Power grid investment scale prediction method, computing device and storage medium

Abstract

The invention relates to the field of power grid intelligent planning, and discloses a power grid investment scale prediction method, computing equipment and a storage medium, and the method comprises the steps: constructing a least square support vector machine model according to a plurality of key influence factors of the power grid investment scale; setting an optimization function of the least square support vector machine model, and enabling each training sample in the training sample set to be on a hyperplane constructed by the least square support vector machine model; determining an adjustment factor of the least square support vector machine model and a kernel parameter of the kernel function according to a bat algorithm to obtain an optimized least square support vector machine model; and training the optimized least square support vector machine model according to the training sample set. Obtaining a trained least square support vector machine model; and in response to received to-be-predicted key influence factor data, inputting the received to-be-predicted key influence factor data into the trained least square support vector machine model, and determining a predicted power grid investment scale.

Description

Technical Field

[0001] The present invention relates to the field of intelligent power grid planning, and in particular to a power grid investment scale prediction method, computing equipment and storage medium. Background Art

[0002] With the vigorous development of emerging industries such as virtual power plants, load aggregators, and energy storage in my country's new power system, how to more accurately predict the scale of grid investment has become an urgent problem to be solved in order to achieve accurate investment of grid enterprises. In the existing technology, the main consideration is to predict the scale of future grid investment based on historical data. However, the power investment scenario is becoming increasingly complex. Using historical grid investment data to predict the scale of future grid investment is not accurate enough.

[0003] Therefore, a new method for predicting the scale of grid investment is needed. Summary of the invention

[0004] To this end, the present invention provides a method for predicting the scale of power grid investment, in an effort to solve or at least alleviate the above problems.

[0005] According to a first aspect of the present invention, there is provided a method for predicting the scale of power grid investment, which is suitable for execution in a computing device, and the method comprises: constructing a least squares support vector machine model according to a plurality of key influencing factors of the scale of power grid investment; setting an optimization function of the least squares support vector machine model so that each training sample in a training sample set is on a hyperplane constructed by the least squares support vector machine model; determining an adjustment factor of the least squares support vector machine model and a kernel parameter of a kernel function according to a bat algorithm to obtain an optimized least squares support vector machine model; training the optimized least squares support vector machine model according to the training sample set to obtain a trained least squares support vector machine model; and in response to receiving key influencing factor data to be predicted, inputting the data into the trained least squares support vector machine model to determine a predicted scale of power grid investment.

[0006] Optionally, in the method according to the present invention, the key influencing factors of the scale of grid investment include multiple items including electricity sales, grid asset-to-income ratio, internal rate of return, increased power supply per unit investment, GDP, total social fixed asset investment, secondary industry structure, power supply reliability, proportion of renewable energy power generation and carbon dioxide emissions.

[0007] Optionally, in the method according to the present invention, constructing a least squares support vector machine model according to multiple key influencing factors of the power grid investment scale includes: mapping the feature vectors corresponding to the key influencing factors according to a nonlinear mapping function to obtain the mapped feature vectors; determining the least squares support vector machine model corresponding to the hyperplane in the feature space according to the mapped feature vectors: F(X)=wT Φ(X)+b, X is a feature vector generated according to multiple key influencing factors in each training sample, Φ(X) is the mapped feature vector, and f(X) is a least squares support vector machine model.

[0008] Optionally, in the method according to the present invention, the optimization function of the least squares support vector machine model includes: sty i =w T Φ(X i )+b+e i , w and b are the parameters of the least squares support vector machine model, C is the adjustment factor, e, e i is the slack variable, e i is the i-th slack variable, the value range of i is 1-m, m is the number of training samples in the training sample set, X i is the feature vector generated based on multiple key influencing factors in the i-th training sample, Φ(X i ) is the feature vector after the i-th training sample is mapped, y i is the scale of grid investment in the i-th training sample.

[0009] Optionally, in the method according to the present invention, the method also includes: determining a least squares support vector machine model including a kernel function based on the least squares support vector machine model and its optimization function, including: introducing Lagrange multipliers, determining the Lagrangian function of the optimization function according to the Lagrangian multiplier method; determining the dual problem of the optimization function based on the Lagrangian function of the optimization function; constructing a kernel function in the dual problem, solving the dual problem, and determining the least squares support vector machine model including the kernel function.

[0010] Optionally, in the method according to the present invention, the least squares support vector machine model including the kernel function includes: α i is the i-th Lagrange multiplier, the value range of i is 1-m, m is the number of training samples in the training sample set, K(X i ,X j ) is the kernel function, X i and X j are the feature vectors generated according to multiple key influencing factors in the i-th training sample and the feature vectors generated according to multiple key influencing factors in the j-th training sample, b is the parameter of the least squares support vector machine model, f(X) is the least squares support vector machine model, and the kernel function K(X i ,X j )=exp[-||X i -X j || 2 / (2σ 2 )], σ is the kernel parameter.

