Parallel computing-based new energy power generation capacity prediction method and system
Through parallel computing and data fusion technology, ARIMA and BSTS models are constructed, which solves the problem of insufficient accuracy and adaptability of the new energy power generation prediction method when processing complex data, and achieves efficient and robust new energy power generation prediction.
Patent Information
- Application Number
- CN202410793741.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-19
- Publication Date
- 2024-11-15
AI Technical Summary
When the existing new energy power generation prediction method processes complex, nonlinear and multivariate data, the prediction accuracy is poor, the adaptability is insufficient, the calculation efficiency is low, and the robustness is poor for outliers and noise data, and the real-time prediction and dynamic update capabilities are weak. .
Using a method based on parallel computing, multi-source heterogeneous data is collected for preprocessing and time series feature extraction, ARIMA and BSTS models are constructed, and the two are fused through the weighted average method to generate a fusion model for prediction, and the Hadoop big data framework is used for Parallel calculations are performed to improve prediction speed and accuracy.
It significantly improves the accuracy and reliability of new energy power generation prediction, enhances the robustness of outliers and noise data, realizes real-time prediction and dynamic update capabilities, and ensures the efficient and stable operation of new energy power generation systems.
Smart Images

Figure CN118970869A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of renewable energy power generation prediction, and in particular to a method and system for predicting renewable energy power generation capacity based on parallel computing. Background Art
[0002] Wind power and photovoltaic power are clean and renewable energy sources with abundant total reserves. However, their power output is easily affected by the natural environment and fluctuates violently, which poses a severe challenge to the development of large-scale new energy power stations.
[0003] In order to maintain the stable operation of the power grid, when there is an excess of renewable energy, the grid dispatcher needs to abandon part of the wind and solar grid-connected power; conversely, when there is insufficient renewable energy, other energy sources are needed to supplement. This not only leads to energy waste, but also aggravates the difficulty of grid dispatching. Therefore, it is urgent to improve the research and technology development of renewable energy power prediction, which is particularly critical to ensure the stable grid connection of renewable energy.
[0004] At present, existing forecasting methods have significant defects in processing complex, nonlinear and multivariate data, lack the ability to capture nonlinear complex relationships, and have poor forecasting accuracy. At the same time, these methods have poor adaptability to seasonal trend changes and are mainly suitable for single-variable forecasting. When processing large-scale multivariate data, the computational efficiency is low. In addition, existing methods lack robustness to outliers and noise data, easily introduce forecasting bias, and have weak real-time forecasting and dynamic updating capabilities. Summary of the invention
[0005] In view of the problems existing in the existing new energy power generation prediction method, the present invention is proposed.
[0006] Therefore, the problem to be solved by the present invention is how to effectively process complex, nonlinear and multivariate data, improve prediction accuracy and adaptability, while enhancing the robustness to outliers and noise data, and improve real-time prediction and dynamic update capabilities.
[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0008] In the first aspect, an embodiment of the present invention provides a method for predicting renewable energy power generation capacity based on parallel computing, which includes collecting multi-source heterogeneous data of renewable energy power generation, and preprocessing and extracting time series features; constructing an ARIMA model and a BSTS model based on the time series features; and using a weighted average method to fuse the ARIMA model and the BSTS model to obtain a fusion model and predict power generation capacity.
[0009] As a preferred solution of the parallel computing method for predicting new energy power generation capacity described in the present invention, the multi-source heterogeneous data includes historical power generation data, weather data and external variable data; the construction process of the ARIMA model includes the following steps: testing the stationarity of the time series data through ADF, drawing the autocorrelation function ACF diagram and the partial autocorrelation function PACF diagram of the time series data respectively, and determining the values of the difference order d, the autoregressive term p and the moving average term q; inputting the values of the difference order d, the autoregressive term p and the moving average term q into the ARIMA model to fit the training data; observing the autocorrelation function ACF diagram and the partial autocorrelation function PACF diagram of the residual sequence, and using the AIC information criterion to evaluate the ARIMA model of different order combinations; predicting future time points through the evaluated ARIMA model to obtain predicted values and corresponding confidence intervals.
