CN111754037A

CN111754037A - Long-term load hybrid prediction method for regional terminal integrated energy supply system

Info

Publication number: CN111754037A
Application number: CN202010571419.8A
Authority: CN
Inventors: 耿翠英; 黄泽华; 郭建宇; 娄北; 王文豪; 林烽; 刘洋; 张龙
Original assignee: State Grid Corp of China SGCC; Economic and Technological Research Institute of State Grid Henan Electric Power Co Ltd
Current assignee: State Grid Corp of China SGCC; Economic and Technological Research Institute of State Grid Henan Electric Power Co Ltd
Priority date: 2020-06-19
Filing date: 2020-06-19
Publication date: 2020-10-09
Anticipated expiration: 2040-06-19
Also published as: CN111754037B

Abstract

The invention belongs to a long-term load forecasting method, and particularly relates to a long-term load hybrid forecasting method for a regional terminal integrated energy supply system, which comprises the following steps: collecting original historical data of each load in the region, and performing normalization processing; calculating the collected coupling degree C of each load; establishing an ARIMA prediction model; determining internal and external variables of a prediction model of each load system; building a VAR prediction model; and obtaining a final prediction result. The method can more accurately perform mixed prediction on the conditions of each long-term load of the regional terminal integrated energy supply system, and can reasonably arrange the construction of energy supply facilities.

Description

Long-term load hybrid prediction method for regional terminal integrated energy supply system

Technical Field

The invention belongs to a long-term load forecasting method, and particularly relates to a long-term load hybrid forecasting method for a regional terminal integrated energy supply system.

Background

With the development of advanced science, the demand of people on energy is increasing day by day, meanwhile, the problems of energy safety, environmental protection and the like due to the fact that the traditional fossil energy is exhausted day by day are concerned, and the situations of low energy conversion efficiency, non-centralized distribution, high use cost and the like of the existing forms of cold, heat, electricity and the like generally exist, so that the energy and the environment become the main bottleneck restricting the sustainable development of national economy. The physical characteristic complementarity of an electric power system, a thermodynamic system and a gas system is strong, and a regional terminal integrated energy supply system is a novel energy system integrating the supply of electric power, natural gas, heat energy and cold energy, and has an important promoting effect on optimizing an energy structure, improving the energy use efficiency and promoting the consumption of renewable energy. The load of the regional multi-energy supply system is influenced by various factors and has the characteristics of uncertainty and nonlinearity. The cold and hot electric loads of the system are not only related to the historical data of the system, but also influence each other.

For regional energy supply, the energy supply range is usually large, the result of load prediction in a region is influenced by various factors, and the influence degree can be represented more accurately only by reasonably quantizing various influence factors. The optimization planning method and the expansion planning method of the integrated energy supply system of the research area terminal are beneficial to realizing the cooperative and efficient supply of cold and hot electricity in a new energy increasing area and the comprehensive and graded utilization of energy in the existing industrial park, and have important practical significance for the comprehensive supply mode of the energy in the power grid innovation area, the improvement of the supply side and the realization of the industrial upgrading.

With regard to load prediction of an electric power system, foreign scholars often perform load prediction by using a method related to a neural network and a support vector machine. The scholars propose to select similar daily loads as input loads, then apply wavelet decomposition to decompose the loads into low-frequency components and high-frequency components, and finally use a single neural network to predict the future loads of the two components. And a learner also adopts the asymmetric quadratic loss function to support vector regression to accurately predict the load, so that the accuracy of the load prediction model is effectively improved. A new hybrid intelligent algorithm based on Wavelet Transform (WT) and fuzzy adaptive resonance theory mapping network is proposed, and the model is proved by extensive prediction comparison. Experts divide a time sequence into two parts by using an empirical mode decomposition method, respectively describe the trend and the energy consumption value of local oscillation, and then the two parts are used for training a support vector regression model. Experts also propose a short-term load prediction method based on a kernel machine, which provides a better short-term load prediction result.

