediction error state transfer probabilities in the matrix are obtained through historical data statistics, the elements in the cumulative state transfer matrix are obtained by summing up the elements in the state transfer matrix, and the state transfer matrix and the cumulative state transfer matrix are expressed as:

[0026]

[0027]

[0028] Among them, p MM is the probability of the prediction error transferring from any state to another state.

[0029] As a preferred solution of the method for generating a new energy power generation scenario taking into account uncertainty described in the present invention, wherein: the generating a new energy power sequence scenario includes setting the number of prediction boxes according to the historical new energy predicted output-actual output data, and determining the predicted output interval of each prediction box through Bayesian optimization, constructing a prediction box, and fitting the probability distribution of the prediction error in each prediction box one by one through kernel density estimation, and obtaining the prediction box at each moment and the probability distribution of the prediction error at each moment according to the new energy output prediction sequence, defining the discrete state of the prediction error, and calculating the state transfer matrix and the cumulative state transfer matrix of the prediction error according to the historical new energy prediction error data, starting to generate a prediction error simulation sequence, and performing a simulation on the initial time. The probability distribution function of the prediction error at the moment is sampled to obtain the value of the prediction error simulation sequence at the initial moment, and the state of the prediction error at this moment is obtained. According to the cumulative state transfer matrix of the prediction error, the state of the prediction error simulation sequence at the next moment is sampled, and the probability distribution of the prediction error at the next moment is sampled, and it is determined whether the state of the sampling result is consistent with the obtained state. If it is consistent, the sampling result is used as the value of the prediction error simulation sequence at the next moment. If it is not consistent, re-sampling is performed. If the preset number of time periods is reached, the prediction error simulation sequence generation is terminated. Otherwise, the state of the prediction error simulation sequence at the next moment is sampled, and the prediction error simulation sequence is superimposed with the predicted output sequence to obtain the new energy power sequence scenario.

[0030] As a preferred solution of the method for generating a new energy power generation scenario considering uncertainty described in the present invention, wherein: the analysis of the prediction error characteristics includes that the prediction error has sequentiality and correlation in time, and the prediction error at any moment has an impact on the prediction error at future moments. By analyzing the autocorrelation coefficient of the daily prediction error sequence, the time series characteristics of the prediction error are expressed. For the time series {x1, x2, ..., x n}, the autocorrelation coefficient with a lag period of l is expressed as:

[0031]

[0032] Among them, X t is a time series {x1,...,x n-l}, X t+l For the time series {x n-l ,...,x n}, cov is the covariance, and D is the variance.

[0033] Another object of the present invention is to provide a new energy power generation scenario generation system that takes uncertainty into account. The system can generate new energy power sequence scenarios based on the modeling process of new energy prediction errors through a scenario generation module, combining error volatility and timing, thereby solving the current problem of insufficient simulation of complex output changes when generating power scenarios.

[0034] As a preferred solution of the renewable energy power generation scenario generation system considering uncertainty described in the present invention, it includes: an error volatility module, an error timing module, and a scenario generation module; the error volatility module is used to model the error volatility combining Bayesian optimization and kernel density estimation; the error timing module is used to model the error timing based on the Markov chain; the scenario generation module is used to generate a new energy power series scenario based on the new energy prediction error modeling process, and analyze the prediction error characteristics.

[0035] A computer device includes a memory and a processor, wherein the memory stores a computer program, and wherein the processor executes the computer program to implement a step of a method for generating a new energy power generation scenario taking into account uncertainty.

[0036] A 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 method for generating a new energy power generation scenario taking into account uncertainty are implemented.

