Load Forecasting Method, System, Device and Medium Based on CPO-VMD-MLKELM Combined Model

Abstract

The present invention belongs to the technical field of load forecasting, and discloses a load forecasting method, system, device and medium based on a CPO-VMD-MLKELM combined model to solve the problems of low generalization ability and poor prediction accuracy of existing models. The method of the present invention includes: obtaining historical load data, meteorological feature data and date feature data of a prediction target and performing preprocessing to obtain a preprocessed data set; performing feature dimensionality reduction on the preprocessed data set using PCA principal component analysis, and retaining the principal components with a cumulative variance contribution rate of not less than 98% to obtain the dimensionality-reduced principal component features; using the crown porcupine optimization algorithm to dynamically optimize the VMD parameters through four defense strategies, and decomposing the historical load data into multiple IMF components based on the optimized VMD parameters; constructing an MLKELM model for each IMF component, using the dimensionality-reduced principal component features as the input to predict each IMF component, and superimposing the predicted values of each IMF component to obtain the load prediction value. The MLKELM model includes a stacked KELM autoencoder and a KELM regressor.

Description

Technical Field

[0001] The present invention belongs to the technical field of load forecasting, and specifically relates to a load forecasting method, system, device and medium based on a CPO-VMD-MLKELM combined model. Background Art

[0002] Power load forecasting and future power generation planning are key links in the efficient operation of power systems. With the continuous rise of global energy demand and the accelerating transformation of the energy structure from fossil fuels to volatile renewable energy sources (such as wind energy and solar energy), the dynamic balance of power loads faces unprecedented complexity. Accurate load forecasting has become a core requirement in modern power system planning - even a small forecasting deviation may lead to power grid stability problems, increase operating costs, and even cause large-scale power outages, resulting in significant economic losses and social impacts. Therefore, improving the accuracy and reliability of load forecasting has always been a research focus in the field of power systems. As a shallow artificial neural network model, the Extreme Learning Machine (ELM) has become an important tool in the field of load forecasting in the past two decades due to its extremely fast training speed (weight optimization can be completed in a single iteration) and strong generalization ability. However, its shallow structure (usually only containing 1-2 hidden layers) limits the model's ability to model complex non-linear relationships and is difficult to capture the multi-scale time dependence (such as daily cycles, seasonal fluctuations, and the impact of emergencies) and high-dimensional feature interactions (such as the coupling of multi-source data such as meteorology, economy, and user behavior) in power loads. To break through the bottleneck of the expression ability of the shallow ELM, researchers have constructed a deep ELM model by stacking multiple hidden layers, attempting to enhance the feature extraction ability with a deep structure. However, the deep ELM has obvious defects: on the one hand, the gradient anomaly in backpropagation easily leads to convergence difficulties; on the other hand, although the complex model performs well on the training data, its generalization ability for new scenarios has decreased. In addition, the deep structure also faces the contradiction between computational efficiency and interpretability - although it can improve performance, it significantly increases the computational cost, and the model decision logic is difficult to trace, thus affecting its application in actual engineering. Summary of the Invention

[0003] Based on the above-mentioned disadvantages and deficiencies existing in the prior art, one of the purposes of the present invention is to at least solve one or more of the above-mentioned problems existing in the prior art. In other words, one of the purposes of the present invention is to provide a load forecasting method, system, device and medium based on a CPO-VMD-MLKELM combined model that meet one or more of the foregoing requirements, so as to significantly improve the accuracy of load forecasting, while ensuring that the model has excellent generalization ability, thereby achieving the purpose of adapting to complex and changeable power load scenarios.

[0004] To achieve the above-mentioned invention purpose, the present invention adopts the following technical solutions:

[0005] In a first aspect, the present invention provides a load forecasting method based on a CPO-VMD-MLKELM combined model, comprising the steps of: S1. Obtain the historical load data, meteorological feature data, and date feature data of the forecasting target and perform preprocessing to obtain a preprocessed data set; S2. Perform feature dimensionality reduction on the preprocessed data set using PCA principal component analysis, and retain the principal components with a cumulative variance contribution rate of not less than 98% to obtain the dimensionality-reduced principal component features; S3. Use the crown porcupine optimization algorithm to dynamically optimize the VMD parameters through four defense strategies, and decompose the historical load data into multiple IMF components based on the optimized VMD parameters, where the VMD parameters include the number of modes K and the penalty factor α; S4. Construct an MLKELM model for each IMF component, use the dimensionality-reduced principal component features as the input to predict each IMF component, and superimpose the predicted values of each IMF component to obtain the load forecasting value, where the MLKELM model includes a stacked KELM autoencoder and a KELM regressor.