[0011] Optionally, in the method according to the present invention, determining the adjustment factor of the least squares support vector machine model and the kernel parameter of the kernel function according to the bat algorithm includes: initializing parameters of the bat group, each bat represents a possible value of the adjustment factor and the kernel parameter of the kernel function in the search space; calculating the fitness of each bat, and determining the current optimal value x according to the fitness of each bat * ; If the current number of iterations is less than the maximum number of iterations, update the bat's speed and position; if the bat p is at x p The fitness at x is * If the fitness difference at is less than 0, the loudness and pulse emission frequency are updated; the optimal value is re-determined according to the updated loudness and pulse emission frequency; it is determined whether the termination condition is met, and if so, the optimal value of this iteration is output as the adjustment factor and the kernel parameter of the kernel function.

[0012] Optionally, in the method according to the present invention, updating the loudness and the pulse emission frequency comprises: A t+1 =α t+1 A t , A t and A t+1 They represent the average loudness of all bats when the iteration number is t and t+1, α t and α t+1 They represent the sound wave loudness attenuation coefficients when the iteration times are t and t+1 respectively. and denote the initial pulse frequency of bat p and the pulse frequency at t+1, γ t and γ t+1 They represent the pulse frequency enhancement coefficients when the iteration times are t and t+1, δ t =g(x p )-g(x * ), indicating x p The fitness at x * The difference in fitness at .

[0013] According to a second aspect of the present invention, there is provided a computing device, comprising: one or more processors; a memory; and one or more devices, wherein the one or more devices include instructions for a method for predicting a power grid investment scale.

[0014] According to a third aspect of the present invention, there is provided a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions which, when executed by a computing device, cause the computing device to execute a method for predicting a power grid investment scale.

[0015] The present invention proposes a method for predicting the scale of power grid investment. A least squares support vector machine model is constructed according to multiple key influencing factors of the scale of power grid investment. The optimization function of the least squares support vector machine model is set so that each training sample in the training sample set is on the hyperplane constructed by the least squares support vector machine model. When solving the optimization function, the process of solving the inequality is converted into the process of solving the equation. There is no need to impose constraints on the slack variables, so that the solution process of the algorithm is simplified and the solution time of the algorithm is accelerated. The adjustment factor of the least squares support vector machine model and the kernel parameter of the kernel function are determined according to the bat algorithm to obtain an optimized least squares support vector machine model, thereby improving the global optimization level. The optimized least squares support vector machine model is further trained according to the training sample set to obtain the trained least squares support vector machine model, thereby realizing more accurate prediction of the scale of power grid investment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to achieve the above and related purposes, the present invention describes certain illustrative aspects in conjunction with the following description and the accompanying drawings, which indicate various ways in which the principles disclosed in the present invention can be practiced, and all aspects and their equivalents are intended to fall within the scope of the claimed subject matter. By reading the following detailed description in conjunction with the accompanying drawings, the above and other purposes, features and advantages disclosed by the present invention will become more apparent. Throughout the present disclosure, the same reference numerals generally refer to the same parts or elements.

[0017] Figure 1 A schematic diagram of a computing device 100 according to an exemplary embodiment of the present invention is shown;

[0018] Figure 2 A schematic diagram of a method 200 for predicting power grid investment scale according to an exemplary embodiment of the present invention is shown;

[0019] Figure 3 A schematic diagram of optimizing least squares support vector machine parameters according to an exemplary embodiment of the present invention is shown;

[0020] Figure 4 A schematic diagram of an evolution curve of a least squares support vector machine model optimized by a bat algorithm according to an exemplary embodiment of the present invention is shown;

[0021] Figure 5 A schematic diagram showing comparison of grid investment scale prediction results under different models according to an exemplary embodiment of the present invention is shown. DETAILED DESCRIPTION

[0022] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to enable the scope of the present disclosure to be fully conveyed to those skilled in the art. The same reference numerals generally refer to the same components or elements.

[0023] According to one embodiment of the present invention, a method for predicting the scale of power grid investment of the present invention is suitable for execution in a computing device. Figure 1 A schematic diagram of a computing device according to an exemplary embodiment of the present invention is shown.

[0024] Figure 1 1 shows a block diagram of a computing device according to an exemplary embodiment of the present invention. In a basic configuration, computing device 100 includes at least one processing unit 120 and system memory 110. According to one aspect, depending on the configuration and type of computing device, system memory 110 includes, but is not limited to, volatile storage (e.g., random access memory), non-volatile storage (e.g., read-only memory), flash memory, or any combination of such memories. According to one aspect, system memory 110 includes operating system 111.

[0025] According to one aspect, operating system 111, for example, is suitable for controlling the operation of computing device 100. Furthermore, examples may be practiced in conjunction with graphics libraries, other operating systems, or any other application programs, and are not limited to any particular application or system. Figure 1 This basic configuration is illustrated in FIG by those components within dashed line 115. According to one aspect, computing device 100 has additional features or functionality. For example, according to one aspect, computing device 100 includes additional data storage devices (removable and / or non-removable), such as magnetic disks, optical disks, or tapes.