Regarding heating system load prediction, Kawashima et al abroad used the ANN model to predict the heat demand of a single family building in Sweden. Kalogirou utilizes an ANN model to predict the long-term characteristics of a Greek solar heating system. Chou and Bui use ANN to predict building thermal load. Yang et al use neural networks to predict building thermal loads and determine optimal startup times for heating systems. Kalogirou et al utilize a BP neural network for predicting building thermal loads and are trained with 225 building data. Ekici and Aksoy establish a BP neural network with input variables of building orientation, building insulation thickness and transmittance parameters to predict annual heat load requirements of swedish independent homes. Olofsson and Andersson establish a neural network model for long-term thermal load prediction using short-term (2-5 weeks) data. Kreider et al predict building thermal loads using a BP neural network with input variables for time-by-time building energy consumption data. And the Yan and the Yao carry out prediction analysis on the building heat load data in different climatic regions by adopting a neural network model.

With respect to load prediction of an electric power system, domestic scholars have also developed various methods and algorithms in order to improve the accuracy and speed of load prediction. The learners take the voltage characteristics as basic quantities for describing the system state characteristics, and propose a state estimation model and an algorithm based on weighted least squares. And experts also realize attribute reduction of a rough set theory by adopting the global search capability of a genetic algorithm, and improve and perfect the load prediction model and algorithm of the least square support vector machine by optimizing the input variables of the model and automatically optimizing the parameters of the model by adopting a real-value genetic algorithm. The students also introduce human comfort indexes, comprehensively consider the influence of meteorological factors, establish a particle swarm parameter Optimization-based Support Vector Machine (PSO-SVM) prediction model by utilizing daily feature vectors and load data of similar days, and experiments prove that the prediction precision is high and the popularization capability is strong. The problems of data preprocessing, kernel function construction and selection, parameter optimization and the like in the application of the SVM in short-term load prediction are analyzed, the existing solution is summarized, and the key problem to be solved next step is provided. The students also study the power load prediction problem of local areas specially, and provide a new multi-mode variable structure load prediction method based on self-adaptive clustering partition and support vector regression, and experiments prove that the prediction method has higher precision and stronger robustness than the traditional neural network prediction method. The professor David in the Doe of cattle utilizes a data mining technology which has advantages in the aspects of processing large data volume, eliminating redundant information and the like to preprocess historical data to form a data sequence with highly similar meteorological characteristics, and the sequence is used as training data of an SVM (support vector machine), so that the data volume is reduced, the prediction speed and precision are improved, and the defects of a support vector machine are overcome.

As for the load prediction of a heating system, Chinese researchers carry out theoretical and time research work on the application of an artificial neural network to the heat load prediction. The Du-jin et al explains how to determine the input variables and the output variables of the heat supply load prediction neural network, the prediction time step length, the number of hidden layer units and the like. Congying carries out comparative study on the effect of stepwise regression analysis and an artificial neural network on building heat load prediction.

In summary, the conventional load prediction methods are all directed to a single energy system, and the coupling relationship between multiple energy sources is not considered, so that it is necessary to perform uniform load prediction on a regional integrated energy supply system with multiple energy sources.

Disclosure of Invention

The invention aims to provide a long-term load hybrid prediction method for a regional terminal integrated energy supply system, aiming at the problems in the prior art, which can more accurately perform hybrid prediction on the conditions of each long-term load of the regional terminal integrated energy supply system and reasonably arrange the construction of energy supply facilities.

The technical scheme of the invention is as follows:

the long-term load hybrid prediction method for the area terminal integrated energy supply system comprises the following steps:

s1, collecting original historical data of each load in a region, and performing normalization processing;

s2, calculating the collected coupling degree C of each load;

s3, establishing an ARIMA prediction model;

s4, determining internal and external variables of a prediction model of each load system;

s5, building a VAR prediction model;

and S6, obtaining a final prediction result.