[0006] As a preferred solution, the preprocessing in step S1 includes performing one-hot encoding on the date feature data, and the one-hot encoding is specifically: encoding weekdays as 0 and holidays as 1; encoding seasonal features as [1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 1] for spring, summer, autumn, and winter respectively.

[0007] As a preferred solution, the four defense strategies are visual defense, auditory defense, odor defense, and physical attack respectively; the visual defense updates the parameters based on the current optimal solution and normal distribution random numbers; the auditory defense determines the search direction through the positions of two random individuals; the odor defense uses a diffusion factor to control the parameter adjustment range; the physical attack corrects the parameters based on the inelastic collision theorem.

[0008] As a preferred solution, the value range of the diffusion factor in the odor defense is [0.3, 2.6].

[0009] As a preferred solution, the stacked KELM autoencoder includes a three-layer structure, and the number of nodes in each layer decreases in the ratio of 60%, 40%, and 20%.

[0010] As a preferred solution, the KELM regressor uses an RBF kernel function, and its bandwidth parameter is optimized and determined by the grid search method.

[0011] As a preferred solution, it also includes using the ten-fold cross-validation method to evaluate the MLKELM model.

[0012] In a second aspect, the present invention provides a load forecasting system based on a CPO-VMD-MLKELM combined model for implementing the load forecasting method as described in the first aspect.

[0013] In a third aspect, the present invention provides an electronic device. The computer device includes a memory, a processor, and a computer program. When the computer program is executed by the processor, the load forecasting method as described in the first aspect is implemented.

[0014] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by the processor, the load forecasting method as described in the first aspect is implemented.

[0015] Compared with the prior art, the present invention has the following beneficial effects:

[0016] 1. By optimizing the VMD parameters (the number of modes K and the penalty factor α) through CPO, the present invention effectively solves the mode mixing problem caused by the traditional VMD's dependence on manual experience for parameter tuning, making the decomposition of IMF components more accurate, thereby improving the accuracy of subsequent predictions.

[0017] 2. The MLKELM model combines an unsupervised stacked autoencoder and a supervised KELM regressor, which can deeply extract features and fit non-linear relationships. Compared with a single ELM or a shallow model, the prediction error for complex load sequences is reduced.

[0018] Further or more detailed beneficial effects will be described in combination with specific embodiments in the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 is a schematic flow chart of the load forecasting method described in the embodiments of the present invention.

[0021] Figure 2 is a schematic structural diagram of an autoencoder based on kernel extreme learning described in the embodiments of the present invention.

[0022] Figure 3 is a schematic diagram of the prediction principle of MLKELM described in the embodiments of the present invention.

[0023] Figure 4 is a structural diagram of the electronic device provided in the embodiments of the present invention.

[0024] Reference numerals in the drawings:

[0025] 400, electronic device;

[0026] 401, processor; 402, communication bus; 403, user interface; 404, network interface; 405, memory. Detailed implementation manners

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention.

[0028] In the following description, multiple embodiments of the present invention are provided. Different embodiments can be replaced or combined. Therefore, the present invention can also be considered to include all possible combinations of the same and / or different embodiments described. Thus, if one embodiment includes features A, B, and C, and another embodiment includes features B and D, then the present invention should also be considered to include embodiments including one or more all other possible combinations of A, B, C, and D, although such embodiments may not be explicitly described in the following content in words.

[0029] The following description provides examples and does not limit the scope, applicability, or examples set forth in the claims. Changes can be made to the functions and arrangements of the described elements without departing from the scope of the content of the present invention. Each example can appropriately omit, substitute, or add various processes or components. For example, the described method can be executed in a different order from the described order, and various steps can be added, omitted, or combined. In addition, the features described in some examples can be combined into other examples.

[0030] To facilitate a better understanding of the embodiments of the present invention, before explaining the detailed implementation manners of the present invention in detail, its application scenarios will be described first.

[0031] The load forecasting method described in the embodiments of this specification is applied to the processes of power system dispatching, power generation planning, and new energy grid connection management. In these scenarios, the application of the load forecasting method aims to balance the supply and demand relationship, reduce the waste of reserve capacity, and improve the consumption capacity of a high-proportion renewable energy power grid through high-precision load forecasting.