[0026] As stated above, according to one aspect, the program module 112 is stored in the system memory 110. According to one aspect, the program module 112 can be implemented as one or more computer program products, and the present application does not limit the type of computer program products, for example, it can include: email, word processing application, spreadsheet application, database application, slide show application, painting or computer-aided application, web browser, etc. In some embodiments according to the present application, the computer program / instructions related to the power grid investment scale prediction method are packaged as a computer program product, and when these computer programs / instructions are executed by the processor (i.e., the processing unit 120), the power grid investment scale prediction method according to the present application is implemented.

[0027] According to one aspect, the examples may be practiced on a circuit comprising discrete electronic components, a packaged or integrated electronic chip containing logic gates, a circuit utilizing a microprocessor, or a single chip containing electronic components or a microprocessor. Figure 1 Each or many components shown in can be integrated in a system on chip (SOC) on a single integrated circuit to practice examples. According to one aspect, such a SOC device may include one or more processing units, a graphics unit, a communication unit, a system virtualization unit, and various application functions, all of which are integrated (or "burned") into a chip substrate as a single integrated circuit. When operated via SOC, the functions described in this article can be operated via a dedicated logic integrated with other components of the computing device 800 on a single integrated circuit (chip). Embodiments of the present invention can also be practiced using other technologies capable of performing logical operations (such as AND, OR, and NOT), including but not limited to mechanical, optical, fluid, and quantum technologies. In addition, embodiments of the present invention can be practiced in a general-purpose computer or in any other circuit or system.

[0028] According to one aspect, the computing device 100 may also have one or more input devices 131, such as a keyboard, a mouse, a pen, a voice input device, a touch input device, etc. Output devices 132 may also be included, such as a display, a speaker, a printer, etc. The aforementioned devices are examples and other devices may also be used. The computing device 100 may include one or more communication connections 133 that allow communication with other computing devices 140. Examples of suitable communication connections 133 include, but are not limited to: RF transmitters, receivers, and / or transceiver circuits; Universal Serial Bus (USB), parallel and / or serial ports. The computing device 100 can be communicatively connected to other computing devices 140 via a communication connection 333.

[0029] The embodiment of the present invention also provides a non-transitory readable storage medium, which stores instructions, and the instructions are used to make the computing device perform a method according to an embodiment of the present invention. The readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be a computer-readable instruction, a data structure, a module of a program, or other data. Examples of readable storage media include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, tape disk storage or other magnetic storage device or any other non-transitory readable storage medium.

[0030] According to one aspect, communication media is implemented by computer readable instructions, data structures, program modules, or other data in a modulated data signal (e.g., a carrier wave or other transport mechanism), and includes any information delivery media. According to one aspect, the term "modulated data signal" describes a signal that has one or more characteristics set or changed in a manner that encodes information in the signal. By way of example and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, radio frequency (RF), infrared, and other wireless media.

[0031] It should be noted that, although the computing device shown above only includes the processing unit 120, the system memory 110, the input device 131, the output device 132, and the communication connection 133, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the device may only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0032] Figure 2 FIG. 2 is a schematic diagram showing a method 200 for predicting the scale of power grid investment according to an exemplary embodiment of the present invention. Figure 2 As shown, firstly, step 210 is performed to construct a least squares support vector machine model according to multiple key influencing factors of the grid investment scale.

[0033] Since my country is still in the early stages of building a new power system, the addition of new business forms such as virtual power plants and load aggregators, new business entities, and the development of new interactive modes among various entities of source, grid, load and storage under the background of digitalization and other internal and external influencing factors have made the investment scenarios faced by various entities increasingly complex. Therefore, the present invention comprehensively considers the current status of power grid investment and divides the influencing factors of power grid investment from both internal and external aspects. The influencing factors of power grid investment include internal financial factors, external policies and economic and social development factors, and external power industry factors.

[0034] Among them, 19 indicators were selected as internal financial indicators, namely electricity sales revenue, grid asset-to-income ratio, electricity sales volume, electricity purchase cost, unit electricity supply cost, total profit, return on net assets, internal rate of return, investment recovery period, financing, debt-to-asset ratio, net asset value ratio, debt-to-asset ratio, return on net assets, power supply per unit asset, increased power supply per unit investment, increased load per unit investment, depreciation, and operation and maintenance expenses.

[0035] Fourteen indicators were selected for external policies and economic and social development factors, namely: GDP, total exports, fiscal revenue, fiscal expenditure, producer price index, wage level, total social fixed asset investment, total social population, permanent population, population size, proportion of urban population, secondary industry structure, contribution rate of technological progress, and regional electricity share.

[0036] Sixteen indicators were selected for external power industry factors, namely: total social power generation, power generation, total social power consumption, regional power consumption, standard coal consumption for power generation, line loss rate, power supply reliability, automation coverage, length of 35kV and above transmission circuits, line length, average sales electricity price, on-grid electricity price, sales electricity price, carbon dioxide emissions, proportion of renewable energy power generation, and peak load.