Specifically, the normalization processing step in step S1 is as follows:

a) the normalization of each data in the analysis range of the electric, gas, cold and heat loads is calculated according to the following formula

Chemical value

Wherein: l is_emax、L_eminThe maximum value and the minimum value in the electric load analysis range are obtained;

L_gmax、L_gminis qiMaximum and minimum values within the load analysis range;

L_cmax、L_cminmaximum and minimum values in the analysis range of the cold load;

L_hmax、L_hminmaximum and minimum values in the thermal load analysis range;

l^* _e(i)、l_e(i) the method comprises the following steps The electricity load normalization value and the actual value of the ith day in the load analysis range;

l^* _g(i)、l_g(i) the method comprises the following steps The ith weather load normalization value and the actual value in the load analysis range;

l^* _c(i)、l_c(i) the method comprises the following steps The cold load normalization value and the actual value on the ith day in the load analysis range;

l^* _h(i)、l_h(i) the method comprises the following steps The normalized value and the actual value of the thermal load of the ith day in the load analysis range;

b) the electric, gas and cold and heat loads account for the specific weight of the system load in the ith day:

of which ∑ l^*(i) Total load for system day i:

c) day i load entropy value e (i):

e(i)＝-(ln 4)-1[r_e(i)ln r_e(i)+r_g(i)ln r_g(i)+r_c(i)ln r_c(i)+r_h(i)ln r_h(i)]；

d) day i differential coefficient of system load g (i): (i) 1-e (i);

e) day i system load weight ω (i):

f) electricity, gas, cold and heat load comprehensive change index:

wherein gamma is_e、γ_g、γ_c、γ_hRespectively are the comprehensive change indexes of the system electricity, gas, cold and heat loads.

Specifically, the load coupling degree calculation formula is as follows:

wherein C is_e,g、C_g,c、C_e,h、C_g,h、C_g,c、C_c,hThe coupling values of electric-gas, electric-cold, electric-heat and gas-cold, gas-heat loads in the analysis range are respectively, and the value range is [0, 1%]。

Specifically, the establishing step of the ARIMA prediction model in step S3 is as follows:

(1) difference processing: d-order difference processing is carried out on the original time sequence [ Xt ] to obtain a stable time sequence [ Xt' ];

(2) model identification and parameter scaling: calculating the autocorrelation and partial autocorrelation functions of a stationary time sequence [ Xt' ], preliminarily determining model categories (AR, MA and ARMA), and determining the values of model parameters p and q by using the minimum information criterion (AIC) information criterion;

(3) parameter estimation: using the correlation moments to carry out parameter estimation on ai and bj, and determining a final ARIMA (p, q, d) model;

(4) and (3) data prediction: single-step or multi-step prediction is realized through the established ARIMA model.

Specifically, the expression of the VAR prediction model in step S5 is as follows:

y_t＝A₁y_t-1+…+A_py_t-p+Bx_t+_tt＝1，2，…，N

wherein y is_tIs a k-dimensional internal variable vector, N is the number of samples, k × k-dimensional matrixes A1 and …, Ap is an internal variable coefficient matrix, B is an external variable coefficient matrix,_tis a k-dimensional perturbation vector。

Specifically, the building of the VAR prediction model in step S5 includes the following steps:

1) determining model variables: determining endogenous variables and exogenous variables of the model through characteristic analysis of the relevant variables;

2) estimating model parameters: determining stationary time series vector [ y ]_1t，y_2t，y_3t，y_Tt，x_t]Calculating an endogenous variable coefficient matrix A and an exogenous variable coefficient matrix B by utilizing maximum likelihood estimation;

3) determining the order of the model: determining the order p of the model by using Akaike's Information Criterion (AIC);

4) and (3) data prediction: single-step or multi-step prediction is realized through the established VAR (p) model.