[0032] Embodiment 1:

[0033] As Figures 1 - 3As shown in the figure, this embodiment provides a load forecasting method based on a CPO-VMD-MLKELM combined model, including the steps of: S1. Obtain the historical load data, meteorological feature data, and date feature data of the forecasting target and perform preprocessing to obtain a preprocessed data set; S2. Use PCA principal component analysis to perform feature dimensionality reduction on the preprocessed data set, and retain the principal components with a cumulative variance contribution rate of not less than 98% to obtain the dimensionality-reduced principal component features; S3. Use the crown porcupine optimization algorithm to dynamically optimize the number of modes K and the penalty factor of the VMD parameters through four defense strategies , and decompose the historical load data into multiple IMF components based on the optimized VMD parameters. The decomposition formula is , where x is the original signal and N is the number of IMFs; S4. Construct an MLKELM model for each IMF component, use the dimensionality-reduced principal component features as the input to predict each IMF component, and superimpose the predicted values of each IMF component to obtain the load prediction value. The MLKELM model includes a stacked KELM autoencoder and a KELM regressor. More specifically, each has a frequency and amplitude modulation signal . In the above formula, is the frequency, is the amplitude. The expression of the observed signal x0 is , where x is the original signal, is the additive Gaussian white noise. Use Tikhonov regularization to restore the original signal to , and then obtain and solve the frequency-domain Euler–Lagrange equation . In the formula, is the variance of the white noise, is the Fourier transform of the observed signal. The expression of the constrained optimization is to determine the bandwidth of a mode. In the formula, is the center frequency, is the Hilbert transform. To convert the constrained optimization model into an unconstrained form, define the penalty factor a and the Lagrange multiplier k, and then reorganize the model . More specifically, in the unsupervised stage of the MLKELM model, the first autoencoder randomly maps the input original data to an ELM hidden space. The subsequent autoencoders use this random space as the input data and train each stacked ELM autoencoder separately. When the training is completed, the output expression of each hidden layer is , where g is the activation function, is the output weight matrix of the i-th autoencoder, and are the outputs of the i-th layer and the i-1-th layer, To remap the hidden layer of the (i - 1)-th layer to the i-th layer. Each of the autoencoders is an independent feature extractor. As the number of autoencoders increases, the obtained features will become more compact. Once these autoencoders are trained separately, the parameters of the stacked autoencoder structure are fixed and do not require any fine-tuning. The kernel matrix of the KELM is , and the solution of the KELM is . The autoencoder of the kernel extreme learning machine maps the input data to the hidden matrix through a kernel function and then learns the transformation matrix, which transforms the hidden representation into the output data. The transformation matrix can be learned using

[0034] , and the input data of the KELM autoencoder of the (i + 1)-th layer can be calculated . The multi-layer KELM performs KELM-based regression in the second stage. The final representation matrix constitutes the input of the KELM regression model. The training process in the final stage is as follows , where T is the target matrix, is the kernel representation matrix, is the output weight matrix in the second stage, is calculated as , where I is the identity matrix and C is the regularization coefficient. The predicted values of each of the IMF components are stacked to obtain the load prediction value. The MLKELM model includes a stacked KELM autoencoder and a KELM regressor.

[0035] Specifically, this embodiment provides a preferred implementation manner. The preprocessing in step S1 includes performing one-hot encoding on the date feature data. The one-hot encoding is specifically as follows: a working day is encoded as 0, and a holiday is encoded as 1; the seasonal features are encoded as [1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], and [0, 0, 0, 1] for spring, summer, autumn, and winter respectively.