[0037] The above indicators construct a basic influencing factor system for the scale of grid investment. Subsequently, the correlation coefficient between each basic influencing factor and the scale of grid investment is determined, and the basic influencing factor with a correlation coefficient greater than a preset value is taken as a key influencing factor. The present invention does not limit the specific method of determining the correlation coefficient. The preset value obtained can be set to 70%, 80%, etc. The present invention does not limit the specific value-taking method of the preset value, and the preset value can be adjusted according to calculation needs. According to one embodiment of the present invention, the Pearson correlation coefficient method is used to calculate the correlation coefficient between the basic influencing factors and the scale of grid investment, so as to extract key influencing factors, including: electricity sales, grid asset-to-income ratio, internal rate of return, unit investment increase in electricity supply, GDP, total social fixed asset investment, secondary industry structure, power supply reliability, proportion of renewable energy power generation and carbon dioxide emissions.

[0038] According to one embodiment of the present invention, constructing a least squares support vector machine model based on multiple key influencing factors of the power grid investment scale includes: mapping the feature vectors corresponding to the key influencing factors according to a nonlinear mapping function to obtain the mapped feature vectors; and determining the least squares support vector machine model corresponding to the hyperplane in the feature space according to the mapped feature vectors:

[0039] f(X)=w T Φ(X)+b

[0040] X is a feature vector generated according to multiple key influencing factors in each training sample, Φ(X) is the mapped feature vector, and f(X) is a least squares support vector machine model.

[0041] The least squares support vector machine is an improvement on the standard support vector machine based on statistical theory. The present invention applies this method to the prediction of power grid investment scale, uses the data of factors affecting power grid investment, such as key factors, to generate sample vectors, and maps them from the original space to the high-dimensional feature space by selecting a nonlinear mapping function, so as to construct the optimal decision function and achieve the prediction purpose. The main process of least squares support vector machine model construction is as follows:

[0042] Given a sample set as an N-dimensional vector, where N is the number of key influencing factors, the training sample set is {(X1,y1),(X2,y2),…,(X m ,y m )}, where X1, X2, …X i …、X m is the input vector of the training sample, X i is the i-th training sample input vector, y i is the prediction vector of the i-th training sample, i ranges from 1 to m, X1, X2, ..., X m ∈R n , which can be realized as a matrix vector. Each matrix vector includes key influencing factor data: power sales, grid asset-to-income ratio, internal rate of return, power supply per unit investment, GDP, total social fixed asset investment, secondary industry structure, power supply reliability, renewable energy power generation ratio and carbon dioxide emissions. Among them, m is the number of training samples in the training sample set. X1, X2, ..., X m Each matrix vector X in consists of x1, x2, ..., x k , k represents the serial number of the key influencing factor, x k is the kth key influencing factor; the value range of k is 1 to N, for example, x1 is the electricity sales amount, and x2 is the grid asset-to-income ratio.

[0043] y m is the prediction vector for each training sample, ym Specifically including the scale of power grid investment.

[0044] The prediction sample data set of the sample to be predicted is {X m+1 ,X m+2 ,…,X m+n}, n is the number of samples to be predicted in the prediction sample data set.

[0045] The present invention sets the optimal decision function of the least squares support vector machine model to be y=f′(X). At the same time, in order to realize the transformation of the sample space from low dimension to high dimension, the sample vector is mapped from the original space to the high-dimensional feature space through a nonlinear mapping function, and the optimal decision nonlinear function f(x) is obtained in the following form:

[0046] f(X)=w T Φ(X)+b

[0047] In the above formula, w and b are the model parameters of the least squares support vector machine model, where w is the weight coefficient, that is, the normal vector in the hyperplane; b is the bias term of the function, that is, the displacement term of the hyperplane, and Φ(X) is the nonlinear mapping function that transforms the sample space from low dimension to high dimension. By transforming the sample space from low dimension to high dimension, the optimal decision nonlinear function can be solved to maximize the fitting of the sample points in the sample space, thereby improving the prediction accuracy based on the nonlinear mapping function.

[0048] Then, step 220 is executed to set the optimization function of the least squares support vector machine model so that each training sample in the training sample set is on the hyperplane constructed by the least squares support vector machine model.

[0049] According to the principle of structural risk minimization, the optimization function of the least squares support vector machine model is:

[0050]

[0051] sty i =w T Φ(X i )+b+e i

[0052] J(w,b,e) is the optimization function, C is the adjustment factor, e is the slack variable, and e i is the i-th slack variable, the value range of i is 1-m, m is the number of training samples in the training sample set, X i is the feature vector generated based on multiple key influencing factors in the i-th training sample, Φ(X i ) is the feature vector after the i-th training sample is mapped, y i is the scale of grid investment in the i-th training sample. i =wT Φ(X i )+b+e i Indicates that each training sample is on the hyperplane constructed by the least squares support vector machine model.

[0053] The method of the present invention also includes: determining the least squares support vector machine model including the kernel function according to the least squares support vector machine model and its optimization function, including: introducing Lagrange multipliers, determining the Lagrange function of the optimization function according to the Lagrange multiplier method; determining the dual problem of the optimization function according to the Lagrange function of the optimization function; constructing the kernel function in the dual problem, solving the dual problem, and determining the least squares support vector machine model including the kernel function.