The invention has the beneficial effects that: the prediction method provided by the invention fully considers the relevance among electric, heat, cold and gas loads in an interval, and performs quantitative analysis by using the coupling relation, and because the dimensions of various influence factors are different, in order to prevent the real action of partial influence factors from being distorted or even annihilated in the whole mapping effect due to the difference of value range, the invention uniformly performs standardized treatment on various influence factors, so that the threshold value fluctuates in the range of 0-1. The method provided by the invention has stronger tracking capability on each load, has smaller prediction error and has obvious advantages.

Drawings

FIG. 1 is a schematic technical route of example 1;

FIG. 2 is a graph of the results of an electrical load coupling prediction;

FIG. 3 is a graph of the results of prediction of electric cooling load coupling;

FIG. 4 is a graph of results of prediction of degree of coupling of electrical heating load;

FIG. 5 is a graph of air cooling load coupling prediction results;

FIG. 6 is a graph of the results of a prediction of gas heat load coupling;

FIG. 7 is a graph of the results of a prediction of cold and heat load coupling;

FIG. 8 is a graph of the prediction results of the electrical load independent prediction model;

FIG. 9 is a graph of electrical load hybrid predictive model prediction results;

FIG. 10 is a graph of the air load independent prediction model prediction results;

FIG. 11 is a graph of the prediction results of the air load hybrid prediction model;

FIG. 12 is a graph of cold load independent prediction model prediction results;

FIG. 13 is a graph of cold load hybrid predictive model prediction results;

FIG. 14 is a graph of thermal load independent prediction model prediction results;

FIG. 15 is a graph of thermal load hybrid predictive model prediction results.

Detailed Description

The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

The invention provides a long-term load hybrid prediction method for a regional terminal integrated energy supply system, which comprises the following steps:

s2, calculating the collected coupling degree C of each load;

s3, establishing an ARIMA prediction model;

s5, building a VAR prediction model;

and S6, obtaining a final prediction result.

In the integrated energy supply system of the area terminal, the result of load prediction is influenced by various factors, and only by reasonably quantizing various influencing factors, the influence degree can be more accurately represented, and the influence degree is introduced into the modeling of a prediction model. Due to different dimensions of various influence factors, in order to prevent the true effect of partial influence factors from being distorted or even annihilated in the whole mapping effect due to the difference of value range, the various influence factors are uniformly standardized, so that the threshold value fluctuates within the range of 0-1. The time-series normalization processing steps for the respective influencing factors in step S1 are as follows:

Chemical value

L_gmax、L_gminthe maximum value and the minimum value in the gas load analysis range are obtained;

L_hmax、L_hminmaximum and minimum values in the thermal load analysis range;

of which ∑ l^*(i) Total load for system day i:

c) day i load entropy value e (i):

d) day i differential coefficient of system load g (i): (i) 1-e (i);

e) day i system load weight ω (i):

f) electricity, gas, cold and heat load comprehensive change index:

The degree of coupling is the degree of interaction of the description systems, and the degree of influence between the loads of the integrated energy supply system of the electric-gas interconnection area terminal is defined as the degree of load coupling, and the value reflects the degree of interaction between the system loads. With reference to a capacity coupling concept (capacity coupling) and a capacity coupling coefficient model in physics, a coupling calculation model of a load of a regional terminal integrated energy supply system can be defined, and the load coupling calculation formula is as follows:

a) When the coupling degree function value C tends to 1, the coupling degree is the maximum, which shows that the coupling relationship between every two four loads of the area terminal integrated energy supply system is strong;

b) when the coupling degree function value C is 0, the coupling degree is extremely low, which indicates that no mutual influence relationship exists between the quadruple loads of the area terminal integrated energy supply system;

c) when C is more than 0 and less than or equal to 0.5, the loads of the area terminal integrated energy supply system are in a coupling state of a lower level, the interaction among the loads is not strong, and the coupling relation of the loads can not be considered in the process of load prediction;

d) when C is more than 0.5 and less than or equal to 1, the coupling degree between the loads is high, the interaction relation between the comprehensive energy loads is strong, and the interaction between the loads is fully considered in establishing a model for predicting the comprehensive energy loads so as to improve the prediction accuracy.

