Port load prediction method and system considering multi-dimensional impact factors

Abstract

The invention discloses a port load prediction method and system considering a multi-dimensional impact factor, and the method comprises the steps: selecting the multi-dimensional impact factor affecting the port load prediction, carrying out the processing of the bad data of a port, and carrying out the normalization processing of the obtained new data; dividing the normalized data into a training sample and a prediction sample, and constructing a suitability function; performing performance improvement on the white whale optimization algorithm to obtain an improved white whale optimization algorithm; a penalty factor, an insensitive loss function and a radial basis function parameter in a support vector machine are optimized based on an improved white whale optimization algorithm, a port load time sequence and meteorological dimension influence factor data are input based on a training sample and a prediction sample, the port load is predicted based on the support vector machine, and a prediction result is obtained. The port load prediction precision can be effectively improved, the energy utilization rate is improved, and the port operation cost is reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of port load prediction, and relates to a port load prediction method and system taking multi-dimensional influencing factors into account. Background Art

[0002] With the rapid development of social economy, people's demand for energy is also increasing. Ports are an important type of energy system in the country, affecting the development of the national economy. Load forecasting is an important basis for realizing the economic dispatch of port energy systems. The timeliness and accuracy of load forecasting have a great impact on the economic and low-carbon operation of port energy systems. If the load forecast value is too low, it will lead to tight energy supply and fail to meet the needs of port operations, greatly reducing the reliability of port operations; if the load forecast value is too high, it will lead to a decrease in the operating efficiency of power generation, transmission, and transformation equipment after they are put into the system, seriously affecting the economic indicators of the system. Due to the variety of load types in port scenarios, and the presence of impact load equipment such as cranes, in addition, port loads are also affected by many factors such as weather, date, time, throughput, etc., showing the characteristics of large randomness and long-term fluctuations. Many existing prediction methods cannot be directly applied to port scenarios to predict the load values ​​of ports. Summary of the invention

[0003] The purpose of the present invention is to solve the problem in the prior art that the port load is easily affected by many factors such as weather, date, time, throughput, etc., and the existing prediction methods cannot be directly applied, and to provide a port load prediction method and system taking into account multi-dimensional influencing factors.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] A port load forecasting method taking into account multi-dimensional influencing factors includes:

[0006] Select multi-dimensional influencing factors that affect port load, including port load category dimension influencing factors, and select port load time series and meteorological dimension influencing factors based on principal component-grey correlation;

[0007] Process bad port data and normalize the new data obtained;

[0008] The normalized data is divided into training samples and prediction samples, and the fitness function is constructed to minimize the total error of the training sample set;

[0009] Improve the performance of the White Whale optimization algorithm and obtain the improved White Whale optimization algorithm;

[0010] Based on the improved Beluga optimization algorithm, the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine are optimized. Based on the training samples and prediction samples, the port load time series and meteorological dimension influencing factor data are input, and the port load is predicted based on the support vector machine to obtain the prediction result.

[0011] A further improvement of the present invention is:

[0012] Furthermore, multi-dimensional influencing factors affecting port load are selected, including the influencing factors of port load category dimension, and the influencing factors of port load time series and meteorological dimension based on principal component-grey correlation are selected, specifically:

[0013] Statistics are collected on various types of port loads, and the loads with the largest energy consumption, the most obvious time series changes, and the closest correlation with port throughput are selected as the influencing factors for port load prediction;

[0014] Construct an index system based on historical load energy consumption values ​​and meteorological factors;

[0015] Based on principal component analysis and grey correlation method, the principal component-grey correlation analysis method is constructed;

[0016] Based on the principal component-grey correlation analysis method, the indicator system is reduced in dimension to obtain the important influencing factors of short-term load.

[0017] Furthermore, the bad port data is processed and the obtained new data is normalized, specifically:

[0018] Improve the wavelet denoising threshold function to obtain an improved wavelet denoising method;

[0019] In view of the shortcomings of the soft and hard threshold methods in the wavelet denoising threshold function, the weighted average method is used to estimate the wavelet coefficients. The improved formula is:

[0020]

[0021] In the formula, w j,k and are the wavelet coefficients of the noisy signal after wavelet transformation and the estimated wavelet coefficients; T is the threshold; β is the weighting factor; sgn() is the sign function; j is the number of decomposition levels;

[0022] The original port load data is denoised based on the improved wavelet denoising method, and the defective data is repaired.

[0023] The repaired data is normalized, and the normalization formula is:

[0024]

[0025] In the formula, is the normalized data value; xi is the original value; x imin is the minimum value in the column; x imax is the maximum value of the column; l is the dimension of the input vector.

[0026] Furthermore, the fitness function is:

[0027]

[0028] In the formula, y i is the actual load value; is the predicted load value; n represents the number of predicted loads.

[0029] Furthermore, the performance of the White Whale optimization algorithm is improved to obtain an improved White Whale optimization algorithm, which is as follows:

[0030] Improve the beluga whale population position initialization strategy to evenly distribute the beluga whale positions;

[0031] Introducing dynamic weights in the process of updating the beluga whale's position to optimize the optimal solution for the beluga whale's position;

[0032] Based on the optimal de-disturbance strategy, the position of the beluga whale in the whale fall phase is updated during the beluga whale's hunting process.

[0033] Furthermore, the beluga whale population position initialization strategy is improved to evenly distribute the beluga whale positions, specifically:

[0034] The position matrix of the white whale in the white whale optimization algorithm is as follows:

[0035]

[0036] Where n represents the number of beluga whales in the population; x i is the white whale position; D is the dimension of the problem to be solved;

[0037] For all beluga whales, the corresponding fitness values ​​are stored as follows:

[0038]

[0039] Where, fix is ​​the fitness matrix of the beluga population; fix n Corresponding to the fitness value of the nth individual in the beluga population;

[0040] The initialization strategy is introduced to enhance the exploration performance of the White Whale optimization algorithm. The function expression of the improved initialization strategy is:

[0041]

[0042] Wherein, h is the adjustment parameter of uniform distribution and h=0.5.

[0043] Furthermore, dynamic weights are introduced in the process of updating the beluga position to optimize the optimal solution of the beluga position. Specifically, the beluga shares each other's position with nearby beluga whales, and determines that the beluga whale with the best distance to the prey is the best candidate, and the other beluga whales are other candidates. The beluga position is updated in real time to continuously determine the best candidate. The beluga position update formula is as follows:

[0044]

[0045] μ=2r2((1-T) / T max ) (9)

[0046] L F =0.05×(δ×γ) / |μ 1 / B (10)

[0047]

[0048] In the formula, and are the position of the i-th beluga whale and the position of a random beluga whale at the current iteration number, respectively; r1 and r2 are random numbers between (0,1); The position of the best white whale is the global optimal position; λ represents the random jump intensity used to measure the flight intensity in the Levy flight strategy; T max is the maximum number of iterations; L F is the Levy flight function; δ and μ are normally distributed random numbers; B is a constant whose value is 1.5;

[0049] Dynamic weights are introduced into the White Whale position update formula. A larger weight in the early development stage is beneficial to global search, while a lower weight in the later stage is more conducive to White Whale searching for the optimal solution within a certain range. The weight formula is as follows:

[0050]

[0051] After the dynamic weight w is introduced, the position update formula of the white whale is as follows:

[0052]

[0053] Furthermore, based on the optimal disturbance resolution strategy, the position of the beluga whale in the whale fall phase is updated during the beluga whale hunting process, specifically:

[0054] Beluga whales may experience whale falls when they are moving or hunting. The mathematical model of whale falls is as follows:

[0055]

[0056] Where r3, r4 and r5 are random numbers between (0,1), X s is the step length of the whale fall; u b and l b are the upper and lower bounds of the variable respectively; ρ f is the probability of a whale fall;

[0057] Introducing small-amplitude neighborhood disturbances can improve the ability of the White Whale optimization algorithm to escape from local optimality. The formula for small-amplitude neighborhood disturbances is as follows:

[0058]

[0059] Among them, c1 is a variable disturbance factor. As the number of iterations increases, the disturbance factor gradually decreases, which helps the algorithm converge to the optimal solution in the later stage. The improved position update formula in the whale fall stage is as follows:

[0060]

[0061] In the formula, rand is a random number between 0 and 1; CL is the probability factor, and the expression is as follows:

[0062]

[0063] Furthermore, the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine are optimized based on the improved Beluga optimization algorithm. Based on the training samples and prediction samples, the port load time series and meteorological dimension influencing factor data are input, and the port load is predicted based on the support vector machine to obtain the prediction results, which are as follows:

[0064] The output y of the processed data model is the load energy consumption value to be predicted, and the input vector port load time series and meteorological dimension influencing factor data;

[0065]