[0054] The Lagrangian function is constructed based on the original optimization function as follows:

[0055]

[0056] L(w,b,α,e) is the Lagrangian function, α, α i is the Lagrange multiplier, i=1,2,…,m.

[0057] By changing the Lagrangian function according to the dual transformation method, it can be converted into an unconstrained optimization problem, that is, a dual problem; after alignment and solving, the least squares support vector machine model including the kernel function can be obtained:

[0058]

[0059] α i is the i-th Lagrange multiplier, the value range of i is 1-m, m is the number of training samples in the training sample set, K(X i ,X j ) is the kernel function, X i and X j are the feature vectors generated according to multiple key influencing factors in the i-th training sample and the feature vectors generated according to multiple key influencing factors in the j-th training sample, b is the parameter of the least squares support vector machine model, f(X) is the least squares support vector machine model, and the kernel function K(X) i ,X j )=exp[-||X i -X j || 2 / (2σ 2 )], σ is the kernel parameter.

[0060] Then, step 230 is executed to determine the adjustment factor of the least squares support vector machine model and the kernel parameter of the kernel function according to the bat algorithm to obtain an optimized least squares support vector machine model.

[0061] The firefly algorithm used in the prior art is prone to fall into the local optimal solution in the optimization process, resulting in the problem that the constructed model cannot obtain the global optimal solution, and thus the prediction accuracy of the power grid investment scale is poor.

[0062] The present invention takes into account the limitation that the firefly algorithm used in the prior art is prone to fall into the local optimal solution when solving high-dimensional problems. Therefore, a bat algorithm with adaptive evolution parameters is designed. By introducing the bat algorithm with adaptive evolution parameters, the diversity of the bat population evolution during the algorithm operation is guaranteed, and the parameters can be dynamically modified according to the number of iterations, thereby improving the global optimization level of the bat algorithm.

[0063] The bat algorithm is a swarm intelligence optimization algorithm proposed in recent years by simulating the echolocation of micro-bats in nature. It has the advantages of fast convergence speed and strong robustness. Based on the existing bat algorithm, the present invention considers introducing adaptive evolution parameters to improve the diversity of population evolution, so that the parameters can be dynamically changed according to the number of iterations.

[0064] The values ​​of the adjustment factor C and the kernel parameter σ (i.e., the bandwidth σ of the Gaussian kernel) play a decisive role in the performance of the least squares support vector machine: the regularization parameter C can adjust the complexity of the model structure and the proportion of empirical risk, and the kernel function width parameter σ reflects the range distribution of the training data; and there are certain difficulties in selecting the optimal parameters. Therefore, the present invention adopts the improved bat algorithm to optimize the least squares support vector machine.

[0065] Figure 3 FIG. 4 is a schematic diagram showing a method for optimizing least squares support vector machine parameters according to an exemplary embodiment of the present invention. Figure 3 As shown, first initialize the parameters and bat positions. The initialization parameters include: group size g, speed v p (p=1,2,…,g), loudness A p and the pulse transmission frequency r p , bat position x p , the maximum number of iterations T. v p , A p 、r p 、x p are the speed, loudness, pulse emission frequency and bat position of the pth bat respectively.

[0066] The present invention searches for the values ​​of the adjustment factor C and the kernel parameter σ according to the bat algorithm. Therefore, the search space is 2-dimensional, the range of the adjustment factor C is [10, 1000], and the range of the kernel parameter σ is [0.01, 0.1]. Each individual bat can be represented by (C, σ). P bats are generated in the search space, and the initial position of each bat is randomly determined in the search space.

[0067] Then, calculate the initial fitness. For the randomly initialized bat group position, learn and predict the (C, σ) of the individual bats in the least squares support vector machine model through sample data, and determine the fitness of the bat algorithm based on the relative error between the training value and the predicted value. The smaller the relative error, the higher the fitness, and the larger the relative error, the smaller the fitness. Calculate the fitness of each bat and find the current optimal value x * .

[0068] When the number of iterations is less than the maximum number of iterations, update the bat speed and position:

[0069]

[0070] f p =f min +(f max -f min )β

[0071] and They represent the position of bat p in the search space when the number of iterations is t and t+1 respectively. and They represent the speed of bat p when the iteration number is t and t+1, respectively. p Indicates that in x p The pulse frequency at max and f min is the maximum and minimum value of the pulse frequency at the current moment; β is a random number in the range of [0,1]; x * is the current optimal value.

[0072] If rand1>r p (rand1∈[0,1]), then a local search is performed to generate a new solution, including: after selecting an optimal individual from the current optimal solution set, the bat individual will update its position:

[0073] x new =x * +εA t

[0074] Among them, A t represents the average loudness of all bat individuals at t iterations, ε is a uniformly distributed random number in the interval [-1,1], and x new is the bat individual after updating its position.

[0075] During the iteration process, the fitness of all bats is sorted, and after finding the current optimal solution and optimal value, the volume and pulse frequency of each individual bat are updated until the global optimal solution is output. According to biological mechanisms, when approaching prey, the volume of individual bats will decrease, while the frequency of pulse emission will increase, then A p When it approaches 0, it means that bat p has found a prey and stops making sounds.