[0036] Specifically, this embodiment provides a preferred implementation manner. The four defense strategies are visual defense, auditory defense, odor defense, and physical attack respectively; the visual defense updates the parameters based on the current optimal solution and a normally distributed random number; the auditory defense determines the search direction through the positions of two random individuals; the odor defense uses a diffusion factor to control the parameter adjustment range; the physical attack corrects the parameters based on the non-elastic collision theorem. More specifically, dynamically optimizing the VMD parameters with the four defense strategies using the crown porcupine optimization algorithm further includes population initialization, that is, each crown porcupine in the crown porcupine group represents a candidate solution. The mathematical expression for the initialization of the crown porcupine group is , where i = 1, 2, …, N, N is the population size, and x iis the position of the i-th individual, rand is a random number between [0, 1], UB and LB are the upper and lower limits of the search interval respectively. CPO introduces a cyclic population reduction strategy. By allowing some crested porcupines to leave the group during the optimization process and then rejoin the group to improve population diversity, its convergence speed is thus accelerated. Its mathematical expression is , where C is the number of cycles. In this embodiment, C = 2 is preferably selected, is the current iteration number, T max is the maximum iteration number, mod represents the modulo operation, N min is the minimum value of the individuals in the newly generated population, generally taking 0.8 times of N. The first defense strategy among the four defense strategies is that the crested porcupine raises and flaps its feathers to warn the predator. The mathematical model is , where is the iteration when the -th individual's position, is the current optimal solution, is a random number based on the normal distribution, is a random number on [0, 1], is the position of the predator at -th iteration. The mathematical expression is , where is a random integer on [1, N]. The second defense strategy among the four defense strategies is that the crested porcupine makes noise and further threatens the predator. The mathematical model is , where and are the positions of two randomly selected crested porcupines, is a random number on [0, 1], randomly takes 0 or 1, y is the position of the predator, between the current crested porcupine and the crested porcupine randomly selected from the population. When = 0, , which means that the second defense strategy has successfully intimidated the predator and the predator no longer moves towards the crested porcupine. When = 1, it means that the predator is moving towards the crested porcupine, but the predator can either approach the crested porcupine or move away from it. The second defense strategy uses two random individuals and to determine whether the predator approaches or moves away from the crested porcupine. When , the predator will approach the crested porcupine. When , the predator will move away from the crested porcupine. The third defense strategy among the four defense strategies is that the crested porcupine will secrete a stench to spread in the surrounding area to prevent the predator from approaching further. The mathematical model is , where r3 is a random integer on [1, N], is a random number on [0, 1], is a parameter for controlling the search direction, and is defined using the formula ; is the defense factor and is defined using the formula ; is the odor diffusion factor. In this embodiment, the range of the odor diffusion factor is preferably [0.3, 2.6]. When U1 = 0, the crested porcupine stops odor diffusion, and the predator stops moving because it is afraid of the crested porcupine. At this time, the distance between the predator and the crested porcupine remains unchanged. When U1 = 1, the crested porcupine emits an obvious odor. The fourth defense strategy among the four defense strategies is that when all previous strategies fail, the crested porcupine will launch a physical attack on the predator, and the mathematical model is , where is the convergence speed factor. In this embodiment, it is preferably = 0.2, is a random number on [0, 1], is the average force of the crested porcupine attacking the i-th predator, which is calculated according to the inelastic collision theorem.

[0037] Specifically, this embodiment provides a preferred implementation manner, and the value range of the diffusion factor in the odor defense is [0.3, 2.6].

[0038] Specifically, this embodiment provides a preferred implementation manner. The stacked KELM autoencoder includes a three-layer structure, and the number of nodes in each layer decreases in the ratio of 60%, 40%, and 20%.

[0039] Specifically, this embodiment provides a preferred implementation manner. The KELM regressor adopts an RBF kernel function, and its bandwidth parameter is optimized and determined by the grid search method.

[0040] Specifically, this embodiment provides a preferred implementation manner, and also includes evaluating the MLKELM model using the ten-fold cross-validation method.

[0041] Embodiment 2:

[0042] This embodiment provides a load forecasting system based on a CPO-VMD-MLKELM combined model for implementing the load forecasting method as described in Embodiment 1.

[0043] Embodiment 3:

[0044] As Figure 4 shown, this embodiment provides an electronic device, which may include: at least one processor, at least one network interface, a user interface, a memory, and at least one communication bus.

[0045] Among them, the communication bus can be used to realize the connection and communication of each of the above components.

[0046] Among them, the user interface may include buttons, and the optional user interface may further include standard wired interfaces and wireless interfaces.

[0047] Among them, the network interface can but is not limited to including a Bluetooth module, an NFC module, a Wi-Fi module, etc.

[0048] Among them, the processor may include one or more processing cores. The processor uses various interfaces and lines to connect each part within the entire electronic device, and by running or executing instructions, programs, code sets, or instruction sets stored in the memory, and by calling the data stored in the memory, it executes various functions of the electronic device and processes data. Optionally, the processor may be implemented in at least one of the hardware forms of DSP, FPGA, and PLA. The processor may integrate one or a combination of several of CPU, GPU, and modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above modem may not be integrated into the processor and may be implemented separately by a single chip.