[0037] Beneficial effects of the present invention: The method for generating a new energy power generation scenario taking uncertainty into consideration provided by the present invention determines the output range of each prediction box through a Bayesian optimization method, thereby improving the utilization efficiency of data in the prediction box, and fitting the prediction error in each prediction box through a kernel density estimation method, thereby improving the description capability of the uncertainty of the new energy output and enhancing the adaptability of the model. The time series correlation of the prediction error is modeled through a Markov chain, thereby enhancing the time series consistency of the generated scenario, and generating a complete new energy power sequence scenario by simulating the error at each moment, thereby improving the adaptability of the power grid to fluctuations in the output of new energy. The present invention achieves better results in terms of utilization efficiency, adaptability and time series consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0039] Figure 1 An overall flow chart of a method for generating a renewable energy power generation scenario taking uncertainty into consideration provided for the first embodiment of the present invention.

[0040] Figure 2 A schematic diagram of prediction error sequence modeling for a method for generating a renewable energy power generation scenario taking uncertainty into account provided in a second embodiment of the present invention.

[0041] Figure 3 A distribution diagram of wind power prediction errors under different prediction value intervals of a new energy power generation scenario generation method considering uncertainty provided in the second embodiment of the present invention.

[0042] Figure 4 A box plot of the prediction error autocorrelation coefficient of a method for generating a new energy power generation scenario taking into account uncertainty provided in the second embodiment of the present invention.

[0043] Figure 5 A wind power sequence scenario set of a method for generating a renewable energy power generation scenario taking uncertainty into consideration is provided in accordance with a second embodiment of the present invention.

[0044] Figure 6 A photovoltaic power sequence scenario set of a method for generating a new energy power generation scenario taking uncertainty into consideration is provided in the second embodiment of the present invention.

[0045] Figure 7 An overall module diagram of a new energy power generation scenario generation system taking uncertainty into consideration provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] Example 1

[0048] Reference Figure 1 , which is an embodiment of the present invention, provides a method for generating a new energy power generation scenario taking uncertainty into consideration, comprising:

[0049] S1: Combining Bayesian optimization and kernel density estimation for error volatility modeling.

[0050] Furthermore, the error volatility of Bayesian optimization and kernel density estimation is combined to model the prediction error distribution fitting considering the predicted output.

[0051] It should be noted that due to the differences in the probability distribution of prediction errors for different predicted outputs, when fitting the prediction error, the predicted output size is taken into account, and the predicted output and actual output data at the same time in the historical data are combined into data pairs. The prediction error of each data pair is calculated, and the data pairs are arranged from small to large according to the value of the predicted output, and divided into N groups, each group is a prediction box, and the prediction error data in a single prediction box are fitted to obtain the probability distribution of the prediction error when the predicted output is within any interval. The prediction error calculation is expressed as:

[0052]

[0053] Among them, E i is the prediction error at time i, P Pi is the actual output at time i, P Mi The prediction contribution for time i, C i is the startup capacity at time i. When fitting the prediction error in the prediction box, due to the different probability distributions of prediction errors of different prediction outputs, the prediction errors in different prediction boxes are fitted by the kernel density estimation method. Let e1, e2, ..., e n is a random sample of prediction errors in any prediction box, let the probability density function of the random variable E be f(e), and the kernel density estimate is expressed as:

[0054]

[0055] Among them, n is the sample size, h is the bandwidth, K() is the kernel function, and e k is the kth error sample. In the kernel density estimation, the kernel function and bandwidth are the selected hyperparameters. The Gaussian kernel function is used as the probability density estimation model of the prediction error in the prediction box. The kernel density estimation is rewritten as:

[0056]

[0057] Combined with the optimal bandwidth calculation method, the optimal bandwidth is expressed as:

[0058]

[0059] Among them, σ is the standard deviation of the sample variable.

[0060] Furthermore, modeling based on the error volatility of Bayesian optimization and kernel density estimation also includes Bayesian optimization parameter adjustment.

[0061] It should be noted that in each iteration of Bayesian optimization, the most likely region where the optimal value of the function is located is inferred by establishing the posterior probability model of the objective function, sampling is performed in the region to obtain new function values, and the posterior probability model is updated. The prediction box is divided based on Bayesian optimization. The objective function is the average value of the determination coefficients of all prediction boxes. The hyperparameters that need to be adjusted are the predicted output range covered by each prediction box, which is expressed as:

[0062] z * = arg max g(z)

[0063] Among them, z is the hyperparameter vector, g(z) is the objective function, and z * It is the optimization result of the hyperparameter. When establishing the prediction box, the coefficient R is determined. 2 As an indicator, the goodness of fit of each prediction box is evaluated, R 2 The closer it is to 1, the higher the goodness of fit. The calculation formula is:

[0064]

[0065] in, is the fitted value of the kth error sample, is the average value of all error samples.

[0066] It should also be noted that the main components of the Bayesian optimization framework include a probabilistic surrogate model and an acquisition function. The probabilistic surrogate model is an estimation model for the objective function. It predicts the value of the objective function at unknown points based on existing sampling points, and gradually approaches the true objective function through dynamic updates during the optimization process. Commonly used probabilistic surrogate models include Gaussian Process (GP) and Random Forest (RF). Since the random forest method has better computational efficiency, the random forest is selected as the model surrogate. The acquisition function determines which points are selected for sampling in each iteration to update the surrogate model and continue to optimize the objective function. Commonly used methods include Probability of Improvement (PI), Expected Improvement (EI), Upper Confidence Bound (UCB), etc. EI is selected as the acquisition function, that is, based on the existing surrogate model, the expectation of the improvement of the objective function at each possible sampling point is calculated, and then the point with the highest expectation is selected as the next sampling point.

[0067] It should also be noted that, currently when establishing prediction boxes, most people adopt the method of dividing the interval into equal parts according to the size of the interval, that is, when grouping the data, the interval covered by the predicted output value is divided into N equal parts, and the N groups of data obtained are N prediction boxes. However, this method has the problem of insufficient data in some prediction boxes, resulting in poor fitting effect.

[0068] S2: Modeling based on the error temporality of the Markov chain.

[0069] Furthermore, modeling the temporal nature of the error based on the Markov chain involves defining discrete states of the prediction error.

[0070] It should be noted that if the maximum value in the historical data of the forecast error is E max , the number of states is M, then the definition falls within the range The prediction error state between is j (j = 1, 2, ..., M), and the Markov chain assumes that the probability of state transition at any time depends only on the previous state. If the prediction error sequence state is represented as X1, X2, ..., X i-1 ,X i ,..., then the state X at time i i The state conditional probability depends only on the state X at time i-1 i-1 , expressed as:

[0071] P(X i|X1,X2,...,X i-1 )=P(X i |X i-1 )

[0072] Among them, P(·|·) is the conditional transition probability.

[0073] Furthermore, modeling based on the error temporality of the Markov chain also includes the Markov chain describing the changing pattern of the prediction error between different moments through a state transfer matrix and a cumulative state transfer matrix.

[0074] It should be noted that the state transfer matrix describes the probability of the prediction error state transition at adjacent moments. All the prediction error state transition probabilities in the matrix are obtained through historical data statistics. The elements in the cumulative state transfer matrix are obtained by summing up the elements in the state transfer matrix. The state transfer matrix and the cumulative state transfer matrix are expressed as:

[0075]

[0076] Among them, p MM is the probability of the prediction error transferring from any state to another state.

[0077] S3: Based on the new energy prediction error modeling process, generate new energy power sequence scenarios and analyze the prediction error characteristics.

[0078] Furthermore, generating a new energy power sequence scenario includes predicting output and actual output data based on historical new energy.

[0079] It should be noted that the number of prediction boxes is set, and the predicted output range of each prediction box is determined through Bayesian optimization. The prediction box is constructed, and the probability distribution of the prediction error in each prediction box is fitted one by one through kernel density estimation. According to the new energy output prediction sequence, the prediction box at each moment and the probability distribution of the prediction error at each moment are obtained, the discrete state of the prediction error is defined, and the state transfer matrix and the cumulative state transfer matrix of the prediction error are calculated according to the historical new energy prediction error data. The prediction error simulation sequence is started to be generated, and the probability distribution function of the prediction error at the initial moment is sampled to obtain the prediction error simulation sequence at the initial moment. value, obtain the state of the prediction error at this moment, sample the state of the prediction error simulation sequence at the next moment according to the cumulative state transfer matrix of the prediction error, sample the probability distribution of the prediction error at the next moment, and judge whether the state of the sampling result is consistent with the obtained state. If it is consistent, the sampling result is used as the value of the prediction error simulation sequence at the next moment. If it is not consistent, re-sample. If the preset number of time periods is reached, the generation of the prediction error simulation sequence is terminated. Otherwise, sample the state of the prediction error simulation sequence at the next moment, superimpose the prediction error simulation sequence with the predicted output sequence, and obtain the new energy power sequence scenario.

[0080] Furthermore, the characteristics of the forecast errors are analyzed including the temporal sequentiality and correlation of the forecast errors.

[0081] It should be noted that the forecast error at any moment has an impact on the forecast error at future moments. By analyzing the autocorrelation coefficient of the daily forecast error sequence, the time series characteristics of the forecast error are expressed. For the time series {x1, x2, ..., x n}, the autocorrelation coefficient with a lag period of l is expressed as:

[0082]

[0083] Among them, X t is a time series {x1,...,x n-l}, X t+l For the time series {x n-l ,...,x n}, cov is the covariance, and D is the variance.

[0084] Example 2

[0085] Reference Figure 2-6 , which is an embodiment of the present invention, provides a method for generating a renewable energy power generation scenario taking uncertainty into consideration. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0086] from Figure 2 It can be seen that the horizontal axis represents the number of time periods, with 24 time periods per day, and the vertical axis represents the prediction error value. Modeling the volatility of the prediction error means taking into account the size of the predicted output and determining the specific value of the prediction error at a certain moment, which is reflected in the figure as the vertical coordinate of each sample point. Modeling the temporal nature of the prediction error means determining the range of the prediction error value at the next moment based on the prediction error value at the previous moment, which is reflected in the figure as the trend of the sample points between each moment.

[0087] The normalized actual wind power output-forecasted output data of a certain place throughout the year is used as the analysis basis. The data collection interval is 1h. There is a correlation between the statistical characteristics of wind power prediction error and the wind power prediction value. Statistics of prediction errors in different wind power prediction value intervals can intuitively reflect this characteristic. The prediction errors in the [0,0.1]pu and [0.3,0.4]pu intervals are selected respectively, and their distribution is statistically analyzed and expressed in a histogram, as shown in Figure 3 As shown in the figure, the distribution of forecast errors is obviously different under different forecast output levels. The autocorrelation coefficient of the daily forecast error sequence in the lag period of 1-6h is calculated, and the calculation results are as follows: Figure 4As shown, it can be seen that the prediction error series has strong autocorrelation within the interval of 1-6 hours, and the prediction errors of adjacent 1 hours have the highest correlation.

[0088] Based on the predicted wind power output data and actual output data of a certain place from January to September 2020, a prediction box is constructed, and the prediction error is modeled. Combined with the predicted wind power output data on October 10, 2020, a wind power sequence scenario set with a scale of 300 and a time period of 24 is generated according to the method of the present invention, as shown in the figure. Figure 5 As shown, the dotted line is the wind power sequence scenario set generated by the method of the present invention. The scenario set shows various possibilities of wind power output. There is no obvious disordered fluctuation in the power sequence, and the actual wind power output sequence is completely within the scenario set range.

[0089] Based on the photovoltaic forecast output data and actual output data of a certain place from January to September 2020, a prediction box is constructed, and the prediction error is modeled. Combined with the photovoltaic forecast output data on October 10, 2020, a photovoltaic power sequence scenario set with a scale of 300 and a time period of 24 is generated according to the method of the present invention, as shown in the figure. Figure 6 As shown, the dotted line is the photovoltaic power sequence scenario set generated by the method of the present invention. The scenario set shows various possibilities of photovoltaic output. There is no obvious disordered fluctuation in the power sequence, and the actual photovoltaic output sequence is completely within the scenario set range.

[0090] Example 3

[0091] Reference Figure 7 , which is an embodiment of the present invention, provides a new energy power generation scenario generation system taking uncertainty into consideration, including: an error volatility module, an error timing module, and a scenario generation module.

[0092] The error volatility module is used to model the error volatility combining Bayesian optimization and kernel density estimation; the error timing module is used to model the error timing based on the Markov chain; the scenario generation module is used to generate new energy power series scenarios based on the new energy prediction error modeling process and analyze the prediction error characteristics.

[0093] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0094] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0095] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0096] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for generating a renewable energy power generation scenario taking uncertainty into account, characterized in that: include: Combining Bayesian optimization and kernel density estimation with error volatility for modeling; Modeling based on the error timing of Markov chains; Based on the new energy prediction error modeling process, new energy power sequence scenarios are generated and the prediction error characteristics are analyzed.

2. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 1, characterized in that: The modeling of error volatility combined with Bayesian optimization and kernel density estimation includes fitting the prediction error distribution considering the predicted output. Since there are differences in the probability distribution of prediction errors of different predicted outputs, when fitting the prediction error, the predicted output size is considered, and the predicted output and actual output data at the same time in the historical data are combined into data pairs, and the prediction error of each data pair is calculated. The data pairs are arranged from small to large according to the value of the predicted output and divided into N groups, each group being a prediction box, and the prediction error data in a single prediction box are fitted to obtain the probability distribution of the prediction error when the predicted output is within any interval. The prediction error calculation is expressed as: Among them, E i is the prediction error at time i, P Pi is the actual output at time i, P Mi The prediction contribution for time i, C i is the startup capacity at time i. When fitting the prediction error in the prediction box, due to the different probability distributions of prediction errors of different prediction outputs, the prediction errors in different prediction boxes are fitted by the kernel density estimation method. Let e1, e2, ..., e n is a random sample of prediction errors in any prediction box, let the probability density function of the random variable E be f(e), and the kernel density estimate is expressed as: Among them, n is the sample size, h is the bandwidth, K() is the kernel function, and e k is the kth error sample. In the kernel density estimation, the kernel function and bandwidth are the selected hyperparameters. The Gaussian kernel function is used as the probability density estimation model of the prediction error in the prediction box. The kernel density estimation is rewritten as: Combined with the optimal bandwidth calculation method, the optimal bandwidth is expressed as: Among them, σ is the standard deviation of the sample variable.

3. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 2, characterized in that: The modeling of error volatility combined with Bayesian optimization and kernel density estimation also includes Bayesian optimization parameter adjustment. In each iteration, Bayesian optimization infers the area where the most likely optimal value of the function is located by establishing a posterior probability model of the objective function, obtains new function values ​​by sampling in the area, and updates the posterior probability model. The prediction box is divided based on Bayesian optimization. The objective function is the average value of the determination coefficients of all prediction boxes. The hyperparameter that needs to be adjusted is the prediction output range covered by each prediction box, which is expressed as: With * =arg max g(z) Among them, z is the hyperparameter vector, g(z) is the objective function, and z * It is the optimization result of the hyperparameter. When establishing the prediction box, the coefficient R is determined. 2 As an indicator, the goodness of fit of each prediction box is evaluated, R 2 The closer it is to 1, the higher the goodness of fit. The calculation formula is: in, is the fitted value of the kth error sample, is the average value of all error samples.

4. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 3, characterized in that: The Markov chain-based error temporal modeling includes defining the discrete state of the prediction error. If the maximum value in the historical data of the prediction error is E max , the number of states is M, then the definition falls within the range The prediction error state between is j (j = 1, 2, ..., M), and the Markov chain assumes that the probability of state transition at any time depends only on the previous state. If the prediction error sequence state is represented as X1, X2, ..., X i-1 ,X i ,..., then the state conditional probability of state Xi at time i depends only on the state X at time i-1 i-1 , expressed as: P(X i |X1,X2,…,X i-1 )=P(X i |X i-1 ) Among them, P(·|·) is the conditional transition probability.

5. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 4, characterized in that: The error temporal modeling based on the Markov chain also includes that the Markov chain describes the change law of the prediction error between different moments through the state transfer matrix and the cumulative state transfer matrix, the state transfer matrix describes the probability of the prediction error state transfer at adjacent moments, all the prediction error state transfer probabilities in the matrix are obtained through historical data statistics, the elements in the cumulative state transfer matrix are obtained by accumulating the elements in the state transfer matrix, and the state transfer matrix and the cumulative state transfer matrix are expressed as: Among them, p MM is the probability of the prediction error transferring from any state to another state.

6. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 5, characterized in that: The generating scenario of the new energy power sequence includes setting the number of prediction boxes according to the historical new energy predicted output-actual output data, determining the predicted output interval of each prediction box through Bayesian optimization, constructing the prediction box, fitting the probability distribution of the prediction error in each prediction box one by one through kernel density estimation, obtaining the prediction box at each moment and the probability distribution of the prediction error at each moment according to the new energy output prediction sequence, defining the discrete state of the prediction error, and calculating the state transfer matrix and the cumulative state transfer matrix of the prediction error according to the historical new energy prediction error data, starting to generate the prediction error simulation sequence, sampling the probability distribution function of the prediction error at the initial moment, and obtaining The value of the prediction error simulation sequence at the initial moment is obtained, the state of the prediction error at this moment is obtained, and the state of the prediction error simulation sequence at the next moment is sampled according to the cumulative state transfer matrix of the prediction error. The probability distribution of the prediction error at the next moment is sampled, and it is determined whether the state of the sampling result is consistent with the obtained state. If it is consistent, the sampling result is used as the value of the prediction error simulation sequence at the next moment. If it is not consistent, re-sampling is performed. If the preset number of time periods is reached, the generation of the prediction error simulation sequence is terminated. Otherwise, the state of the prediction error simulation sequence at the next moment is sampled, and the prediction error simulation sequence is superimposed with the predicted output sequence to obtain the new energy power sequence scenario.

7. The method for generating a renewable energy power generation scenario taking uncertainty into consideration according to claim 6, characterized in that: The analysis of the forecast error characteristics includes that the forecast error has temporal order and correlation, and the forecast error at any moment has an impact on the forecast error at future moments. By analyzing the autocorrelation coefficient of the daily forecast error sequence, the time series characteristics of the forecast error are expressed. For the time series {x1, x2, ..., x n }, the autocorrelation coefficient with a lag period of l is expressed as: Among them, X t is a time series {x1,...,x n-l }, X t+l For the time series {x n-l ,...,x n }, cov is the covariance, and D is the variance.

8. A system using the method for generating a new energy power generation scenario taking into account uncertainty as described in any one of claims 1 to 7, characterized in that: Including error volatility module, error timing module, and scenario generation module; The error volatility module is used to model the error volatility by combining Bayesian optimization and kernel density estimation; The error timing module is used to model the error timing based on the Markov chain; The scenario generation module is used to generate new energy power sequence scenarios and analyze prediction error characteristics based on the new energy prediction error modeling process.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for generating a new energy power generation scenario taking into account uncertainty described in any one of claims 1 to 7 are implemented.

10. A 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 the method for generating a new energy power generation scenario taking into account uncertainty described in any one of claims 1 to 7 are implemented.

Images