[0076] If rand2<A p (rand2∈[0,1]) and g(x p )<g(x * ), then the new solution generated by the reception is recorded as the current optimal solution, and the loudness A is updated as follows t and the pulse transmission frequency r p .

[0077] The update equation is as follows:

[0078] A t+1 =α t+1 a t

[0079]

[0080] a t and a t+1 They represent the average loudness of all bats when the iteration number is t and t+1, α t and α t+1 They represent the sound wave loudness attenuation coefficients when the iteration times are t and t+1 respectively. and denote the initial pulse frequency of bat p and the pulse frequency at t+1, γ t and γ t+1 They represent the pulse frequency enhancement coefficients when the iteration times are t and t+1 respectively. In the present invention, the sound wave loudness attenuation coefficient and the pulse frequency enhancement coefficient are set, the bat (BA) algorithm is improved, and an improved bat (IBA) algorithm is realized.

[0081] Among them, define δ t =g(x p )-g(x * ), indicating x p The fitness at x * The difference in fitness at t <0, it means that the performance of the global optimal solution has declined and the local search capability needs to be strengthened. p ) is x p The fitness at g(x * ) is x * The adaptability of the place.

[0082] Then, sort all bat fitness values ​​and find the current optimal solution x * .

[0083] Then the output is judged. If the maximum number of iterations is reached or the optimal solution condition is met, the algorithm ends and the global optimal value of (C, σ) is output. Otherwise, t = t + 1, and the iteration continues until the termination condition is met. The optimal solution condition can be specifically set as when the optimal solution is obtained, the training value of the least squares support vector machine model is the same as the preset value, or the difference is less than a preset value.

[0084] Subsequently, step 240 is executed to train the optimized least squares support vector machine model according to the training sample set to obtain a trained least squares support vector machine model, that is, the optimal value of (C, σ) is used as the optimal parameter combination to train the support vector machine model, thereby constructing a power grid investment scale prediction model based on the improved bat algorithm optimized least squares support vector machine (IBA-LSSVM)

[0085] Finally, step 250 is executed, in response to receiving the key influencing factor data to be predicted, inputting it into the trained least squares support vector machine model to determine the predicted scale of power grid investment. For example, the key influencing factor data includes: when the scale of power grid investment is to be predicted, the values ​​of power sales, power grid asset income ratio, internal rate of return, power supply per unit investment, GDP, total social fixed asset investment, secondary industry structure, power supply reliability, proportion of renewable energy power generation and carbon dioxide emissions.

[0086] The present invention takes into account the key influencing factors of the new power system on the scale of power grid investment, improves the bat algorithm with stronger global optimization capabilities, and designs a power grid investment scale prediction method based on the improved bat algorithm to optimize the least squares support vector machine. The power grid investment scale prediction method of the present invention can adapt to the complex development trend of future power grid investment scenarios under the new power system, meet the needs of new power system construction, improve the accuracy of power grid investment scale prediction, realize accurate investment of power grid enterprises, promote the construction of new power systems, and further improve power supply access capabilities and power supply capabilities.

[0087] The present invention improves the three-dimensional standard support vector machine algorithm in the prior art, and uses the improved least squares support vector machine algorithm to predict the scale of power grid investment. Compared with the traditional standard support vector machine algorithm, the least squares support vector machine algorithm uses equality constraints instead of inequality constraints, and uses equality constraints for each sample point without imposing any constraints on the relaxation vector, so that the algorithm solution process is simplified and the algorithm solution time is accelerated, so that the present invention can also run efficiently when making predictions based on large data sets.

[0088] The present invention takes into account the limitation that the firefly algorithm used in the prior art is prone to fall into the local optimal solution when solving high-dimensional problems. Therefore, a bat algorithm with adaptive evolution parameters is designed. By introducing the bat algorithm with adaptive evolution parameters, the diversity of the bat population evolution during the algorithm operation is guaranteed, and the parameters can be dynamically modified according to the number of iterations, thereby improving the global optimization level of the bat algorithm.

[0089] In summary, the power grid investment scale prediction method designed by the present invention enriches and improves the existing method and can better meet the current demand for power grid investment prediction.

[0090] The prediction of the scale of power grid investment has the characteristics of high dimension, nonlinearity and small sample. The prediction accuracy of the traditional prediction method is not high and the calculation is complicated. The present invention constructs a power grid investment scale prediction model based on the IBA-LSSVM algorithm, and applies the prediction model to the prediction of the scale of power grid investment of a certain project. The key influencing factor data of the region from 2016 to 2020 are used as training samples, and the key influencing factor data from 2021 to 2023 are used as test samples of the investment scale prediction model, and the changes in the investment of power grid enterprises in this stage with the changes in the power grid structure and the external environment are predicted and analyzed. The power grid investment scale prediction model proposed in the present invention is a nonlinear optimization model, and the MATLAB 2018b software can be used to calculate and analyze the example.