[0010] As a preferred solution of the parallel computing new energy power generation capacity prediction method of the present invention, the relevant formula of the ADF test time series data is as follows:
[0011] Δy t =α+βt+γy t-1 +δ1Δy t-1 +...+δ p Δy t-p +ε t
[0012] Among them, Δy t-p is the first-order difference of the time series, α is the constant term, βt is the time trend term, γ is the autoregressive coefficient, δ p is the coefficient of the difference term p, ε t is the error term.
[0013] The test results of the ADF test time series data are used to determine the stationarity of the data; when the first-order difference value is less than the set threshold, the data is considered to be stationary; when the first-order difference value is greater than or equal to the set threshold, the data is considered to be disordered; if the data is disordered, the data is stabilized by difference, and this operation is repeated until the series is stable.
[0014] As a preferred solution of the parallel computing new energy power generation capacity prediction method of the present invention, it also includes that the relevant formula of the observed residual sequence is as follows:
[0015]
[0016] Among them, ρ t is the time series y t and the correlation with the lagged values, n is the number of lag orders, y t is the observed value of the time series, is the lagged value of the time series.
[0017] If the ACF of the observed residual sequence is truncated to 0 after a certain lag order, then this lag order is the moving average term q; if the ACF of the observed residual sequence is truncated to 1 after a certain lag order, then this lag order is the autoregressive term p; the relevant formula of the ARIMA model is as follows:
[0018] y t =d+φ1y t-1 +...+φ p y t-p +θ1ε t-1 +...+θ q ε t-q +ε t
[0019] Among them, y t is the observed value of the time series, φ p is the autoregressive coefficient, θ q is the moving average coefficient, p is the autoregressive term, q is the moving average term, is the lag value of the time series, and d is the number of differences.
[0020] As a preferred solution of the parallel computing new energy power generation capacity prediction method described in the present invention, the construction step of the BSTS model includes the following steps: determining the trend type, trend cycle, external prediction variables and regression terms, and inputting them into the BSTS model, and fitting the regression terms using the Bayesian regression method; testing the BSTS model parameters and checking the fitting residuals; using the Bayesian inference method to estimate the parameters, process the uncertainty in the data, and obtain parameter estimates; comparing the fitting effects of the BSTS models under different order combinations, selecting the BSTS model with the best fitting effect, and using cross-validation to evaluate the model prediction accuracy; predicting future time points through the evaluated BSTS model, and giving predicted values and confidence intervals.
[0021] As a preferred solution of the parallel computing new energy power generation capacity prediction method described in the present invention, the ARIMA model and the BSTS model are fused to generate a fusion model through a weighted average method; the construction process of the fusion model includes the following steps: using the Hadoop big data framework to allocate computing resources and set up a parallel computing environment; dividing data preprocessing, feature extraction, model training and prediction tasks into several subtasks; training the ARIMA model and the BSTS model separately in a parallel environment, and optimizing model parameters using historical data; obtaining parameter estimates through the ARIMA model and the BSTS model respectively, and using the ARIMA model and the BSTS model in Predict at a future time point to obtain the prediction result; use the weighted average method to weight the prediction results of the ARIMA model and the BSTS model to obtain a fusion model, and use the prediction result obtained by weighted average as the prediction output of the fusion model; predict new energy power generation through the fusion model; obtain real-time multi-source data and input it into the fusion model for real-time prediction; use the parallel computing framework to perform parallel computing of several tasks to improve the prediction speed; provide a confidence interval for the prediction result through Bayesian inference to quantify uncertainty; dynamically adjust the power generation plan by judging the prediction results and confidence intervals of the fusion model to improve the efficiency of new energy utilization; the specific formula of the fusion model is as follows:
[0022]
[0023] in, is the prediction result of the fusion model, ω ARIMA is the weight of the ARIMA model, is the prediction result of the ARIMA model, ω BSTS is the weight of the BSTS model, This is the prediction result of the BSTS model.
[0024] As a preferred solution of the parallel computing new energy power generation capacity prediction method described in the present invention, the prediction result of the fusion model includes: if the prediction result of the fusion model is greater than the observed value, and the upper confidence bound is much greater than the lower confidence bound, it is expected that the new energy supply is sufficient and the power generation plan is increased; if the prediction result of the fusion model is less than the observed value, and the lower confidence bound is much lower than the upper confidence bound, it is expected that the new energy supply is insufficient and a corresponding supplementary plan is formulated; if the gap between the prediction result of the fusion model and the observed value is within the confidence interval and the confidence interval is narrow, the prediction result is accurate and the power generation plan is arranged according to the prediction result of the fusion model; if the gap between the prediction result of the fusion model and the observed value is outside the confidence interval and the confidence interval is wide, there is uncertainty in the prediction, the actual situation should be closely monitored and the power generation plan should be adjusted in time; if in a continuous prediction cycle, the actual value continues to deviate from the predicted value by an amount less than the predicted amount, the ARIMA model and the BSTS model are re-fused to optimize the fusion model structure.
[0025] In the second aspect, an embodiment of the present invention provides a new energy power generation capacity prediction system based on parallel computing, which includes: an extraction module for collecting multi-source heterogeneous data of new energy power generation, and preprocessing and extracting time series features; a construction module for constructing an ARIMA model and a BSTS model based on time series features; a prediction module for fusing the ARIMA model and the BSTS model using a weighted average method, obtaining a fusion model and performing real-time power generation capacity prediction.
[0026] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of the parallel computing new energy power generation capacity prediction method as described in the first aspect of the present invention are implemented.
[0027] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of the parallel computing method for predicting new energy power generation capacity as described in the first aspect of the present invention are implemented.
[0028] The beneficial effects of the present invention are as follows: the present invention ensures the integrity and consistency of data through data preprocessing and time series feature extraction; constructs ARIMA and BSTS models, fully utilizes their respective advantages, and realizes multi-angle analysis of time series data; and finally fuses the two models through a weighted average method, significantly improving the accuracy and reliability of power generation capacity prediction, providing robust prediction results, realizing efficient and stable operation of new energy power generation systems, and improving economic and social benefits. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] 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 work. Among them:
[0030] Figure 1 This is a flow chart of the new energy power generation capacity prediction method using parallel calculation in Example 1. DETAILED DESCRIPTION
[0031] In order to make the above-mentioned objects, 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 accompanying drawings.
[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0033] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0034] Example 1
[0035] Reference Figure 1 , which is the first embodiment of the present invention, and provides a new energy power generation capacity prediction method based on parallel computing, including:
[0036] S1: Collect multi-source heterogeneous data of renewable energy power generation, preprocess it and extract time series features.
[0037] Specifically, by systematically collecting historical power generation data from renewable energy power plants, energy management systems, and public energy databases, as well as weather data from meteorological agencies, weather forecast websites, and satellite data providers, and obtaining seasonal factors and holiday information, the comprehensiveness and high quality of the data are ensured. Through API calls, data integration, and storage, the data is synchronized and integrated, and the missing values and outliers are processed using linear interpolation methods to ensure the integrity and consistency of the data; trends and seasonality in time series are eliminated to ensure the stability of the data, and the data is standardized or normalized to provide a solid foundation for building accurate and reliable time series prediction models. Multi-source heterogeneous data includes historical power generation data, weather data, and external variable data.
[0038] Furthermore, the linear trend is eliminated by first-order difference, the quadratic parabola trend is eliminated by second-order difference, and the seasonal difference is used to eliminate the influence of seasonal cycles. The function-related formula of the first-order difference is as follows:
[0039] y' t =x t -x t-1
[0040] Among them, y t ' is the first-order difference data at time t, x t is the original data at time t.
[0041] Furthermore, the function-related formula of the second-order difference is as follows:
[0042] y" t =y' t -y' t-1
[0043] y" t =(x t -x t-1 )-(x t-1 -x t-2 )
[0044] y" t =x t -2x t-1 +x t-2
[0045] Among them, y t is the second-order difference data at time t, y' t is the first-order difference data at time t, x t is the original data at time t.
[0046] Specifically, the function-related formula of seasonal difference is as follows:
[0047]
[0048] in, is the difference data of seasonal cycle length s at time t, z t is the seasonal difference raw data at time t.
[0049] Specifically, the average value of the data in the window is calculated, the time series is smoothed, different weights are given to historical data, and newer data has a higher weight. The local polynomial fitting smoothes the time series. By calculating the average value of the data in the window, the time series is smoothed and the noise is reduced. The specific formula is as follows:
[0050]
[0051] Among them, MA t is the moving average at time t, y t-i is the original data at time ti, and n is the size of the moving window.
[0052] Furthermore, the time series is smoothed by local polynomial fitting, which is suitable for more complex trends and seasonal changes; the Pearson correlation coefficient between each extracted time series feature and the target variable is calculated, and the feature with a higher correlation with the target variable is selected according to the size of the correlation coefficient. The specific formula of the correlation coefficient is as follows:
[0053]
[0054] Among them, r is the correlation coefficient, m is the number of samples, and x j is the characteristic variable value of the jth sample, z j is the target variable value of the jth sample, is the mean of the characteristic variable values, is the 2-mean of the target variable value.
[0055] S2: Based on the time series characteristics, construct ARIMA model and BSTS model.
[0056] Specifically, the construction process of the ARIMA model includes the following steps: test the stationarity of the time series data through ADF, draw the autocorrelation function ACF diagram and partial autocorrelation function PACF diagram of the time series data, and determine the values of the difference times d, autoregressive term p and moving average term q; the relevant formula for ADF test time series data is as follows:
[0057] Δy t =α+βt+γy t-1 +δ1Δy t-1 +...+δ p Δy t-p +ε t
[0058] Among them, Δy t-p is the first-order difference of the time series, α is the constant term, βt is the time trend term, γ is the autoregressive coefficient, δ p is the coefficient of the difference term p, ε t is the error term;
[0059] Furthermore, the test results of the ADF test time series data are used to determine the stationarity of the data; when the first-order difference value is less than the set threshold, the data is determined to be stationary; when the first-order difference value is greater than or equal to the set threshold, the data is determined to be disordered; if the data is disordered, the data is stabilized by difference, and this operation is repeated until the series is stable.
[0060] Furthermore, the values of the number of differences d, the autoregressive term p, and the moving average term q are input into the ARIMA model to fit the training data. The relevant formula for the degree of fit is as follows:
[0061] AIC=2k-2ln(L)
[0062] Where k is the number of ARIMA model parameters and ln(L) is the likelihood term of the ARIMA model.
[0063] Specifically, by observing the autocorrelation function ACF diagram and partial autocorrelation function PACF diagram of the residual sequence, and using the AIC information criterion to evaluate the ARIMA model of different order combinations; the relevant formula for observing the residual sequence is as follows:
[0064]
[0065] Among them, ρ t is the time series y t and the correlation with the lagged values, n is the number of lag orders, y t is the observed value of the time series, is the lagged value of the time series.
[0066] Furthermore, if the observed residual sequence ACF is truncated to 0 after a certain lag order, then this lag order is the moving average term q; if the observed residual sequence ACF is truncated to 1 after a certain lag order, then this lag order is the autoregressive term p.
[0067] It should be noted that the ADF test shows that the p value is 0.1 (greater than 0.05), and the data is not stable. After a difference is performed, the ADF test is performed again, and the p value is 0.01 (less than 0.05), and the data is stable, so d = 1; observe the data after a difference, the ACF graph is truncated at lag 2, and the PACF graph is truncated at lag 1, so p = 1, q = 2; ARIMA (1,1,2) model fits the data to obtain the fitting results. Observe the ACF and PACF graphs of the residuals to confirm that the residuals have no significant autocorrelation. Calculate AIC and BIC, AIC = 320.5, BIC = 330.2; Use the fitted ARIMA (1,1,2) model to predict the new energy power generation in the next 12 months, and obtain the predicted value and 95% confidence interval.
[0068] Furthermore, the ARIMA model is evaluated to predict future time points and obtain the predicted value and the corresponding confidence interval. The relevant formula of the ARIMA model is as follows:
[0069] y t =d+φ1y t-1 +...+φ p y t-p +θ1εt-1 +...+θ q ε t-q +ε t
[0070] Among them, y t is the observed value of the time series, φ p is the autoregressive coefficient, θ q is the moving average coefficient, p is the autoregressive term, q is the moving average term, is the lag value of the time series, and d is the number of differences.
[0071] Specifically, the construction steps of the BSTS model include the following steps: determine the trend type, trend cycle, external prediction variables and regression terms, and input them into the BSTS model, and use the Bayesian regression method to fit the regression terms; test the BSTS model parameters and check the fitting residuals; use the Bayesian inference method to estimate the parameters, deal with the uncertainty in the data, and obtain the parameter estimates; compare the fitting effects of the BSTS model under different order combinations, select the BSTS model with the best fitting effect, and use cross-validation to evaluate the model prediction accuracy; use the evaluated BSTS model to predict future time points and give the prediction value and confidence interval.
[0072] S3: The ARIMA model and the BSTS model are fused using the weighted average method to obtain the fusion model and perform real-time power generation capacity prediction.
[0073] Furthermore, the ARIMA model and the BSTS model are fused to generate a fusion model through a weighted average method; the construction process of the fusion model includes the following steps: using the Hadoop big data framework to allocate computing resources and set up a parallel computing environment; dividing data preprocessing, feature extraction, model training and prediction tasks into several subtasks; training the ARIMA model and the BSTS model separately in a parallel environment, and optimizing model parameters using historical data; obtaining parameter estimates through the ARIMA model and the BSTS model respectively, and using the ARIMA model and the BSTS model to predict at a future time point to obtain a prediction result; using the weighted average method to weighted average the prediction results of the ARIMA model and the BSTS model to obtain a fusion model, and using the prediction result obtained by the weighted average as the prediction output of the fusion model.
[0074] Furthermore, the fusion model is used to predict the generation of renewable energy; real-time multi-source data is obtained and input into the fusion model for real-time prediction; a parallel computing framework is used to perform parallel computing of several tasks to improve the prediction speed; the prediction result confidence interval is provided through Bayesian inference to quantify uncertainty; the power generation plan is dynamically adjusted by judging the prediction results and confidence intervals of the fusion model to improve the utilization efficiency of renewable energy; the specific formula of the fusion model is as follows:
[0075]
[0076] in, is the prediction result of the fusion model, ω ARIMA is the weight of the ARIMA model, is the prediction result of the ARIMA model, ω BSTS is the weight of the BSTS model, This is the prediction result of the BSTS model.
[0077] Specifically, the prediction results of the fusion model include: if the prediction result of the fusion model is greater than the observed value, and the upper confidence bound is much greater than the lower confidence bound, then the supply of new energy is expected to be sufficient, and the power generation plan is increased; if the prediction result of the fusion model is less than the observed value, and the lower confidence bound is much lower than the upper confidence bound, then the supply of new energy is expected to be insufficient, and a corresponding supplementary plan is formulated; if the gap between the prediction result of the fusion model and the observed value is within the confidence interval, and the confidence interval is narrow, then the prediction result is accurate, and the power generation plan is arranged according to the prediction result of the fusion model; if the gap between the prediction result of the fusion model and the observed value is outside the confidence interval, and the confidence interval is wide, then there is uncertainty in the prediction, and the actual situation should be closely monitored and the power generation plan should be adjusted in time; if in a continuous prediction cycle, the actual value continues to deviate from the predicted value by an amount less than the predicted value, then the ARIMA model and the BSTS model are re-integrated to optimize the fusion model structure.
[0078] Furthermore, this embodiment also provides a new energy power generation capacity prediction system based on parallel computing, including: an extraction module, which is used to collect multi-source heterogeneous data of new energy power generation, and preprocess and extract time series features; a construction module, which constructs an ARIMA model and a BSTS model based on time series features; a prediction module, which uses a weighted average method to fuse the ARIMA model and the BSTS model, obtains a fusion model and performs real-time power generation capacity prediction.
[0079] This embodiment also provides a computer device, which is suitable for the case of a parallel computing method for predicting new energy power generation capacity, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the parallel computing method for predicting new energy power generation capacity as proposed in the above embodiment.
[0080] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.
[0081] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the following steps are implemented: collecting multi-source heterogeneous data of renewable energy power generation, and preprocessing and extracting time series features; constructing an ARIMA model and a BSTS model based on the time series features; and fusing the ARIMA model and the BSTS model using a weighted average method to obtain a fusion model and predict power generation capacity.
[0082] In summary, the present invention ensures the integrity and consistency of data through data preprocessing and time series feature extraction; constructs ARIMA and BSTS models, fully utilizes their respective advantages, and realizes multi-angle analysis of time series data; finally, the two models are integrated through the weighted average method, which significantly improves the accuracy and reliability of power generation capacity prediction, provides robust prediction results, realizes efficient and stable operation of new energy power generation systems, and improves economic and social benefits.
[0083] Example 2
[0084] Referring to Table 1, which is the second embodiment of the present invention, this embodiment provides a new energy power generation capacity prediction method based on parallel computing. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.
[0085] Specifically, we selected photovoltaic power generation data, wind power generation data, meteorological data (including temperature, wind speed and sunshine duration) and other relevant external variables (such as seasonal factors and holiday information) of a certain region over the past five years.
[0086] Furthermore, interpolation and moving average methods are used to handle missing values, outliers and noise to ensure data quality; time series data are differentiated and smoothed to ensure their stability; sliding window technology is used to extract time series features from preprocessed data; feature selection is performed through correlation analysis to reduce dimensions and reduce model complexity; based on the selected time series features, an ARIMA model is constructed to capture linear relationships and short-term dependencies, and a BSTS model is constructed in combination with external predictors and structural components (such as trends, seasonality and regression terms) to capture complex nonlinear relationships and long-term dependencies. In the BSTS model, Bayesian inference methods are used for parameter estimation to handle uncertainty in the data.
[0087] Furthermore, in order to improve computing efficiency, the Hadoop big data processing framework is used to allocate computing resources and set up a parallel computing environment; data preprocessing, feature extraction, model training and prediction tasks are divided into multiple subtasks and processed in parallel through a parallel computing environment; in the parallel computing environment, the ARIMA model and BSTS model are trained separately, and the model parameters are optimized using historical data.
[0088] Specifically, after the ARIMA model and BSTS model are trained, the weighted average method is used to combine the two models to obtain a fusion model. The prediction of new energy power generation is carried out through the prediction results of the fusion model of the fusion model. In order to make real-time predictions, real-time multi-source data are obtained and substituted into the fusion model. In the prediction process, multi-task parallel computing is performed through the parallel computing framework to improve the prediction speed; Bayesian inference is used to provide the confidence interval of the prediction results of the fusion model and quantify the uncertainty; the prediction results and model performance of the fusion model are displayed through charts and dashboards for easy understanding and decision support.
[0089] Specifically, as shown in Table 1, the average prediction error of the fusion model of the present invention is 6.2 kWh, while the average prediction errors of the prior art are 12.5 kWh and 10.8 kWh, respectively. This shows that the fusion model of the present invention can more accurately predict the power generation of renewable energy and reduce the prediction error. The peak prediction error of the fusion model of the present invention is 16.8 kWh, which is much lower than 28.7 kWh of the prior art and 24.5 kWh of the BSTS model. This shows that the fusion model of the present invention also performs well in dealing with extreme situations and can effectively control the maximum value of the prediction error.
[0090] Table 1 Test data table
[0091] Test subjects Prior art Prior art The present invention Average forecast error (kWh) 12.5 10.8 6.2 Peak prediction error (kWh) 28.7 24.5 16.8 Prediction accuracy (%) 82.3 85.6 92.7 Training time (seconds) 105 215 310 Prediction time (seconds) 2.1 3.8 4.2 Storage space occupied (MB) 25 37 55
[0092] Furthermore, the prediction accuracy of the fusion model of the present invention is as high as 92.7%, while the accuracy of the prior art is 82.3% and 85.6%, respectively. This fully demonstrates the superiority of the fusion model of the present invention in improving prediction accuracy; the training time of the fusion model of the present invention is relatively long (310 seconds), but this is due to the simultaneous training of the ARIMA model and the BSTS model, as well as the model fusion. In contrast, the training time of using a single ARIMA model or BSTS model is 105 seconds and 215 seconds, respectively. Through parallel computing technology, the present invention can effectively improve training efficiency. The prediction time of the fusion model of the present invention is 4.2 seconds, which is only slightly higher than the 2.1 seconds and 3.8 seconds of the prior art. This shows that the present invention takes into account prediction efficiency while maintaining high prediction accuracy.
[0093] Furthermore, the storage space occupied by the fusion model of the present invention is 55MB, which is higher than the 25MB of the prior art and the 37MB of the BSTS model. This is because the fusion model needs to store the parameters and results of two separate models. However, compared with the improvement of prediction accuracy and efficiency, this cost is acceptable.
[0094] 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 new energy power generation capacity prediction method based on parallel computing, characterized by: include, Collect multi-source heterogeneous data of renewable energy power generation, pre-process and extract time series features; Based on the time series characteristics, construct ARIMA model and BSTS model; The ARIMA model and the BSTS model are fused using the weighted average method to obtain the fusion model and perform power generation capacity prediction.
2. The parallel computing method for predicting renewable energy power generation capacity according to claim 1, characterized in that: The multi-source heterogeneous data includes historical power generation data, weather data and external variable data; the construction process of the ARIMA model includes the following steps: The stationarity of time series data is tested by ADF, the autocorrelation function ACF diagram and partial autocorrelation function PACF diagram of time series data are drawn respectively, and the values of the difference times d, autoregressive term p and moving average term q are determined; Input the difference times d, autoregressive term p and moving average term q into the ARIMA model to fit the training data; By observing the ACF graph and PACF graph of the residual sequence, and using the AIC information criterion to evaluate the ARIMA model of different order combinations; The future time points are predicted by the evaluated ARIMA model to obtain the predicted values and corresponding confidence intervals.
3. The parallel computing method for predicting renewable energy power generation capacity according to claim 2, characterized in that: The relevant formula for the ADF test time series data is as follows: Δy t =α+βt+γy t-1 +δ1Δy t-1 +...+d p Δy t-p +e t Among them, Δy t-p is the first-order difference of the time series, α is the constant term, βt is the time trend term, γ is the autoregressive coefficient, δ p is the coefficient of the difference term p, ε t is the error term; The test results of time series data are tested by ADF to determine the stationarity of the data; When the first-order difference value is less than the set threshold, the data is judged to be stable; When the first-order difference value is greater than or equal to the set threshold, the data is judged to be disordered; If the data is disordered, then stationary the data by differencing and repeating this operation until the series becomes stationary.
4. The parallel computing method for predicting new energy generation capacity according to claim 2, characterized in that: Also includes, The relevant formula for the observed residual sequence is as follows: Among them, ρ t is the time series y t and the correlation with the lagged values, n is the number of lag orders, y t is the observed value of the time series, is the lagged value of the time series; If the ACF of the observed residual sequence is truncated to 0 after a certain lag order, then this lag order is the moving average term q; If the ACF of the observed residual sequence is truncated to 1 after a certain lag order, then this lag order is the autoregressive term p; The relevant formula of the ARIMA model is as follows: y t =d+φ1y t-1 +...+φ p y t-p +θ1ε t-1 +...+θ q e t-q +e t Among them, y t is the observed value of the time series, φ p is the autoregressive coefficient, θ q is the moving average coefficient, p is the autoregressive term, q is the moving average term, is the lag value of the time series, and d is the number of differences.
5. The parallel computing method for predicting renewable energy power generation capacity according to claim 1, characterized in that: The steps of constructing the BSTS model include the following steps: Determine the trend type, trend period, external predictor variables and regression terms, input them into the BSTS model, and use the Bayesian regression method to fit the regression terms; Test the BSTS model parameters and check the fitting residuals; Use Bayesian inference methods to estimate parameters, handle uncertainty in data, and obtain parameter estimates; Compare the fitting effects of BSTS models under different order combinations, select the BSTS model with the best fitting effect, and use cross-validation to evaluate the prediction accuracy of the model; The evaluated BSTS model is used to predict future time points and provide predicted values and confidence intervals.
6. The parallel computing method for predicting new energy generation capacity according to claim 5, characterized in that: The ARIMA model and the BSTS model are fused by a weighted average method to generate a fusion model; the construction process of the fusion model includes the following steps: Use the Hadoop big data framework to allocate computing resources and set up a parallel computing environment; Divide data preprocessing, feature extraction, model training, and prediction tasks into several subtasks; Train the ARIMA model and BSTS model separately in a parallel environment and optimize the model parameters using historical data; The parameter estimates are obtained through the ARIMA model and the BSTS model respectively, and the ARIMA model and the BSTS model are used to make predictions at future time points to obtain the prediction results; The prediction results of the ARIMA model and the BSTS model are weighted averaged by the weighted average method to obtain a fusion model, and the prediction results obtained by the weighted average are used as the prediction output of the fusion model; Forecasting of renewable energy generation through fusion model; Acquire real-time multi-source data and input it into the fusion model for real-time prediction; Use the parallel computing framework to perform parallel computing of several tasks to improve the prediction speed; Provide confidence intervals for prediction results through Bayesian inference to quantify uncertainty; By judging the prediction results and confidence intervals of the fusion model, the power generation plan can be adjusted dynamically to improve the efficiency of new energy utilization; The specific formula of the fusion model is as follows: in, is the prediction result of the fusion model, ω ARIMA is the weight of the ARIMA model, is the prediction result of the ARIMA model, ω BSTS is the weight of the BSTS model, This is the prediction result of the BSTS model.
7. The parallel computing method for predicting renewable energy power generation capacity according to claim 6, characterized in that: The prediction results of the fusion model include: If the prediction result of the fusion model is greater than the observed value, and the upper confidence bound is much greater than the lower confidence bound, it is expected that the supply of new energy is sufficient and the power generation plan is increased; If the prediction result of the fusion model is less than the observed value, and the lower confidence bound is much lower than the upper confidence bound, it is expected that the new energy supply is insufficient, and a corresponding supplementary plan is formulated; If the difference between the prediction result of the fusion model and the observed value is within the confidence interval, and the confidence interval is narrow, the prediction result is accurate, and the power generation plan is arranged according to the prediction result of the fusion model; If the difference between the prediction result of the fusion model and the observed value is outside the confidence interval, and the confidence interval is wide, there is uncertainty in the prediction, and the actual situation should be closely monitored and the power generation plan should be adjusted in a timely manner; If the actual value continues to deviate from the predicted value by an amount less than the predicted value during the continuous forecast period, the ARIMA model and the BSTS model will be fused again to optimize the fusion model structure.
8. A new energy power generation capacity prediction system based on parallel computing, a new energy power generation capacity prediction method based on parallel computing according to any one of claims 1 to 7, characterized in that: include, The extraction module is used to collect multi-source heterogeneous data of renewable energy generation, and to perform preprocessing and extract time series features; Building module, constructing ARIMA model and BSTS model based on time series characteristics; The prediction module uses the weighted average method to fuse the ARIMA model and the BSTS model, obtains the fusion model and performs real-time power generation capacity prediction.
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 parallel computing new energy power generation capacity prediction method according to 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 parallel computing new energy power generation capacity prediction method according to any one of claims 1 to 7 are implemented.