[0049] Among them, the memory may include RAM and may also include ROM. Optionally, the memory includes a non-transitory computer-readable medium. The memory can be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing each of the above method embodiments, etc.; the data storage area may store the data involved in each of the above method embodiments. Optionally, the memory may further be at least one storage device located far from the aforementioned processor. The memory as a computer storage medium may include an operating system, a network communication module, a user interface module, and a prediction application program. The processor may be used to call the prediction application program stored in the memory and execute the steps of the load prediction method mentioned in the foregoing embodiments.

[0050] Embodiment 4:

[0051] This embodiment provides a computer-readable storage medium, in which instructions are stored. When it runs on a computer or a processor, it causes the computer or the processor to execute the above Figure 1Steps of one or more of the illustrated embodiments. When each component module of the above electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in the computer-readable storage medium.

[0052] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of this specification are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (for example, floppy disk, hard disk, magnetic tape), an optical medium (for example, Digital Versatile Disc (DVD)), or a semiconductor medium (for example, Solid State Disk (SSD)), etc.

[0053] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above Embodiment 1 can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The foregoing storage medium includes various media that can store program codes, such as ROM, RAM, magnetic disk, or optical disc. Without conflict, the technical features in this embodiment and the implementation scheme can be combined arbitrarily.

[0054] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0055] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not described in detail in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0056] The above are only exemplary embodiments of the present invention and should not be used to limit the scope of the present invention. That is, any equivalent changes and modifications made in accordance with the teachings of the present invention still fall within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the disclosure herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not recorded in the present invention. The specification and embodiments are only regarded as exemplary, and the scope and spirit of the present invention are defined by the claims.

Claims

1. A load forecasting method based on the CPO-VMD-MLKELM combined model, characterized in that: Includes steps: S1, obtaining historical load data, meteorological characteristic data and date characteristic data of the forecast target and preprocessing them to obtain a preprocessed data set; S2, using PCA principal component analysis to perform feature dimensionality reduction on the preprocessed data set, and retaining the principal components with cumulative variance contribution rate not less than 98% to obtain the principal component features after dimensionality reduction; S3, using the crown porcupine optimization algorithm to dynamically optimize VMD parameters through four defense strategies, and decomposing the historical load data into multiple IMF components based on the optimized VMD parameters, wherein the VMD parameters include a modal number K and a penalty factor α; The four defense strategies are visual defense, auditory defense, odor defense and physical attack; The visual defense updates parameters based on the current optimal solution and a normally distributed random number; The auditory defense determines the search direction through two random individual positions; The odor defense adopts a diffusion factor to control the parameter adjustment range; The physical attack is based on the correction parameters of the inelastic collision theorem; S4. Construct an MLKELM model for each of the IMF components, use the principal component features after dimensionality reduction as input to predict each of the IMF components, and superimpose the predicted values ​​of each of the IMF components to obtain a load prediction value. The MLKELM model includes a stacked KELM autoencoder and a KELM regressor.

2. The load forecasting method based on the CPO-VMD-MLKELM combined model according to claim 1, characterized in that: The preprocessing in step S1 includes one-hot encoding the date feature data, and the one-hot encoding is specifically: Weekdays are coded as 0 and holidays are coded as 1; Seasonal features are coded as [1,0,0,0], [0,1,0,0], [0,0,1,0], [0,0,0,1] for spring, summer, autumn, and winter, respectively.

3. The load forecasting method based on the CPO-VMD-MLKELM combined model according to claim 2, characterized in that: The value range of the diffusion factor in the odor defense is [0.3, 2.6].

4. The load forecasting method based on the CPO-VMD-MLKELM combined model according to claim 3 is characterized by: The stacked KELM autoencoder includes a three-layer structure, and the number of nodes in each layer decreases in the ratio of 60%, 40%, and 20%.

5. The load forecasting method based on the CPO-VMD-MLKELM combined model according to claim 4, characterized in that: The KELM regressor adopts the RBF kernel function, and its bandwidth parameter is optimized and determined by the grid search method.

6. The load forecasting method based on the CPO-VMD-MLKELM combined model according to claim 5, characterized in that: It also includes evaluating the MLKELM model using a ten-fold cross validation method.

7. A load forecasting system based on the CPO-VMD-MLKELM combined model, characterized in that: Used to implement the load forecasting method as described in any one of claims 1 to 6.

8. A computer device, comprising a memory, a processor and a computer program, characterized in that: When the computer program is executed by a processor, the load forecasting method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the load forecasting method according to any one of claims 1 to 6 is implemented.

Images