[0091] The data analysis and processing process is as follows: In this embodiment, the data mainly comes from the statistical yearbook and power grid company of a certain region. Based on the analysis of factors affecting investment demand, the power sales volume x1, power grid asset income ratio x2, internal rate of return x3, power supply per unit investment x4, GDP x5, total social fixed asset investment x6, secondary industry structure x7, power supply reliability x8, renewable energy power generation ratio x9, carbon dioxide emissions x1 from 2016 to 2023 in the region are used. 10 The data of ten factors such as y and y are used as the input variable training sample set of the IBA-LSSVM prediction model, and the total investment y is the output variable, and the training samples are normalized. The descriptive statistical analysis of the training sample and the test sample data is shown in the following table.

[0092] Table 1 Some training sample and test sample data

[0093]

[0094]

[0095] Since the units of various indicators are different, the size of their values ​​will have a certain impact on the machine learning process. Therefore, before machine learning, it is necessary to normalize all the data samples of the influencing factors involved in the training, that is, to make them dimensionless. The dimensionless formula is as follows:

[0096]

[0097] Where: x i is the original value of the sample; is the dimensionless value; x max and x m i n are the maximum and minimum values ​​of each group of sample data respectively.

[0098] The training results and analysis are as follows: The present invention establishes an LSSVM power grid investment scale prediction model based on IBA algorithm optimization, selects 10 indicators such as regional GDP as input variables, normalizes the sample data, and then uses the improved BA algorithm to optimize the parameters in the LSSVM model. In order to verify the prediction effect of LSSVM integrated with the improved BA algorithm, the present invention uses the least squares support vector machine (LSSVM), BP neural network and the improved least squares vector machine (IBA-LSSVM) to predict the investment demand of power grid enterprises. The present invention uses the radial basis kernel function (RBF) and optimizes the parameters in the LSSVM prediction model through the IBA algorithm. The maximum number of iterations is initially set to 100 times, that is, after the number of iterations is completed, the parameter optimization automatically terminates.

[0099] Figure 4 FIG. 2 shows a schematic diagram of an evolution curve of a least squares support vector machine model optimized by a bat algorithm according to an exemplary embodiment of the present invention. Figure 4 As shown in the figure, it can be seen from the change of the convergence curve that when the number of iterations is 52, the curve is completely converged. The change trend of the curve shows that the IBA algorithm has high convergence efficiency and fast operation speed. At this time, the obtained LSSVM parameter combination can be considered to be optimal. The optimal solution of the global search of the IBA algorithm, namely the regularization C and the kernel function width parameter σ value, is shown in Table 2.

[0100] Table 2 Optimization model parameter selection results

[0101]

[0102] Figure 5 FIG. 2 shows a schematic diagram comparing the prediction results of the power grid investment scale under different models according to an exemplary embodiment of the present invention. Figure 5As shown in the figure, different models are used to analyze the scale of grid investment in the region from 2016 to 2023, and a curve chart is drawn based on the forecast results.

[0103] Table 3 shows the average error between the actual and predicted values ​​of the three models and the maximum error in the three-year predicted values.

[0104] Table 3 Forecast errors of different investment demand forecasting models

[0105]

[0106] contrast Figure 5 From the prediction results in Table 3, we can see that compared with the BP neural network algorithm with strong learning ability, the two errors of the prediction results of the IBA-LSSVM and LSSVM models are both smaller than those of the BP neural network algorithm, indicating that the least squares support vector machine algorithm is more suitable for the data characteristics of investment demand forecasting than the BP neural network; at the same time, the average relative error and maximum relative error of the prediction results of the IBA-LSSVM model are smaller than those of the LSSVM model under the same kernel function, indicating that the IBA-LSSVM model with better optimization ability can obtain better parameters and solve the investment demand forecasting problem with higher accuracy. Therefore, considering the factors affecting the user-side investment demand after the extension of the investment interface of the power grid enterprise, the optimal parameters are selected in combination with the characteristics of the IBA algorithm with strong optimization ability, and the LSSVM optimization model with higher investment demand prediction accuracy and strong convergence ability is obtained, which can realize more accurate prediction of investment demand and provide a theoretical basis for the investment strategy formulation and investment benefit analysis of the power grid enterprise.

[0107] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0108] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof.

[0109] In the case where the program code is executed on a programmable computer, the computing device generally includes a processor, a storage medium readable by the processor (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device. The memory is configured to store the program code; the processor is configured to execute the user identity authentication method of the present invention according to the instructions in the program code stored in the memory.

[0110] By way of example and not limitation, computer readable media include computer storage media and communication media. Computer readable media include computer storage media and communication media. Computer storage media stores information such as computer readable instructions, data structures, program modules or other data. Communication media generally embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and includes any information delivery medium. Any combination of the above is also included within the scope of computer readable media.

[0111] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and instructional purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art. The disclosure of the present invention is illustrative, rather than restrictive, with respect to the scope of the present invention.

[0112] As used herein, unless otherwise specified, the use of ordinal numbers "first," "second," "third," etc. to describe common objects merely indicates that different instances of similar objects are involved, and is not intended to imply that the objects so described must have a given order in time, space, order, or in any other manner.

[0113] Although the present invention has been described according to a limited number of embodiments, it will be apparent to those skilled in the art, with the benefit of the above description, that other embodiments may be envisioned within the scope of the invention thus described. In addition, it should be noted that the language used in this specification is selected primarily for readability and instructional purposes, rather than for explaining or defining the subject matter of the present invention. Therefore, many modifications and variations will be apparent to those skilled in the art. The disclosure of the present invention is illustrative, rather than restrictive, with respect to the scope of the present invention.

Claims

1. A method for predicting the scale of power grid investment, suitable for execution in a computing device, the method comprising: A least squares support vector machine model is constructed based on multiple key influencing factors of power grid investment scale; Setting the optimization function of the least squares support vector machine model so that each training sample in the training sample set is on the hyperplane constructed by the least squares support vector machine model; According to the bat algorithm, the adjustment factor of the least squares support vector machine model and the kernel parameter of the kernel function are determined to obtain the optimized least squares support vector machine model; Training the optimized least squares support vector machine model according to the training sample set to obtain a trained least squares support vector machine model; In response to receiving the key influencing factor data to be predicted, the data is input into the trained least squares support vector machine model to determine the predicted scale of power grid investment.

2. The method of claim 1, wherein: The key influencing factors of the scale of grid investment include electricity sales, grid asset-to-income ratio, internal rate of return, increased power supply per unit investment, GDP, total social fixed asset investment, secondary industry structure, power supply reliability, proportion of renewable energy power generation and carbon dioxide emissions.

3. The method according to claim 1 or 2, wherein: The least squares support vector machine model is constructed based on multiple key factors affecting the scale of power grid investment, including: Mapping the characteristic vector corresponding to the key influencing factor according to a nonlinear mapping function to obtain a mapped characteristic vector; Determine the least squares support vector machine model corresponding to the hyperplane in the feature space according to the mapped feature vector: f(X)=w T Φ(X)+b X is a feature vector generated according to multiple key influencing factors in each training sample, Φ(X) is the mapped feature vector, and f(X) is a least squares support vector machine model.

4. The method of claim 3, wherein: The optimization function of the least squares support vector machine model includes: s.t.y i =w T Φ(X i )+b+e i w and b are the parameters of the least squares support vector machine model, C is the adjustment factor, e and e i is the slack variable, e i is the i-th slack variable, the value range of i is 1-m, m is the number of training samples in the training sample set, X i is the feature vector generated based on multiple key influencing factors in the i-th training sample, φ(X i ) is the feature vector after the i-th training sample is mapped, y i is the scale of grid investment in the i-th training sample.

5. The method of claim 1, wherein: The method further comprises: According to the least squares support vector machine model and its optimization function, a least squares support vector machine model including a kernel function is determined, including: Introduce Lagrange multipliers, and determine the Lagrange function of the optimization function according to the Lagrange multiplier method; Determine the dual problem of the optimization function according to the Lagrangian function of the optimization function; A kernel function is constructed in the dual problem, the dual problem is solved, and a least squares support vector machine model including the kernel function is determined.

6. The method of claim 5, wherein: The least squares support vector machine model including the kernel function includes: α i is the i-th Lagrange multiplier, the value range of i is 1-m, m is the number of training samples in the training sample set, K(x i ,x j ) is the kernel function, X i and X j are the feature vectors generated according to multiple key influencing factors in the i-th training sample and the feature vectors generated according to multiple key influencing factors in the j-th training sample, b is the parameter of the least squares support vector machine model, f(X) is the least squares support vector machine model, and the kernel function K(x i ,x j )=exp[-||X i -X j || 2 / (2σ 2 )], σ is the kernel parameter.

7. The method of claim 1, wherein: The adjustment factor of the least squares support vector machine model and the kernel parameter of the kernel function determined according to the bat algorithm include: Initialize the parameters of the bat population, and each bat represents a possible value of the adjustment factor and the kernel parameter of the kernel function in the search space; Calculate the fitness of each bat and determine the current optimal value x based on the fitness of each bat * ; If the current number of iterations is less than the maximum number of iterations, update the bat's speed and position; If bat p is at x p The fitness at x is * If the fitness difference at is less than 0, the loudness and pulse emission frequency are updated; Re-determine the optimal value according to the updated loudness and pulse emission frequency; Determine whether the termination condition is met. If so, output the optimal value of this iteration as the adjustment factor and the kernel parameter of the kernel function.

8. The method of claim 7, wherein: The updating loudness and pulse emission frequency includes: A t+1 =a t+1 a t A t and A t+1 They represent the average loudness of all bats when the iteration number is t and t+1, α t and α t+1 They represent the sound wave loudness attenuation coefficients when the iteration times are t and t+1 respectively. and denote the initial pulse frequency of bat p and the pulse frequency at t+1, γ t and γ t+1 They represent the pulse frequency enhancement coefficients when the iteration times are t and t+1, δ t =g(x p )-g(x * ), indicating x p The fitness at x * The fitness difference at .

9. A computing device, characterized in that include: one or more processors; Memory; as well as One or more devices comprising instructions for executing the method according to any one of claims 1-8.

10. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions which, when executed by a computing device, cause the computing device to perform a method according to any one of claims 1-8.

Images