The establishing step of the ARIMA prediction model in the step S3 is as follows:

The independent prediction model of the electric, gas, cold and heat loads of the area terminal integrated energy supply system is as follows:

electric load independent prediction model:

air load independent prediction model:

cold load independent prediction model:

heat load independent prediction model:

in step S5, VAR is a vector auto regression (vector auto regression) model established based on statistical properties of data, and the basic principle is that a time sequence is regarded as a random process, and a mathematical model is established to describe or simulate the time sequence, so that dynamic characteristics and persistence of linear components of the time sequence can be well reflected, and the correlation between the past and future time of the time sequence and the present time is revealed. Generally denoted as var (p), where p is the model order, the general expression for var (p) is as follows:

y_t＝A₁y_t-1+…+A_py_t-p+Bx_t+_tt＝1，2，…，N

wherein yt is a k-dimensional internal variable vector, N is the number of samples, k × k-dimensional matrices A1, …, Ap is an internal variable coefficient matrix, B is an external variable coefficient matrix,_tis a k-dimensional perturbation vector.

The building of the VAR prediction model in step S5 includes the following steps:

Example 1

In the embodiment, the VAR is used as a basic model for system load hybrid prediction, the endogenous variable and the exogenous variable of the VAR (p) are determined by combining the load coupling degree of a week before the load with the predicted daily load, the modeling is performed by utilizing each load data of the energy supply system integrated by the area terminal from 12 and 1 days in 2017 to 12 and 11 days in 2018, and each load condition of the system from 12 and 12 days in 2018 to 1 and 10 days in 2019 is predicted. Fig. 1 is a circuit diagram of a load prediction technology of an area terminal integrated energy supply system.

As shown in fig. 2-8, the curves of the load coupling degree prediction results of a week before the daily load of the area terminal integrated energy supply system from 12 days in 2018 and 12 months to 10 days in 2019 and 1 month, the prediction result values are shown in table 1 below, and it can be seen from the graphs that the prediction accuracy is high. It can be seen from the table that the air-cooling load coupling degree is less than 0.5 from day 22 to day 29, the cooling-heating load coupling degree is less than 0.5 from day 21 to day 29, the load coupling degrees of the other cases are all more than 0.5, the coupling conditions of the loads are comprehensively considered to determine the internal and external variables of the VAR prediction model of the loads, the load hybrid prediction model of the regional terminal integrated energy supply system is established, and the internal and external variable determination table of the load prediction VAR model of the regional terminal integrated energy supply system is shown in the following table 2.

TABLE 1

	C_e，g	C_e，c	C_e，h	C_g，c	C_g，h	C _c，h
							1	0.963437	1.004296	0.878808	0.962141	0.967959	0.87458
2	0.950128	1.000862	0.852101	0.938362	0.966143	0.834486
							3	0.908921	0.998254	0.795197	0.910508	0.9704	0.8018
4	0.880099	0.997168	0.775733	0.863026	0.976234	0.755012
							5	0.867569	0.998744	0.753112	0.880378	0.970673	0.773016
6	0.82645	0.995223	0.700012	0.872622	0.967626	0.753991
							7	0.833792	0.993932	0.713403	0.874939	0.968008	0.757079
8	0.830698	0.993916	0.704748	0.87048	0.968509	0.752051
							9	0.828322	0.994297	0.705659	0.871643	0.968375	0.754629
10	0.843782	0.992172	0.714766	0.899022	0.963679	0.779896
							11	0.869187	0.987495	0.732348	0.935502	0.956345	0.815099
12	0.885563	0.991394	0.754873	0.928974	0.9602	0.810583
							13	0.929578	0.998001	0.817621	0.946011	0.963306	0.842516
14	0.8915	0.98999	0.741661	0.949791	0.95231	0.830787
							15	0.877639	0.983451	0.732355	0.938232	0.950002	0.807081
16	0.882935	0.988303	0.735795	0.925889	0.952946	0.793041
							17	0.846487	0.997061	0.706094	0.8698	0.963646	0.73348
18	0.801971	1.002996	0.679536	0.783301	0.973987	0.659776
							19	0.767692	0.999268	0.653648	0.717908	0.976048	0.605837
20	0.719618	0.989873	0.612723	0.636617	0.978554	0.533974
							21	0.719223	0.93723	0.634587	0.510694	0.986218	0.43201
22	0.672029	0.881615	0.576674	0.42949	0.983424	0.357404
							23	0.652076	0.814813	0.578747	0.366169	0.988111	0.311786
24	0.670301	0.753233	0.597514	0.344596	0.989021	0.296285
							25	0.680893	0.743666	0.608673	0.353476	0.988937	0.304211
26	0.690526	0.757554	0.613232	0.364945	0.986657	0.310955
							27	0.725195	0.713525	0.64812	0.343704	0.986634	0.289613
28	0.762506	0.734068	0.681925	0.401153	0.985945	0.343279
							29	0.816763	0.783949	0.748862	0.495596	0.989198	0.436165
30	0.870904	0.850816	0.792387	0.595839	0.983663	0.516448