[0066] Among them, matrix A is the port load energy consumption data at the prediction time of the day before the prediction day, and each row represents the selected different important load energy consumption load types; matrix B is the port load energy consumption data at the moment before the prediction time of the day before the prediction day, and each row represents different load energy consumption load types; vector C is meteorological data, including maximum temperature, minimum temperature, average temperature, etc.; in vector D, d1 represents working days, and d2 represents rest days;

[0067] Set the parameters of the improved Beluga optimization algorithm, including the number of individuals in the population, dimension, and maximum number of iterations;

[0068] The optimal penalty factor C, insensitive loss function ε and radial basis function parameter σ of the support vector machine obtained based on the White Whale optimization algorithm;

[0069] A regression objective function is established based on the training samples and minimized; the regression objective function is:

[0070]

[0071] In the formula, α i and α i * (i=1,2,…,l) is the Lagrange multiplier; K(x i ,x j )=φ(x i )·φ(x j ) is the kernel function; minimize the regression objective function and get the optimal solution as

[0072] Substituting the optimal solution into equation (22), we obtain the decision regression equation;

[0073]

[0074] In the formula, The calculation formula is:

[0075]

[0076] The port load at future times is predicted based on the forecast samples and the obtained decision regression equation.

[0077] A port load forecasting system taking into account multi-dimensional influencing factors, comprising:

[0078] A selection module, wherein the selection module selects multi-dimensional influencing factors affecting port load, including port load category dimension influencing factors, and selects port load time series and meteorological dimension influencing factors based on principal component-grey correlation;

[0079] A processing module, which processes the bad port data and normalizes the obtained new data;

[0080] A construction module, wherein the construction module divides the normalized data into training samples and prediction samples, and constructs a fitness function so that the total error of the training sample set is minimized;

[0081] An improvement module, wherein the improvement module improves the performance of the White Whale optimization algorithm to obtain an improved White Whale optimization algorithm;

[0082] The prediction module optimizes the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine based on the improved Beluga optimization algorithm, inputs the port load time series and meteorological dimension influencing factor data based on the training samples and the prediction samples, predicts the port load based on the support vector machine, and obtains the prediction result.

[0083] Compared with the prior art, the present invention has the following beneficial effects:

[0084] The present invention selects multi-dimensional influencing factors that affect port load prediction, improves the White Whale optimization algorithm, optimizes the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine, inputs the influencing factor data, and uses the support vector machine to predict the port load. The present invention can effectively improve the port load prediction accuracy, improve energy utilization and reduce port operation costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0086] Figure 1 It is a schematic flow chart of a port load forecasting method taking into account multi-dimensional influencing factors of the present invention;

[0087] Figure 2 It is a structural schematic diagram of a port load forecasting system taking into account multi-dimensional influencing factors of the present invention;

[0088] Figure 3 This is a curve chart of monthly energy consumption values ​​for various loads at a port in Jiangsu;

[0089] Figure 4 This is a monthly throughput curve of a port in Jiangsu;

[0090] Figure 5 It is the flow chart of the principal component-grayscale correlation method for extracting port influencing factors;

[0091] Figure 6 This is a curve chart of port load energy consumption data for a port in Jiangsu Province in June 2024.

[0092] Figure 7 This is a schematic diagram of the load forecast result curve of the port predicted by the model for a single day;

[0093] Figure 8 It is a schematic diagram of the relative error curve of the model's prediction of the port's load for a single day. DETAILED DESCRIPTION

[0094] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0095] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0096] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0097] In the description of the embodiments of the present invention, it should be noted that if the terms "upper", "lower", "horizontal", "inner", etc. indicate an orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is usually placed when in use, it is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0098] In addition, if the term "horizontal" appears, it does not mean that the component must be absolutely horizontal, but can be slightly tilted. For example, "horizontal" only means that its direction is more horizontal than "vertical", which does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0099] In the description of the embodiments of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0100] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0101] See also Figure 1 The present invention discloses a port load forecasting method taking into account multi-dimensional influencing factors, including:

[0102] S101, selecting multi-dimensional influencing factors affecting port load, including port load category dimension influencing factors, and selecting port load time series and meteorological dimension influencing factors based on principal component-grey correlation;

[0103] Statistics are collected on various types of port loads, and the loads with the largest energy consumption, the most obvious time series changes, and the closest correlation with port throughput are selected as the influencing factors for port load prediction;

[0104] Construct an index system based on historical load energy consumption values ​​and meteorological factors;

[0105] Based on principal component analysis and grey correlation method, the principal component-grey correlation analysis method is constructed;

[0106] Based on the principal component-grey correlation analysis method, the indicator system is reduced in dimension to obtain the important influencing factors of short-term load.

[0107] S102, processing bad port data and normalizing the obtained new data;

[0108] S102.1: improving the wavelet denoising threshold function to obtain an improved wavelet denoising method;

[0109] In view of the shortcomings of the soft and hard threshold methods in the wavelet denoising threshold function, the weighted average method is used to estimate the wavelet coefficients. The improved formula is:

[0110]

[0111] In the formula, w j,k and are the wavelet coefficients of the noisy signal after wavelet transformation and the estimated wavelet coefficients; T is the threshold; β is the weighting factor; sgn() is the sign function; j is the number of decomposition levels;

[0112] S102.2: De-noise the original port load data based on the improved wavelet denoising method and repair the defective data;

[0113] S102.3: Normalize the repaired data. The normalization formula is:

[0114]

[0115] In the formula, is the normalized data value; xi is the original value; x imin is the minimum value in the column; x imax is the maximum value of the column; l is the dimension of the input vector.

[0116] S103, dividing the normalized data into training samples and prediction samples, and constructing a fitness function so that the total error of the training sample set is minimized;

[0117] The fitness function is:

[0118]

[0119] In the formula, y i is the actual load value; is the predicted load value; n represents the number of predicted loads.

[0120] S104, improving the performance of the White Whale optimization algorithm to obtain an improved White Whale optimization algorithm;

[0121] S104.1, improve the beluga population position initialization strategy to evenly distribute the beluga positions;

[0122] The position matrix of the white whale in the white whale optimization algorithm is as follows:

[0123]

[0124] Where n represents the number of beluga whales in the population; x i is the white whale position; D is the dimension of the problem to be solved;

[0125] For all beluga whales, the corresponding fitness values ​​are stored as follows:

[0126]

[0127] Where, fix is ​​the fitness matrix of the beluga population; fix n Corresponding to the fitness value of the nth individual in the beluga population;

[0128] The initialization strategy is introduced to enhance the exploration performance of the White Whale optimization algorithm. The function expression of the improved initialization strategy is:

[0129]

[0130] Wherein, h is the adjustment parameter of uniform distribution and h=0.5.

[0131] S104.2, dynamic weights are introduced in the process of updating the beluga position to optimize the optimal solution of the beluga position. The beluga shares each other's position with nearby beluga whales, and determines that the beluga whale with the best distance from the prey position is the best candidate, and the other beluga whales are other candidates. The beluga position is updated in real time to continuously determine the best candidate; wherein, the beluga position update formula is as follows:

[0132]

[0133] μ=2r2((1-T) / T max ) (9)

[0134] L F =0.05×(δ×γ) / |μ 1 / B (10)

[0135]

[0136] In the formula, and are the position of the i-th beluga whale and the position of a random beluga whale at the current iteration number, respectively; r1 and r2 are random numbers between (0,1); The position of the best white whale is the global optimal position; λ represents the random jump intensity used to measure the flight intensity in the Levy flight strategy; T max is the maximum number of iterations; L F is the Levy flight function; δ and μ are normally distributed random numbers; B is a constant whose value is 1.5;

[0137] Dynamic weights are introduced into the White Whale position update formula. A larger weight in the early development stage is beneficial to global search, while a lower weight in the later stage is more conducive to White Whale searching for the optimal solution within a certain range. The weight formula is as follows:

[0138]

[0139] After the dynamic weight w is introduced, the position update formula of the white whale is as follows:

[0140]

[0141] S104.3, based on the optimal disturbance resolution strategy, the position of the beluga whale in the whale fall phase is updated during the beluga whale hunting process.

[0142] Beluga whales may experience whale falls when they are moving or hunting. The mathematical model of whale falls is as follows:

[0143]

[0144] Where r3, r4 and r5 are random numbers between (0,1), X sis the step length of the whale fall; u b and l b are the upper and lower bounds of the variable respectively; ρ f is the probability of a whale fall;

[0145] Introducing small-amplitude neighborhood disturbances can improve the ability of the White Whale optimization algorithm to escape from local optimality. The formula for small-amplitude neighborhood disturbances is as follows:

[0146]

[0147]

[0148] Among them, c1 is a variable disturbance factor. As the number of iterations increases, the disturbance factor gradually decreases, which helps the algorithm converge to the optimal solution in the later stage. The improved position update formula in the whale fall stage is as follows:

[0149]

[0150] In the formula, rand is a random number between 0 and 1; CL is the probability factor, and the expression is as follows:

[0151]

[0152] S105, based on the improved Beluga optimization algorithm, the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine are optimized, based on the training samples and the prediction samples, the port load time series and meteorological dimension influencing factor data are input, and the port load is predicted based on the support vector machine to obtain the prediction result.

[0153] S105.1: Input the output y of the processed data model as the load energy consumption value to be predicted, and input the vector port load time series and meteorological dimension influencing factor data;

[0154]

[0155] Among them, matrix A is the port load energy consumption data at the prediction time of the day before the prediction day, and each row represents the selected different important load energy consumption load types; matrix B is the port load energy consumption data at the moment before the prediction time of the day before the prediction day, and each row represents different load energy consumption load types; vector C is meteorological data, including maximum temperature, minimum temperature, average temperature, etc.; in vector D, d1 represents working days, and d2 represents rest days;

[0156] S105.2: Set the parameters of the improved Beluga optimization algorithm, including the number of individuals in the population, the dimension, and the maximum number of iterations;

[0157] S105.3: The optimal penalty factor C, insensitive loss function ε, and radial basis function parameter σ of the support vector machine obtained based on the White Whale optimization algorithm;

[0158] S105.4: Establish a regression objective function based on the training samples and minimize the regression objective function; the regression objective function is:

[0159]

[0160] In the formula, α i and α i * (i=1,2,…,l) is the Lagrange multiplier; K(x i ,x j )=φ(x i )·φ(x j ) is the kernel function; minimize the regression objective function and get the optimal solution as

[0161] S105.5: Substitute the optimal solution into equation (22) to obtain the decision regression equation;

[0162]

[0163] In the formula, The calculation formula is:

[0164]

[0165] S105.6: Predict the port load at future times based on the forecast samples and the obtained decision regression equation.

[0166] See also Figure 2 The present invention discloses a port load forecasting system taking into account multi-dimensional influencing factors, including:

[0167] A selection module, wherein the selection module selects multi-dimensional influencing factors affecting port load, including port load category dimension influencing factors, and selects port load time series and meteorological dimension influencing factors based on principal component-grey correlation;

[0168] A processing module, which processes the bad port data and normalizes the obtained new data;

[0169] A construction module, wherein the construction module divides the normalized data into training samples and prediction samples, and constructs a fitness function so that the total error of the training sample set is minimized;

[0170] An improvement module, wherein the improvement module improves the performance of the White Whale optimization algorithm to obtain an improved White Whale optimization algorithm;

[0171] The prediction module optimizes the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine based on the improved Beluga optimization algorithm, inputs the port load time series and meteorological dimension influencing factor data based on the training samples and the prediction samples, predicts the port load based on the support vector machine, and obtains the prediction result.

[0172] Example:

[0173] Step 1: Select the important influencing factors of port load in multiple dimensions, including the selection of influencing factors of port load category dimension and the extraction of influencing factors of port load time series and meteorological dimension based on principal component-grey correlation;

[0174] Step 1.1: Statistical analysis of various loads of a port in Jiangsu from March to December 2023. Figure 3 and Figure 4 As shown in the figure, the energy consumption of yard operation, horizontal transportation, shore loading and unloading, and ship energy consumption, which have the largest energy consumption proportion, the most obvious time series change, and the closest correlation with port throughput, are selected as the influencing factors of port load prediction;

[0175] Step 1.2, after obtaining the energy consumption of yard operations, horizontal transportation, shore loading and unloading, and ship energy consumption as the influencing factors of port load prediction, filter the load historical data and construct an indicator system based on the load historical energy consumption value and meteorological factors;

[0176] The load historical energy consumption values ​​and meteorological factors of a port in Jiangsu in June 2023 are selected to form an index system, including: the load energy consumption values ​​at the same time of the day before and the previous one, two and three hours, the load energy consumption values ​​at the same time of the two days before and the previous one, two and three hours, and the load energy consumption values ​​at the same time of the same day two weeks before and the previous one, two and three hours; the relevant values ​​of the highest temperature, lowest temperature, average temperature, rainfall and air pressure on the day, the previous three days and the same day two weeks before;

[0177] Step 1.3, design the principal component-grey correlation analysis process as follows Figure 5 As shown;

[0178] Step 1.4, apply the principal component-grey correlation analysis method to reduce the dimension of the above index system, and get the ten factors with the greatest impact on the predicted daily load: the load energy consumption value at the same time of the day before the forecast, the load energy consumption value at the moment before the day before the forecast, the load energy consumption value at the same time two days before the forecast, the load energy consumption value at the moment two days before the forecast, the load energy consumption value at the same time of the day one week before the forecast, the load energy consumption value at the same time of the day two weeks before the forecast, the maximum temperature, the minimum temperature, the average temperature and the date type of the forecast day;

[0179] Step 2, using wavelet denoising to pre-process the bad port data and normalize the new data obtained;

[0180] Step 2.1, using the weighted average method to estimate the wavelet coefficients according to formula (1) and formula (2) to improve the denoising effect;

[0181] Step 2.2, denoising the original port load data by using an improved wavelet denoising method to repair the defective data;

[0182] Step 2.3, normalize the preprocessed new data according to formula (3);

[0183] The present invention selects two weeks of port load energy consumption data of a port in Jiangsu for model training and testing. Figure 6 As shown, the actual load energy consumption data of the port is counted every 30 minutes;

[0184] Step 3, establish training samples and prediction samples, use the input data of the first two weeks as training samples, use the load energy consumption data of the last day as prediction samples, and use the formed training samples to establish the fitness function according to formula (4);

[0185] Step 4, improve the performance of the White Whale optimization algorithm to obtain an improved White Whale optimization algorithm. Improved White Whale optimization algorithm;

[0186] Step 4.1, improve the population position initialization strategy according to formula (7);

[0187] Step 4.2, introduce dynamic weight, its expression is (12), and the position update formula of the improved Beluga optimization algorithm development stage is shown in formula (13);

[0188] Step 4.3, introduce the optimal de-disturbance strategy, and equations (14)-(20) are the update formulas of the whale fall of the improved white whale optimization algorithm;

[0189] Step 5, input data, use the improved Beluga Optimization (IBWO) algorithm to optimize the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine, and use the support vector machine to predict the port load to obtain the prediction result;

[0190] Step 5.1, the output y of the support vector machine is the load energy consumption value to be predicted, and the input vector includes the following items:

[0191]

[0192] Among them, matrix A is the port load energy consumption data at the prediction time of the day before the prediction, and each row represents the selected different important load energy consumption load types; matrix B is the port load energy consumption data at the moment before the prediction time of the day before the prediction, and each row represents different load energy consumption load types; vector C is meteorological data, including maximum temperature, minimum temperature, average temperature, etc.; in vector D, d1 represents working days, and d2 represents rest days;

[0193] Step 5.2, set the number of beluga whales in the beluga whale population in IBWO to N = 50, the position update related parameter B = 1.5, the maximum number of iterations to = 100, the boundary constant h of the algorithm to 0.5, and the value range determined based on experience;

[0194] Step 5.3, apply IBWO iteration to find the optimal penalty factor C, insensitive loss function ε and optimal value of radial basis function parameter σ that affect the performance of support vector machine;

[0195] Step 5.4: Based on the training samples, the SVM regression objective function is established according to formula (21), and the regression objective function is minimized to obtain the optimal solution:

[0196] Step 5.4, substitute the optimal solution into equation (22) to obtain the SVM decision regression equation;

[0197] Step 5.5, use the forecast sample and the obtained decision regression equation to predict the port load at future times.

[0198] In order to better demonstrate the proposed port short-term load forecasting model (IBWOSVM) based on the improved Beluga optimization algorithm, the prediction results of IBWOSVM are compared with those of BP neural network, extreme learning machine (ELM) and the support vector machine port short-term load forecasting model (BWOSVM) based on the Beluga optimization algorithm. The relative error (RE) and root mean square error (RMSE) are used to evaluate the fitting error of the model. The coefficient of determination (r2) is used to evaluate the goodness of fit. The prediction results of each model are shown in Figure 2. Figure 7 and Figure 8 shown. Figure 6 The predicted load curves for the four models are: Figure 8 The relative errors of the four models are shown in Table 1. The prediction error distribution intervals of the ELM model, BP model, BWOSVM model and IBWOSVM model are [0.02%, 7.03%], [0.05%, 6.88%], [0.03%, 7.68%] and [0.15%, 6.61%] respectively. The error interval obtained by the IBWOSVM model is the smallest, indicating that the IBWOSVM model has higher prediction stability and accuracy. For the one-day load forecasting results, the RMSE and r2 values ​​of the four prediction models are shown in Table 1.

[0199] Table 1 Evaluation and analysis of prediction results

[0200]

[0201] As shown in Table 1, the coefficient of determination r2 of the four models is about 90%, and the highest r2 value of the IBWOSVM model is 90.66%, indicating that the IBWOSVM model has a good fit. For the RMSE evaluation index, the minimum RMSE value of the IBWOSVM model is 16.16, and the maximum RMSE value of the SVM model is 17.92, indicating that the prediction error of the IBWOSVM model is small. The above analysis proves the effectiveness of the IBWOSVM prediction model.

[0202] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A port load forecasting method taking into account multi-dimensional influencing factors, characterized in that: include: Select multi-dimensional influencing factors that affect port load, including port load category dimension influencing factors, and select port load time series and meteorological dimension influencing factors based on principal component-grey correlation; Process bad port data and normalize the new data obtained; The normalized data is divided into training samples and prediction samples, and the fitness function is constructed to minimize the total error of the training sample set; Improve the performance of the White Whale optimization algorithm and obtain the improved White Whale optimization algorithm; Based on the improved Beluga optimization algorithm, the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine are optimized. Based on the training samples and prediction samples, the port load time series and meteorological dimension influencing factor data are input, and the port load is predicted based on the support vector machine to obtain the prediction result.

2. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 1 is characterized in that: The multi-dimensional influencing factors affecting the port load are selected, including the influencing factors of the port load category dimension, and the influencing factors of the port load time series and meteorological dimension based on the principal component-grey correlation degree are selected, specifically: Statistics are collected on various types of port loads, and the loads with the largest energy consumption, the most obvious time series changes, and the closest correlation with port throughput are selected as the influencing factors for port load prediction; Construct an index system based on historical load energy consumption values ​​and meteorological factors; Based on principal component analysis and grey correlation method, the principal component-grey correlation analysis method is constructed; Based on the principal component-grey correlation analysis method, the indicator system is reduced in dimension to obtain the important influencing factors of short-term load.

3. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 2 is characterized in that: The bad port data is processed and the obtained new data is normalized, specifically: Step 3.1: Improve the wavelet denoising threshold function to obtain an improved wavelet denoising method; In view of the shortcomings of the soft and hard threshold methods in the wavelet denoising threshold function, the weighted average method is used to estimate the wavelet coefficients. The improved formula is: In the formula, w j,k and are the wavelet coefficients of the noisy signal after wavelet transformation and the estimated wavelet coefficients; T is the threshold; β is the weighting factor; sgn() is the sign function; j is the number of decomposition levels; Step 3.2: De-noise the original port load data based on the improved wavelet denoising method and repair the defective data; Step 3.3: Normalize the repaired data. The normalization formula is: In the formula, is the normalized data value; xi is the original value; x imin is the minimum value in the column; x imax is the maximum value of the column; l is the dimension of the input vector.

4. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 3 is characterized in that: The fitness function is: In the formula, y i is the actual load value; is the predicted load value; n represents the number of predicted loads.

5. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 4 is characterized in that: The performance of the White Whale optimization algorithm is improved to obtain an improved White Whale optimization algorithm, which is specifically: Improve the beluga whale population position initialization strategy to evenly distribute the beluga whale positions; Introducing dynamic weights in the process of updating the beluga whale's position to optimize the optimal solution for the beluga whale's position; Based on the optimal de-disturbance strategy, the position of the beluga whale in the whale fall phase is updated during the beluga whale's hunting process.

6. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 5 is characterized in that: The strategy for initializing the positions of the beluga whale population is improved to evenly distribute the positions of the beluga whales, specifically: The position matrix of the white whale in the white whale optimization algorithm is as follows: Where n represents the number of beluga whales in the population; x i is the white whale position; D is the dimension of the problem to be solved; For all beluga whales, the corresponding fitness values ​​are stored as follows: Where, fix is ​​the fitness matrix of the beluga population; fix n Corresponding to the fitness value of the nth individual in the beluga population; The initialization strategy is introduced to enhance the exploration performance of the White Whale optimization algorithm. The function expression of the improved initialization strategy is: Wherein, h is the adjustment parameter of uniform distribution and h=0.

5.

7. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 6 is characterized in that: In the process of updating the beluga position, dynamic weights are introduced to optimize the optimal solution of the beluga position. Specifically, the beluga shares each other's position with nearby beluga whales, and determines that the beluga whale with the best distance to the prey is the best candidate, and the other beluga whales are other candidates. The beluga position is updated in real time to continuously determine the best candidate; wherein, the beluga position update formula is as follows: μ=2r2((1-T) / T max ) (9) L F =0.05×(δ×γ) / |μ 1 / B (10) In the formula, and are the position of the i-th beluga whale and the position of a random beluga whale at the current iteration number, respectively; r1 and r2 are random numbers between (0,1); The position of the best white whale is the global optimal position; λ represents the random jump intensity used to measure the flight intensity in the Levy flight strategy; T max is the maximum number of iterations; L F is the Levy flight function; δ and μ are normally distributed random numbers; B is a constant whose value is 1.5; Dynamic weights are introduced into the White Whale position update formula. A larger weight in the early development stage is beneficial to global search, while a lower weight in the later stage is more conducive to White Whale searching for the optimal solution within a certain range. The weight formula is as follows: After the dynamic weight w is introduced, the position update formula of the white whale is as follows:

8. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 7 is characterized in that: Based on the optimal de-disturbance strategy, the position of the beluga whale in the whale fall phase is updated during the beluga whale hunting process, specifically: Beluga whales may experience whale falls when they are moving or hunting. The mathematical model of whale falls is as follows: Where r3, r4 and r5 are random numbers between (0,1), X s is the step length of the whale fall; u b and l b are the upper and lower bounds of the variable respectively; ρ f is the probability of a whale fall; Introducing small-amplitude neighborhood disturbances can improve the ability of the White Whale optimization algorithm to escape from local optimality. The formula for small-amplitude neighborhood disturbances is as follows: Among them, c1 is a variable disturbance factor. As the number of iterations increases, the disturbance factor gradually decreases, which helps the algorithm converge to the optimal solution in the later stage. The improved position update formula in the whale fall stage is as follows: In the formula, rand is a random number between 0 and 1; CL is the probability factor, and the expression is as follows:

9. The port load forecasting method taking into account multi-dimensional influencing factors according to claim 8, characterized in that: The penalty factor, insensitive loss function and radial basis function parameters in the support vector machine are optimized based on the improved Beluga optimization algorithm. Based on the training samples and the prediction samples, the port load time series and the meteorological dimension influencing factor data are input, and the port load is predicted based on the support vector machine to obtain the prediction results, which are specifically as follows: Step 9.1: Input the output y of the processed data model as the load energy consumption value to be predicted, and input the vector port load time series and meteorological dimension influencing factor data; Among them, matrix A is the port load energy consumption data at the prediction time of the day before the prediction day, and each row represents the selected different important load energy consumption load types; matrix B is the port load energy consumption data at the moment before the prediction time of the day before the prediction day, and each row represents different load energy consumption load types; Vector C is the meteorological data, including the highest temperature, the lowest temperature, the average temperature, etc. In vector D, d1 represents the working day, and d2 represents the rest day; Step 9.2: Set the parameters of the improved Beluga optimization algorithm, including the number of individuals in the population, the dimension, and the maximum number of iterations; Step 9.3: The optimal penalty factor C, insensitive loss function ε and radial basis function parameter σ of the support vector machine are obtained based on the White Whale optimization algorithm; Step 9.4: Establish a regression objective function based on the training samples and minimize the regression objective function; the regression objective function is: In the formula, α i and α i * (i=1,2,…,l) is the Lagrange multiplier; K(x i ,x j )=φ(x i )·φ(x j ) is the kernel function; minimize the regression objective function and get the optimal solution as Step 9.5: Substitute the optimal solution into equation (22) to obtain the decision regression equation; In the formula, The calculation formula is: Step 9.6: Predict the port load at future times based on the forecast samples and the obtained decision regression equation.

10. A port load forecasting system taking into account multi-dimensional influencing factors, characterized in that: include: A selection module, wherein the selection module selects multi-dimensional influencing factors affecting port load, including port load category dimension influencing factors, and selects port load time series and meteorological dimension influencing factors based on principal component-grey correlation; A processing module, which processes the bad port data and normalizes the obtained new data; A construction module, wherein the construction module divides the normalized data into training samples and prediction samples, and constructs a fitness function so that the total error of the training sample set is minimized; An improvement module, wherein the improvement module improves the performance of the White Whale optimization algorithm to obtain an improved White Whale optimization algorithm; The prediction module optimizes the penalty factor, insensitive loss function and radial basis function parameters in the support vector machine based on the improved Beluga optimization algorithm, inputs the port load time series and meteorological dimension influencing factor data based on the training samples and the prediction samples, predicts the port load based on the support vector machine, and obtains the prediction result.

Images