Comprehensive energy load short-term prediction method, system, equipment and medium

Abstract

The invention discloses a comprehensive energy load short-term prediction method, system and device and a medium, and the method comprises the steps: screening meteorological factors which affect a load through employing a Pearson correlation coefficient, and determining a meteorological sensitive factor; calculating the similarity between the historical dates and the to-be-predicted dates through a grey correlation analysis algorithm, selecting a plurality of historical dates with the highest similarity, and constructing a similar day training set; carrying out adaptive decomposition on the load sequence by adopting an improved variational mode decomposition algorithm to obtain an intrinsic mode function component set; optimizing core parameters of the support vector machine model through an improved sparrow search algorithm, and predicting each sub-component by using the optimized model to obtain a predicted value of each sub-component; and adding and merging the sub-component prediction values of the same load type according to a time sequence to obtain a comprehensive energy load short-term prediction value. The method effectively improves the precision, stability and adaptability of short-term load prediction of the integrated energy system.

Description

Technical Field

[0001] This invention relates to the field of smart energy technology, and in particular to a method, system, device and medium for short-term prediction of integrated energy load. Background Technology

[0002] Multi-load forecasting in integrated energy systems is crucial for dispatch optimization, but existing technologies face three major bottlenecks: 1) Full sequence input leads to model redundancy; 2) A single model is difficult to adapt to the characteristics of multi-energy coupling; 3) The dynamic relationship between meteorological and date factors has not been fully utilized. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] Therefore, this invention provides a comprehensive energy load short-term forecasting method, system, device, and medium to address the problem of insufficient forecasting accuracy in existing short-term forecasting methods.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for short-term forecasting of comprehensive energy load, comprising: Historical meteorological parameters and multivariate load data were obtained, and Pearson correlation coefficient was used to screen meteorological factors that affect the load and determine meteorological sensitive factors. Based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, the similarity between historical dates and the date to be predicted is calculated using a grey relational analysis algorithm. Several historical days with the highest similarity are selected to construct a training set of similar days. Based on each date in the similar day training set, an improved variational mode decomposition algorithm is used to adaptively decompose the load sequence to obtain the set of intrinsic mode function components; Based on the intrinsic mode function component set, the core parameters of the support vector machine model are optimized by improving the sparrow search algorithm, and the optimized model is used to predict each sub-component to obtain the predicted value of each sub-component. Based on the predicted values ​​of each sub-component, the short-term predicted value of the comprehensive energy load is obtained by adding and merging the predicted values ​​of the sub-components of the same load type in chronological order.

[0006] As a preferred embodiment of the integrated energy load short-term forecasting method of the present invention, the determination of meteorological sensitive factors includes: Align historical meteorological parameters with corresponding multivariate load data by timestamp; Calculate the Pearson correlation coefficient between each meteorological parameter and the four loads; For each calculated correlation coefficient, determine whether the absolute value of the calculated Pearson correlation coefficient is greater than the preset correlation threshold. The Pearson correlation coefficient with an absolute value greater than the aforementioned correlation threshold is used as the meteorological sensitivity factor.

[0007] As a preferred embodiment of the integrated energy load short-term forecasting method described in this invention, the construction of the similar day training set includes: Based on meteorological sensitivity factors, date type factors, and real-time electricity price data, a comprehensive matching coefficient model is constructed that integrates meteorological similarity, date matching degree, and electricity price influence. Using the comprehensive matching coefficient model, the similarity value between each historical day and the day to be predicted is calculated; The similarity value is compared with a preset similarity threshold to determine whether the similarity condition is met. Select historical days that meet the similarity criteria to construct a training set of similar days.

[0008] The beneficial effects of this preferred technical solution are that by constructing a comprehensive matching coefficient model that integrates meteorological similarity, date matching degree and electricity price influence, and dynamically filtering historical days that meet the conditions based on similarity thresholds, the accuracy and adaptability of similar day selection can be effectively improved, thereby enhancing the training quality and short-term prediction accuracy of the load forecasting model.

[0009] As a preferred embodiment of the integrated energy load short-term forecasting method described in this invention, the step of obtaining the intrinsic mode function component set includes: Initialize the number of modes and penalty factor in the variational mode decomposition; The load sequence is decomposed into intermediate components based on the current number of modes and the penalty factor to obtain a set of intrinsic mode functions and calculate the envelope entropy of each component. Determine whether the current decomposition result meets the modal stability requirement based on the envelope entropy value of each component; If the envelope entropy of any component exceeds the preset complexity threshold, the parameter adaptive adjustment mechanism is activated and the adjustable range of the number of modes and the penalty factor is set. Multiple candidate parameter combinations of modal number and penalty factor are generated within the adjustable range; For each set of candidate parameter combinations, the load sequence is re-decomposed into intermediate values, the envelope entropy of each component is calculated, and the sum of the envelope entropies of all components is obtained. By comparing the total envelope entropy under different parameter combinations, the parameter combination with the minimum total envelope entropy is determined as the optimal parameter combination. The load sequence is finally decomposed based on the optimal parameter combination to obtain the set of intrinsic mode function components.

[0010] The beneficial effects of this preferred technical solution are that by adaptively optimizing the number of modes and the penalty factor of variational mode decomposition with the goal of minimizing the envelope entropy, it can effectively avoid the under-decomposition or over-decomposition problems caused by manually setting parameters, significantly improve the stability and accuracy of load sequence decomposition, and provide a high-quality input basis for subsequent component-level prediction.

[0011] As a preferred embodiment of the integrated energy load short-term forecasting method of the present invention, wherein obtaining the predicted values ​​of each sub-component includes: For each eigenmode function component obtained by variational mode decomposition, a corresponding least squares support vector machine prediction sub-model is constructed. By introducing the improved sparrow search algorithm based on the Levy flight strategy, the width factor and regularization parameter of the kernel function of each prediction sub-model are globally optimized. The least squares support vector machine sub-models are trained based on the optimized parameters to obtain the predicted values ​​of each sub-component.

[0012] As a preferred embodiment of the integrated energy load short-term forecasting method described in this invention, the parameter adaptive adjustment mechanism includes: If the current number of modes is less than the preset maximum number of modes, the number of modes will be increased by one unit, the penalty factor will remain unchanged, and the load sequence will be decomposed again. If the current number of modes has reached the maximum upper limit of the number of modes, but there are still intrinsic mode function components whose envelope entropy is greater than the complexity threshold, then keep the number of modes unchanged, and gradually increase the penalty factor within the allowed range with a preset increment step size, and decompose again after each update.

[0013] As a preferred embodiment of the integrated energy load short-term forecasting method described in this invention, wherein: the integrated matching coefficient model Represented as: in, , , These are the weighting coefficients for weather, date type, and electricity price factors, respectively. This represents the absolute value of the temperature difference between two days. and These represent the minimum and maximum temperature differences across all historical days. The resolution coefficient, For date type matching functions, This represents the maximum value of the electricity price data.

[0014] Secondly, the present invention provides a comprehensive energy load short-term forecasting system, comprising: The meteorological sensitive factor identification module is used to acquire historical meteorological parameters and multivariate load data, and uses Pearson correlation coefficient to screen meteorological factors that affect the load and determine meteorological sensitive factors. The similar day dynamic screening module is used to calculate the similarity between historical dates and the date to be predicted based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, and selects several historical days with the highest similarity to construct a similar day training set; The load sequence adaptive decomposition module is used to adaptively decompose the load sequence based on each date in the similar day training set, using an improved variational mode decomposition algorithm to obtain the set of intrinsic mode function components. The component-level intelligent prediction modeling module is used to optimize the core parameters of the support vector machine model based on the set of intrinsic mode function components by improving the sparrow search algorithm, and then use the optimized model to predict each sub-component to obtain the predicted value of each sub-component. The prediction result fusion output module is used to obtain the short-term prediction value of comprehensive energy load by adding and merging the prediction values ​​of sub-components of the same load type in chronological order based on the prediction values ​​of each sub-component.

[0015] Thirdly, the present invention provides an electronic device, comprising: Memory, used to store programs; A processor for executing the computer-executable instructions, which, when executed by the processor, implement the steps of the integrated energy load short-term forecasting method.

[0016] Fourthly, the present invention provides a computer-readable storage medium comprising: when the program is executed by a processor, the steps of implementing the integrated energy load short-term forecasting method.

[0017] The beneficial effects of this invention are as follows: By constructing a comprehensive matching coefficient model that integrates meteorological sensitive factors, date type factors, and real-time electricity price data, and dynamically screening similar days based on grey relational analysis, this invention achieves accurate selection of the training set of similar days, effectively improving the relevance and representativeness of the input data of the prediction model; by introducing the envelope entropy minimization criterion to adaptively optimize the number of modes and penalty factors of variational mode decomposition, this invention achieves high-quality adaptive decomposition of the load sequence, significantly enhancing the stationarity and physical interpretability of each intrinsic mode function component; by employing an improved sparrow search algorithm (introducing the Levy flight strategy) to optimize the kernel parameters and regularization parameters of the least squares support vector machine, and establishing independent prediction models for each component, this invention achieves high-precision component-level prediction, effectively improving the model's nonlinear fitting ability and global search performance; by linearly superimposing and fusing the prediction results of each intrinsic mode function component, this invention achieves multi-model collaborative output, taking into account the trend, periodicity, and randomness characteristics of load changes, ultimately significantly improving the overall accuracy and robustness of short-term prediction of integrated energy loads such as cooling, heating, electricity, and gas. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the basic process of a short-term comprehensive energy load forecasting method provided in one embodiment of the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0020] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for short-term forecasting of comprehensive energy load is provided, comprising: S100: Obtain historical meteorological parameters and multivariate load data, use Pearson correlation coefficient to screen meteorological factors that affect the load, and determine meteorological sensitive factors; S200: Based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, the similarity between historical dates and the date to be predicted is calculated using a grey relational analysis algorithm. Several historical days with the highest similarity are selected to construct a training set of similar days. S300: Based on each date in the similar day training set, the improved variational mode decomposition algorithm is used to adaptively decompose the load sequence to obtain the set of intrinsic mode function components; S400: Based on the intrinsic mode function component set, the core parameters of the support vector machine model are optimized by improving the sparrow search algorithm, and the optimized model is used to predict each sub-component to obtain the predicted value of each sub-component. S500: Based on the predicted values ​​of each sub-component, the short-term predicted value of the comprehensive energy load is obtained by adding and merging the predicted values ​​of the sub-components of the same load type in chronological order.

[0021] It should be noted that existing short-term load forecasting methods face a series of challenges during operation: traditional models typically rely on manual experience to select similar days, making it difficult to comprehensively consider the coupling effects of multiple dimensions such as weather, date type, and electricity price, resulting in insufficient representativeness of the training set; when dealing with non-stationary, highly fluctuating load sequences, fixed-parameter signal decomposition methods (such as VMD and EMD) are prone to mode aliasing or over-decomposition problems, and the number of modes and penalty factors depend on trial and error, lacking adaptability; at the same time, single forecasting models have limited ability to fit complex nonlinear load characteristics, are prone to getting trapped in local optima, and have poor generalization performance; in addition, most methods do not fully explore the multi-scale characteristics of the load, lack a refined modeling paradigm of "divide and conquer," and are difficult to balance forecast accuracy and stability, thus restricting the reliability and economy of integrated energy system dispatch.

[0022] Therefore, to address the problem of insufficient prediction accuracy in existing short-term forecasting methods, the following steps are taken from S100 to S500: meteorological sensitive factors are screened using Spearman correlation coefficient, and a training set is constructed by selecting similar days using grey relational analysis; an improved variational mode decomposition technique is used to decompose the load data into multi-frequency components; an improved sparrow search algorithm is used to optimize the support vector machine parameters for component prediction; and finally, the fusion result is output as the predicted value.

[0023] Example 2, this is an embodiment of the present invention, which provides a comprehensive energy load short-term forecasting method based on the previous embodiment, including: In this embodiment, the historical meteorological parameters in step S100 include temperature, humidity, wind speed, sunshine duration, and weather condition coding values; the multi-variable load data includes hourly measured power values ​​of cooling load, heating load, electrical load, and gas load. Data preprocessing includes: aligning the meteorological data and load data with hourly timestamps, completing missing values ​​using linear interpolation, and removing outliers using the 3σ criterion.

[0024] In this embodiment of the application, the meteorological sensitive factor screening method in step S100 includes obtaining historical meteorological parameters and multivariate load data, using the Pearson correlation coefficient to calculate the linear correlation between meteorological parameters and load, and selecting parameters whose absolute value is greater than a preset threshold as meteorological sensitive factors.

[0025] In an optional implementation, the meteorological sensitive factor screening method in step S100 can also acquire historical meteorological parameters and multivariate load data, calculate the Spearman rank correlation coefficient between each meteorological parameter and the load sequence, and select parameters whose absolute value is greater than a preset threshold as meteorological sensitive factors.

[0026] In an optional implementation, the meteorological sensitive factor screening method in step S100 can also acquire historical meteorological parameters and multivariate load data, calculate the mutual information value between each meteorological parameter and the load sequence, sort them according to the information gain, and select parameters that are higher than a preset threshold as meteorological sensitive factors.

[0027] In this embodiment of the application, step S100, determining the meteorological sensitive factor, includes: Align historical meteorological parameters with corresponding multivariate load data by timestamp; Calculate the Pearson correlation coefficient between each meteorological parameter and the four loads; For each calculated correlation coefficient, determine whether the absolute value of the calculated Pearson correlation coefficient is greater than the preset correlation threshold (0.5). The Pearson correlation coefficient with an absolute value greater than the correlation threshold is used as a meteorological sensitivity factor.

[0028] In this embodiment, the similar day matching algorithm includes using grey relational analysis, combining temperature difference, date type matching, and electricity price fluctuation to construct a comprehensive matching coefficient, and selecting historical days with a matching degree higher than a threshold.

[0029] In an optional implementation, the similar day matching algorithm in step S200 can also construct a multi-dimensional feature vector based on meteorological sensitive factors, date category features and real-time electricity price data, calculate the standardized Euclidean distance and cosine similarity between the day to be predicted and each historical day, generate a comprehensive similarity index through weighted fusion, and select several historical days with the highest similarity to form a similar day training set.

[0030] In an optional implementation, the similar day matching algorithm in step S200 can also construct a multi-dimensional time series based on meteorological sensitive factors, date category features and real-time electricity price data, calculate the dynamic time regularization distance between the day to be predicted and each historical day series, select the K historical days with the smallest distance as similar days, and form a similar day training set.

[0031] In this embodiment of the application, step S200, which involves constructing a training set of similar days, includes: Based on meteorological sensitivity factors, date type factors, and real-time electricity price data, a comprehensive matching coefficient model is constructed that integrates meteorological similarity, date matching degree, and electricity price influence. The similarity value between each historical day and the day to be predicted is calculated using a comprehensive matching coefficient model. The similarity value is compared with a preset similarity threshold to determine whether the similarity condition is met. Select historical days that meet the similarity criteria to construct a training set of similar days.

[0032] In this embodiment, the date category feature is represented by a seven-dimensional binary vector from Monday to Sunday, with an additional binary flag indicating whether it is a statutory holiday. The real-time electricity price data is the time-of-use electricity price sequence for the day before the forecast date, and its maximum value is taken as the normalization benchmark.

[0033] In this embodiment, the comprehensive matching coefficient ξ between each historical day and the day to be predicted is calculated, and historical days with ξ > 0.8 are included in the similar day training set. If there are fewer than 3 historical days that meet the condition, the threshold is gradually reduced to 0.75 until the number of training set samples is no less than 3.

[0034] In this application embodiment, the comprehensive matching coefficient model Represented as: in, , , These are the weighting coefficients for weather, date type, and electricity price factors, respectively. ; To construct a training set, we select historical days with ξ > 0.8, representing the absolute value of the temperature difference between two days. and These represent the minimum and maximum temperature differences across all historical days. The resolution coefficient (usually taken as 0.5). This is a function for matching date types (1 for the same, 0 for different). This represents the maximum value of the electricity price data.

[0035] In the embodiments of this application, the signal decomposition method includes using improved variational mode decomposition (VMD), adaptively determining the number of modes K and the penalty factor α by minimizing the envelope entropy, and decomposing the load sequence into multiple intrinsic mode functions (IMFs).

[0036] In an optional implementation, the signal decomposition method in step S300 can also add white noise of different amplitudes to the load sequences in the similar day training set and perform multiple empirical mode decompositions. By eliminating the influence of noise through ensemble averaging, a series of intrinsic mode function components are adaptively obtained, and finally the original signal is decomposed into multi-scale sub-components with physical meaning.

[0037] In an optional implementation, the signal decomposition method in step S300 can also select a suitable wavelet basis for multi-level wavelet packet decomposition of the load sequences in the similar day training set, determine the optimal decomposition structure based on the energy distribution or entropy criterion, and adaptively decompose the load sequences into multiple frequency band components, which are then used as sub-components to input into the subsequent prediction model for modeling.

[0038] In this embodiment of the application, step S300, which obtains the set of intrinsic mode function components, includes: Initialize the number of modes and penalty factor in the variational mode decomposition; The load sequence is decomposed into intermediate components based on the current number of modes and the penalty factor to obtain a set of intrinsic mode functions and calculate the envelope entropy of each component. Based on the envelope entropy value E of each component k Determine whether the current decomposition result meets the modal stationarity requirement; If there exists an envelope entropy E for any component k If the complexity exceeds the preset threshold (0.8), it indicates that the modal component still contains a lot of noise or is not fully decomposed. In this case, the parameter adaptive adjustment mechanism is activated and the adjustable range of the number of modes and the penalty factor is set. Otherwise, it is considered that the current decomposition quality meets the requirements. The current number of modes and the penalty factor are directly used as the optimal parameter combination, and the load sequence is finally decomposed based on the parameters to output the set of intrinsic modal function components.

[0039] Generate multiple candidate parameter combinations of modal number and penalty factor within an adjustable range; For each set of candidate parameter combinations, the load sequence is re-decomposed into intermediate values, the envelope entropy of each component is calculated, and the sum of the envelope entropies of all components is obtained. By comparing the total envelope entropy under different parameter combinations, the parameter combination with the minimum total envelope entropy is determined as the optimal parameter combination. The load sequence is finally decomposed based on the optimal parameter combination to obtain the set of intrinsic mode function components.

[0040] In this embodiment, the initial modality number K=5 and the penalty factor α=1000; In this embodiment of the application, the parameter adaptive adjustment mechanism in step S300 includes: If the current number of modes K is less than the preset maximum number of modes, the number of modes is incremented by one unit, the penalty factor remains unchanged, and the load sequence is decomposed again; for example, if K < 10, the number of modes K is increased by 1, and the variational mode decomposition step is returned.

[0041] If the current number of modes has reached the maximum upper limit of modes, but there are still intrinsic mode function components whose envelope entropy is greater than the complexity threshold, then the number of modes remains unchanged, and the penalty factor is gradually increased within the allowed range with a preset increment step size. The decomposition is performed again after each update. For example, if K=10 but there are still E... k If the value is greater than 0.8, then keep K=10 and adjust the penalty factor. The step size is increased by 200 (i.e., α = α + 200) until α = 3000. After each update, the variational mode decomposition is returned to decompose again.

[0042] In the embodiments of this application, the construction of the sub-component prediction model includes optimizing the LSSVM parameters using an improved sparrow search algorithm for each decomposed sub-component and establishing a least squares support vector machine (LSSVM) prediction sub-model.

[0043] In an optional implementation, the sub-component prediction model construction in step S400 can also construct a long short-term memory network model for each load sub-component obtained by signal decomposition, using historical time series data as input features, training the LSTM network to learn the long-term dependencies of the sub-components, and using the trained model to make point-by-point predictions to obtain the predicted values ​​of each sub-component.

[0044] In an optional implementation, the sub-component prediction model construction in step S400 can also construct XGBoost regression models for each load sub-component obtained by signal decomposition, using historical load values, time characteristics and lagged variables as input features, training the model and optimizing hyperparameters through gradient boosting tree algorithm, and using the trained model to predict each sub-component to obtain its future time series prediction value.

[0045] In this embodiment of the application, obtaining the predicted value of each sub-component in step S400 includes: For each eigenmode function component obtained by variational mode decomposition, a corresponding least squares support vector machine prediction sub-model is constructed. By introducing the improved sparrow search algorithm based on the Levy flight strategy, the width factor and regularization parameter of the kernel function of each prediction sub-model are globally optimized. The least squares support vector machine sub-models are trained based on the optimized parameters to obtain the predicted values ​​of each sub-component.

[0046] In this embodiment, the kernel function of the Least Squares Support Vector Machine (LSSVM) is represented by the radial basis function as follows: In the formula, It is a kernel function, representing a nonlinear mapping from the input space to a high-dimensional feature space, i.e., two data points. and Kernel function values ​​between; It is the width factor of the kernel function, which is an adjustable parameter; There are two data points. and The Euclidean distance between them.

[0047] In this embodiment, the improved sparrow search algorithm incorporating the Levy flight strategy is represented as follows: In the formula, It is a gamma function that satisfies the following when x is a positive integer: For non-integer inputs, numerical methods are used to calculate their function values. It is a constant. Take now ; Random numbers.

[0048] Example 3 is an embodiment of the present invention. This embodiment differs from the first embodiment in that it provides a comprehensive energy load short-term forecasting system.

[0049] It should be noted that the technical solution of the integrated energy load short-term forecasting system and the technical solution of the integrated energy load short-term forecasting method described above belong to the same concept. For details not described in detail in the technical solution of the integrated energy load short-term forecasting system in this embodiment, please refer to the description of the technical solution of the integrated energy load short-term forecasting method described above.

[0050] This embodiment provides a comprehensive energy load short-term forecasting system, comprising: The meteorological sensitive factor identification module is used to acquire historical meteorological parameters and multivariate load data, and uses Pearson correlation coefficient to screen meteorological factors that affect the load and determine meteorological sensitive factors. The similar day dynamic screening module is used to calculate the similarity between historical dates and the date to be predicted based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, and selects several historical days with the highest similarity to construct a similar day training set; The load sequence adaptive decomposition module is used to adaptively decompose the load sequence based on each date in the similar day training set, using an improved variational mode decomposition algorithm to obtain the set of intrinsic mode function components. The component-level intelligent prediction modeling module is used to optimize the core parameters of the support vector machine model based on the set of intrinsic mode function components by improving the sparrow search algorithm, and then use the optimized model to predict each sub-component to obtain the predicted value of each sub-component. The prediction result fusion output module is used to obtain the short-term prediction value of comprehensive energy load by adding and merging the prediction values ​​of sub-components of the same load type in chronological order based on the prediction values ​​of each sub-component.

[0051] This embodiment also provides an electronic device applicable to a comprehensive energy load short-term forecasting method, including: The system includes a memory and a processor. The memory stores computer-executable instructions, and the processor executes these instructions to implement a short-term comprehensive energy load forecasting method as described in the above embodiments.

[0052] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements a short-term comprehensive energy load forecasting method as proposed in the above embodiments.

[0053] The storage medium proposed in this embodiment belongs to the same inventive concept as the method for short-term prediction of comprehensive energy load proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0054] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0055] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A comprehensive energy load short-term forecasting method, characterized in that, include: Historical meteorological parameters and multivariate load data were obtained, and Pearson correlation coefficient was used to screen meteorological factors that affect the load and determine meteorological sensitive factors. Based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, the similarity between historical dates and the date to be predicted is calculated using a grey relational analysis algorithm. Several historical days with the highest similarity are selected to construct a training set of similar days. Based on each date in the similar day training set, an improved variational mode decomposition algorithm is used to adaptively decompose the load sequence to obtain the set of intrinsic mode function components; Based on the intrinsic mode function component set, the core parameters of the support vector machine model are optimized by improving the sparrow search algorithm, and the optimized model is used to predict each sub-component to obtain the predicted value of each sub-component. Based on the predicted values ​​of each sub-component, the short-term predicted value of the comprehensive energy load is obtained by adding and merging the predicted values ​​of the sub-components of the same load type in chronological order.

2. The integrated energy load short-term forecasting method as described in claim 1, characterized in that: The determination of meteorological sensitive factors includes: Align historical meteorological parameters with corresponding multivariate load data by timestamp; Calculate the Pearson correlation coefficient between each meteorological parameter and the four loads; For each calculated correlation coefficient, determine whether the absolute value of the calculated Pearson correlation coefficient is greater than the preset correlation threshold. The Pearson correlation coefficient with an absolute value greater than the aforementioned correlation threshold is used as the meteorological sensitivity factor.

3. The integrated energy load short-term forecasting method as described in claim 1 or 2, characterized in that: The construction of the similar day training set includes: Based on meteorological sensitivity factors, date type factors, and real-time electricity price data, a comprehensive matching coefficient model is constructed that integrates meteorological similarity, date matching degree, and electricity price influence. Using the comprehensive matching coefficient model, the similarity value between each historical day and the day to be predicted is calculated; The similarity value is compared with a preset similarity threshold to determine whether the similarity condition is met. Select historical days that meet the similarity criteria to construct a training set of similar days.

4. The integrated energy load short-term forecasting method as described in claim 3, characterized in that: The obtained intrinsic mode function component set includes: Initialize the number of modes and penalty factor in the variational mode decomposition; The load sequence is decomposed into intermediate components based on the current number of modes and the penalty factor to obtain a set of intrinsic mode functions and calculate the envelope entropy of each component. Determine whether the current decomposition result meets the modal stability requirement based on the envelope entropy value of each component; If the envelope entropy of any component exceeds the preset complexity threshold, the parameter adaptive adjustment mechanism is activated and the adjustable range of the number of modes and the penalty factor is set. Multiple candidate parameter combinations of modal number and penalty factor are generated within the adjustable range; For each set of candidate parameter combinations, the load sequence is re-decomposed into intermediate values, the envelope entropy of each component is calculated, and the sum of the envelope entropies of all components is obtained. By comparing the total envelope entropy under different parameter combinations, the parameter combination with the minimum total envelope entropy is determined as the optimal parameter combination. The load sequence is finally decomposed based on the optimal parameter combination to obtain the set of intrinsic mode function components.

5. The integrated energy load short-term forecasting method as described in claim 4, characterized in that: The process of obtaining the predicted values ​​for each sub-component includes: For each eigenmode function component obtained by variational mode decomposition, a corresponding least squares support vector machine prediction sub-model is constructed. By introducing an improved sparrow search algorithm based on the Levy flight strategy, the width factor and regularization parameter of the kernel function of each prediction sub-model are globally optimized. The least squares support vector machine sub-models are trained based on the optimized parameters to obtain the predicted values ​​of each sub-component.

6. The integrated energy load short-term forecasting method as described in claim 5, characterized in that: The parameter adaptive adjustment mechanism includes: If the current number of modes is less than the preset maximum number of modes, the number of modes will be increased by one unit, the penalty factor will remain unchanged, and the load sequence will be decomposed again. If the current number of modes has reached the maximum upper limit of the number of modes, but there are still intrinsic mode function components whose envelope entropy is greater than the complexity threshold, then keep the number of modes unchanged, and gradually increase the penalty factor within the allowed range with a preset increment step size, and decompose again after each update.

7. The integrated energy load short-term forecasting method as described in claim 6, characterized in that: The comprehensive matching coefficient model Represented as: ； in, , , These are the weighting coefficients for weather, date type, and electricity price factors, respectively. This represents the absolute value of the temperature difference between two days. and These represent the minimum and maximum temperature differences across all historical days. The resolution coefficient, For date type matching functions, This represents the maximum value of the electricity price data.

8. A comprehensive energy load short-term forecasting system, using the method described in any one of claims 1-7, characterized in that, include: The meteorological sensitive factor identification module is used to acquire historical meteorological parameters and multivariate load data, and uses Pearson correlation coefficient to screen meteorological factors that affect the load and determine meteorological sensitive factors. The similar day dynamic screening module is used to calculate the similarity between historical dates and the date to be predicted based on meteorological sensitivity factors, combined with date category characteristics and real-time electricity price fluctuations, and selects several historical days with the highest similarity to construct a similar day training set; The load sequence adaptive decomposition module is used to adaptively decompose the load sequence based on each date in the similar day training set, using an improved variational mode decomposition algorithm to obtain the set of intrinsic mode function components. The component-level intelligent prediction modeling module is used to optimize the core parameters of the support vector machine model based on the set of intrinsic mode function components by improving the sparrow search algorithm, and then use the optimized model to predict each sub-component to obtain the predicted value of each sub-component. The prediction result fusion output module is used to obtain the short-term prediction value of comprehensive energy load by adding and merging the prediction values ​​of sub-components of the same load type in chronological order based on the prediction values ​​of each sub-component.

9. An electronic device, characterized in that, include: Memory, used to store programs; A processor for loading the program to perform the steps of the method as claimed in any one of claims 1-7.

10. A computer-readable storage medium storing a program, characterized in that, When the program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.

Images