TABLE 2

Are respectively provided withThe time sequence of the electric load, the gas load, the cold load and the heat load of the area terminal integrated energy supply system is recorded as follows: { l_g(t)},{l_c(t)},{l_h(t), wherein t is 1-N, and N is 406 is the number of sample spaces, namely the total days from 12 months 1 days in 2017 to 1 month 10 days in 2018, modeling is carried out by using the front 376 groups of the samples, and the load condition of the last 30 days (one month) is predicted.

1) Electric load hybrid prediction model

The electrical load hybrid prediction VAR model comprises the following internal variables of an electrical load and a cold load, the air load and the heat load which are relatively weak in correlation with the electrical load are external variables, and the model order p is determined to be 3 by using an AIC (air interface) criterion:

wherein,

L_e＝[l_e(t-1)，l_e(t-2)，l_e(t-3)]^T，L_c＝[l_c(t-1)，l_c(t-2)，l_c(t-3)]^T。

2) gas load hybrid prediction model

On days 1 to 21 and 30 of the air load hybrid prediction VAR model, the endogenous variables are air load and heat load, the exogenous variables are electric load and cold load, and the model order p is determined to be 2 by using an AIC (air interface) criterion; from day 22 to day 29, the endogenous variables are gas load and heat load, the exogenous variables are electric load, and the model order is p-3:

wherein: l is_g＝[l_g(t-1)，l_g(t-2)]^T，L_h＝[l_h(t-1)，l_h(t-2)]^T，

L′_g＝[l_g(t-1)，l_g(t-2)，l_g(t-3)]^T，L′_h＝[l_h(t-1)，l_h(t-2)，l_h(t-3)]^T。

3) Cold load hybrid prediction model

The cold load hybrid prediction VAR model comprises the following steps that (1) day 1 to 21 day 30 internal variables are cold load and electric load, external variables are gas load and heat load, and the model order p is determined to be 3 by using an AIC (air interface) criterion; from day 22 to day 29, the endogenous variables are the cooling load and the electrical load, and the model order is p-2:

wherein L is_e＝[l_e(t-1)，l_e(t-2)，l_e(t-3)]^T，

L_c＝[l_c(t-1)，l_c(t-2)，l_c(t-3)]^T，

L′_e＝[l′_e(t-1)，l′_e(t-2)]^T，L′_c＝[l′_c(t-1)，l′_c(t-2)]^T。

4) Heat load collaborative prediction model

The method comprises the steps that on days 1 to 20 and 30 of a VAR model for hybrid prediction of heat load of a regional terminal integrated energy supply system, endogenous variables are air load and heat load, exogenous variables are electric load and cold load, and the model order p is determined to be 2 by using an AIC (air interface) criterion; from day 21 to day 29, the endogenous variables are gas load and heat load, the exogenous variables are electric load, and the model order is p-3:

wherein L is_g＝[l_g(t-1)，l_g(t-2)]^T，L_h＝[l_h(t-1)，l_h(t-2)]^T，

L′_g＝[l_g(t-1)，l_g(t-2)，l_g(t-3)]^T，

L′_h＝[l_h(t-1)，l_h(t-2)，l_h(t-3)]^T，。

In order to verify the effectiveness of the load hybrid prediction method for the area terminal integrated energy supply system, an independent prediction ARIMA model and a hybrid prediction model are respectively adopted to predict the sample load, and as shown in FIGS. 8-15, the prediction result curves of the area terminal integrated energy supply system under two models of electricity, gas, cold and heat loads are respectively shown.

By comparing the prediction curves of the mixed prediction method and the independent load method of the regional terminal integrated energy supply system, the method provided by the invention has strong tracking capability on the load curve. In order to further evaluate the prediction effect of the electric, gas, cold and heat loads of the area terminal integrated energy supply system, the average absolute percentage error is adopted_MAPEAnd root mean square error_RMSEMeasuring the integral error degree and the deviation degree between the predicted value and the true value, and adopting the maximum relative error_MReflecting the degree of local prediction error.

Wherein l_iThe actual load value, l' i, and n are the total predicted load values. The following table 3 shows the average absolute percentage error and the root mean square difference of each load prediction under the two prediction methods, and the following table 4 shows the maximum relative error of each load prediction under the two prediction methods. It can be obviously seen that the load prediction method provided by the invention has stronger tracking capability on each load, has smaller prediction error and has obvious advantages.

TABLE 3

TABLE 4

Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present invention and not to limit the same; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will understand that: modifications to the specific embodiments of the invention or equivalent substitutions for parts of the technical features may be made; without departing from the spirit of the present invention, it is intended to cover all aspects of the invention as defined by the appended claims.

Claims

1. The long-term load hybrid prediction method of the area terminal integrated energy supply system is characterized by comprising the following steps:

s2, calculating the collected coupling degree C of each load;

s3, establishing an ARIMA prediction model;

s5, building a VAR prediction model;

and S6, obtaining a final prediction result.

2. The long-term load hybrid prediction method for the area terminal integrated energy supply system according to claim 1, wherein the normalization processing in step S1 is as follows:

Chemical value

L_hmax、L_hminmaximum and minimum values in the thermal load analysis range;

of which ∑ l^*(i) Total load for system day i:

c) day i load entropy value e (i):

e(i)＝-(ln 4)^-1[r_e(i)ln r_e(i)+r_g(i)ln r_g(i)+r_c(i)ln r_c(i)+r_h(i)ln r_h(i)]；

d) day i differential coefficient of system load g (i): (i) 1-e (i);

e) day i system load weight ω (i):

f) electricity and gasAnd comprehensive change index of cold and heat load:

3. The long-term load hybrid prediction method of the area terminal integrated energy supply system according to claim 1, wherein the load coupling degree calculation formula is as follows:

4. The method for forecasting long-term load hybrid of an area terminal integrated energy supply system according to claim 1, wherein the establishing step of the ARIMA forecasting model in the step S3 is as follows:

5. The method for predicting the long-term load hybrid of the area terminal integrated energy supply system according to claim 1, wherein the expression of the VAR prediction model in the step S5 is as follows:

y_t＝A₁y_t-1+…+A_py_t-p+Bx_t+_tt＝1，2，…，N

wherein y is_tIs a k-dimensional internal variable vector, N is the number of samples, k × k-dimensional matrixes A1 and …, Ap is an internal variable coefficient matrix, B is an external variable coefficient matrix,_tis a k-dimensional perturbation vector.

6. The method for predicting the long-term load hybrid of the area terminal integrated energy supply system according to claim 5, wherein the building of the VAR prediction model in the step S5 comprises the following steps: