A method and system for energy demand analysis

Abstract

The application relates to the technical field of data mining, and provides an energy demand analysis method and system. The method screens and groups initial indexes of a predetermined target energy demand, to generate multiple index groups of the target energy demand; performs a frequency mixing operation on data of different frequencies in pre-acquired target energy demand data, and extracts common factors of multiple demand evaluation indexes in each index group in the mixed data generated by the frequency mixing operation in each quarter based on a mixed dynamic factor model of a Markov switching mechanism, while determining energy demand attributes corresponding to the common factors of each quarter in each index group; finally, according to a synthetic index model based on a Markov chain constructed in advance, the demand index of the target energy in each quarter is calculated according to the mixed data and the common factors of the multiple demand evaluation indexes in each index group in each quarter and the corresponding energy demand attributes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data mining, in particular to an energy demand analysis method and system. BACKGROUND

[0002] Energy has a profound impact on economic and social development and stability. With the global energy transition, traditional energy is facing unprecedented challenges and pressures.

[0003] The existing demand analysis model reflecting the change of energy demand mainly includes dynamic factor model (DFM), diffusion index model (DI) and composite index model (CI). Among them, the dynamic factor model can process multiple index data, and extract common factors from multiple energy indicators to describe the synchronous fluctuations of energy demand. However, the dynamic factor model has limitations in capturing and reflecting the asymmetry and dynamic changes of energy demand, especially in the accuracy and timeliness of dynamically adjusting the index weight to adapt to different time periods.

[0004] The composite index model and the diffusion index model are widely used to evaluate the overall situation of energy demand. Among them, the composite index model generates a comprehensive index reflecting the overall situation of energy demand by synthesizing multiple energy indicators. The core idea is to weight different energy indicators according to certain weights to obtain a comprehensive score to measure and compare energy demand conditions. The diffusion index model is mainly used to measure the diffusion degree of energy indicators to reflect the overall situation of energy demand. The core idea is to construct an index reflecting the energy demand situation by counting the proportion of the diffusion state (contraction or diffusion) in the energy indicators. SUMMARY

[0005] The purpose of the present application is to provide an energy demand analysis method and system to solve or alleviate the problems existing in the prior art.

[0006] In order to achieve the above purpose, the present application provides the following technical scheme:

[0007] The present application provides an energy demand analysis method, comprising: step S101, screening and grouping the initial indicators of the predetermined target energy demand, generating multiple indicator groups of the target energy demand; wherein each of the indicator groups contains multiple demand evaluation indicators;

[0008] Step S102, mixing the data of different frequencies in the pre-acquired target energy demand data, and based on the mixed frequency dynamic factor model of Markov switching mechanism, extracting the common factors of each of the demand evaluation indicators in each of the indicator groups every quarter generated by the mixed frequency operation, and determining the energy demand attributes corresponding to the common factors of each of the indicator groups every quarter.

[0009] Step S103: Based on the pre-constructed synthetic index model based on Markov chains, and according to the mixed frequency data and the common factors of multiple demand evaluation indicators in each quarter within each index group and their corresponding energy demand attributes, calculate the demand index of the target energy for each quarter.

[0010] Preferably, in step S101, based on the time difference correlation coefficient analysis method, multiple initial indicators of the predetermined target energy demand are screened and grouped to generate a leading indicator group, a consistency indicator group, and a lagging indicator group for the target energy demand.

[0011] Preferably, in step S102, the mixing dynamic factor model includes: a single-factor observation equation, a factor dynamic equation, and a state-space equation;

[0012] The single-factor observation equation is as follows:

[0013]

[0014] In the formula, Y m,t,1 Y represents the mixing data for the monthly frequency within the t-th quarter. q,t,2 The mixing data for the quarterly frequency in the t-th quarter; α is the intercept term, and β is the factor loading coefficient; The common factor of multiple demand evaluation indicators in the j-th indicator group in the t-th quarter; The data error of the mixed-frequency data for the j-th indicator group in the t-th quarterly time period; t and j are both positive integers, and t>1;

[0015] The dynamic equation for the factor is:

[0016]

[0017] In the formula, φ f The lag operator represents the common factor of multiple demand evaluation indicators in the j-th indicator group in the (t-1)th quarter. The common factor of multiple demand evaluation indicators in the j-th indicator group in the t-th quarter The lag effect; S t For the multiple demand evaluation indicators of the j-th indicator group in the t-th quarter, ν represents the Markov state variable, characterizing the probability of demand shift of the target energy in the j-th indicator group from the demand in the (t-1)-th quarter to the demand in the t-th quarter; t Let be the random error of the mixing data of the j-th indicator group in the t-th quarter, which follows a mean of 0 and a variance of ∈ v The normal distribution;

[0018] The state-space equation is:

[0019]

[0020] wherein Y t is the mixed frequency data in the tth quarter, including the mixed frequency data Y m,t,1 in the tth quarter of monthly frequency and the mixed frequency data Y q,t,2 in the tth quarter of quarterly frequency; A is a constant term vector, H is a load matrix, X t is the state variable matrix of the tth quarter of the jth index group, Ξ t is an intercept vector; is the state variable matrix of the t-1th quarter of the jth index group X t-1 is a coefficient matrix of the lagged effect of the state variable matrix X t of the tth quarter of the jth index group; V t is the error matrix of the mixed frequency data of the tth quarter of the jth index group.

[0021] Preferably, the step S103 comprises: calculating a symmetric change rate C ij (t) of the ith demand evaluation index in the jth index group in the tth quarter, so as to determine a normalization factor A ij of the ith demand evaluation index in the jth index group in n quarters, and determining a normalized change rate S ij (t) of the ith demand evaluation index in the jth index group in the tth quarter according to the symmetric change rate C ij (t) and the normalization factor A ij ; wherein t, i, j and n are positive integers, and t>1.

[0022] According to the normalized change rate S ij (t), an average change rate R j (t) of k j demand evaluation indexes in the jth index group in the tth quarter is calculated, so as to determine a normalization factor F j of the jth index group in n quarters; wherein k j is a positive integer.

[0023] Based on the normalization factor F j , the average change rate R j (t) is corrected to obtain a first normalized average change rate V j (t, S t ) of the k j demand evaluation indexes in the jth index group in the tth quarter.

[0024] The average growth rate r′ of the initial composite index of the predetermined multiple sets of indicators j And the predetermined target energy demand trend G r For the first standardized average rate of change V j (t,S t ) Perform trend adjustments to obtain k in the j-th indicator group. j The second standardized rate of change V′ of the aforementioned demand evaluation indicators in the t-th quarter j (t,S t );

[0025] According to the second standardized rate of change V′ j (t,S t Calculate the trend-adjusted composite index I′ of the j-th indicator group in the t-th quarter. j (t), and based on the average composite index of the j-th indicator group after trend adjustment over n quarters. and the pre-determined benchmark index I′ of the j-th group of indicators j (1) Calculate the demand index CI corresponding to the j-th indicator group of the target energy in the t-th quarter. j (t).

[0026] Preferably, in step S103, based on the mixing data, according to the formula:

[0027]

[0028] Calculate the symmetric rate of change C of the i-th demand evaluation indicator in the j-th indicator group in the t-th quarter. ij (t); where Y is... ij (t) is the data matrix of the mixed-frequency data of the i-th demand evaluation indicator in the j-th indicator group in the t-th quarter; Y ij (t-1) is the data matrix of the mixed-frequency data of the i-th demand evaluation indicator in the j-th indicator group in the (t-1)-th quarter; λ is the common factor of multiple demand evaluation indicators in the j-th indicator group in the t-th quarter. The adjustment parameters are: t, i, and j are all positive integers, and t > 1;

[0029] According to the formula:

[0030]

[0031] Determine the standardized factor A of the i-th demand evaluation indicator in the j-th indicator group over n quarters. ij Where n is a positive integer;

[0032] According to the formula:

[0033]

[0034] determining a standardized change rate S ij (t) of the i

[0035] Preferably, in step S103, according to the formula:

[0036]

[0037] determining an average change rate R j (t) of k j demand evaluation indexes in the j ij th index group in the t ij th quarter; in the formula, w j is a weight of the standardized change rate S j (t) of the i t th demand evaluation index in the j j th index group in the t th quarter; wherein t, i, j, and k are positive integers, and t>1;

[0038] According to the formula:

[0039]

[0040] determining a standardized factor F j of the j th index group in n quarters; in the formula, S t is a data error of the mixed data of the j j th index group in the t th quarter; S j 2(t) is an average change rate of k j demand evaluation indexes in the 2 j nd index group in the t j th quarter; wherein the 2 t nd index group is a consistency index group; and n is a positive integer;

[0041] According to the formula:

[0042]

[0043] correcting the average change rate R j (t) of the k j demand evaluation indexes in the j j th index group in the t j th quarter to obtain a first standardized average change rate V t (t, S

[0044] Preferably, the index group comprises: a leading index group, a consistency index group and a lagging index group;

[0045] The average growth rate of each index sequence in the consistency index group is calculated by the compound interest formula to determine the target demand trend G of the target energy r ;

[0046] Based on the exponential synthesis method, the initial synthetic index of the leading index group, the consistency index group and the lagging index group is determined respectively, and then the average growth rate r′ of the initial synthetic index of the leading index group, the consistency index group and the lagging index group is calculated by the compound interest formula j ;

[0047] Wherein, according to the formula:

[0048]

[0049] The Markov state variable S of the jth index group based on the jth index group is determined t The initial synthetic index I j of the jth index group in the tth quarter is adjusted; in the formula, I j (t-1) is the Markov state variable S of the jth index group based on the jth index group in the t-1th quarter t-1 The initial synthetic index of the jth index group in the t-1th quarter is adjusted; V j (t,S t ) is the first standardized average change rate V j of the k j th demand evaluation index in the jth index group in the tth quarter t (t,S j ); t, j, k j are positive integers, and t>1.

[0050] Preferably, in step S103, according to the formula:

[0051] V′ t (t,S j )=V t (t,S r )+(G j -r′ j )

[0052] The first standardized average change rate V j of the k t th demand evaluation index in the jth index group in the tth quarter is adjusted to obtain the second standardized change rate V′ j (t,St wherein G r is a predetermined target demand trend of the target energy source, r′ j is an average growth rate of the initial composite index of the plurality of index groups;

[0053] According to the formula:

[0054]

[0055] calculating a demand index CI j (t) corresponding to the jth index group of the target energy source in the tth quarter; wherein I′ j (t) is the trend-adjusted composite index of the jth index group in the tth quarter; I′ j (t-1) is the trend-adjusted composite index of the jth index group in the (t-1)th quarter; is the average trend-adjusted composite index of the jth index group in n quarters; I′ j (1) is a predetermined baseline index of the jth index group; wherein t, j, k j are all positive integers, and t>1.

[0056] Preferably, further comprising: determining a demand index dynamic of each quarter of the target energy source based on a preset index dynamic interval, to generate a demand transition probability matrix of the target energy source in n quarters according to the demand index of each quarter of the target energy source;

[0057] determining a demand dynamic transition vector of the current quarter of the target energy source according to the demand index dynamic of the current quarter of the target energy source;

[0058] calculating a demand transition probability vector of the next quarter of the target energy source according to the demand transition probability matrix and the demand dynamic transition vector of the current quarter of the target energy source, and selecting an index dynamic interval of a maximum element value in the demand transition probability vector as a demand index dynamic of the next quarter of the target energy source.

[0059] The present application also provides an energy demand analysis system, comprising: an index screening and grouping unit configured to screen and group initial indexes of a predetermined target energy demand to generate a plurality of index groups of the target energy demand; wherein each of the index groups comprises a plurality of demand evaluation indexes;

[0060] The factor extraction and attribute determination unit is configured to perform mixing operation on data of different frequencies in the target energy demand data obtained in advance, and extract common factors of each of the demand evaluation indexes in each of the index groups in the mixed data generated by the mixing operation based on a mixing dynamic factor model based on Markov switching mechanism, and determine energy demand attributes corresponding to the common factors of each of the index groups in each quarter.

[0061] The demand index determination unit is configured to calculate demand indexes of the target energy in each quarter according to the mixed data and the common factors of each of the demand evaluation indexes in each of the index groups in each quarter and the corresponding energy demand attributes based on a synthetic index model based on Markov chain constructed in advance.

[0062] Beneficial effects:

[0063] In the energy demand analysis method provided in the specification, the initial indexes of the target energy demand are screened and grouped to generate a plurality of index groups of the target energy demand, and each index group contains a plurality of demand evaluation indexes. Meanwhile, the data of different frequencies in the target energy demand data obtained in advance is subjected to mixing operation, and the common factors of each of the demand evaluation indexes in each index group in the mixed data generated by the mixing operation are extracted based on a mixing dynamic factor model based on Markov switching mechanism, and the energy demand attributes corresponding to the common factors of each index group in each quarter are determined. Finally, the demand indexes of the target energy in each quarter are calculated according to the mixed data and the common factors of each of the demand evaluation indexes in each of the index groups in each quarter and the corresponding energy demand attributes based on a synthetic index model based on Markov chain constructed in advance.

[0064] Through the processing of the mixed data based on the mixing dynamic factor model based on Markov switching mechanism, the inconsistency of data and the information loss problem existing in the traditional synthetic index method when facing data of different frequencies are effectively avoided. The Markov switching mechanism is introduced, so that the model can automatically identify different stages of energy demand, and flexibly adjust the index weight according to the stage change of energy demand, effectively avoiding the artificial subjective intervention existing in the traditional method, ensuring that the weight distribution is more scientific and reasonable, improving the objectivity and adaptability of the model. At the same time, combined with the Markov switching mechanism and the dynamic factor model, the asymmetric change in energy demand is dynamically identified and captured, and the weight is adjusted according to the demand change, accurately reflecting the asymmetric characteristics of energy demand, improving the adaptability and accuracy of the demand index to the change of energy demand, enhancing the real-time monitoring and prediction ability of the demand index, providing more accurate and timely demand trend analysis, and improving the adaptability and foresight of the model. BRIEF DESCRIPTION OF DRAWINGS

[0065] Figure 1A flowchart of an energy demand analysis method according to some embodiments of the present application;

[0066] Figure 2 A logic diagram of an energy demand analysis method according to some embodiments of the present application;

[0067] Figure 3 A hardware diagram of an energy demand analysis method according to some embodiments of the present application;

[0068] Figure 4 A structure diagram of an energy demand analysis system according to some embodiments of the present application. DETAILED DESCRIPTION

[0069] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. Each example is provided by way of explanation of the present application, rather than limitation of the present application. In fact, those skilled in the art will appreciate that modifications and variations can be made in the present application without departing from the scope or spirit of the present application. For example, features shown or described as part of one embodiment can be used in another embodiment to produce yet another embodiment. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application should fall within the scope of protection of the embodiments of the present application.

[0070] In existing demand analysis models reflecting changes in energy demand, whether dynamic factor model (DFM), diffusion index (DI) model or composite index model (CI), there are great problems in identifying and predicting changes in energy demand, mainly as follows:

[0071] First, data processing problem: (1) Insufficient integration of different frequency data - existing demand analysis models are usually based on data of the same frequency, such as quarterly or monthly data, which makes it difficult to flexibly handle monthly and quarterly frequency data contained in energy-related indicator data, resulting in insufficient data utilization, lack of real-time and accuracy in reflecting energy demand; (2) Low system automation and data processing efficiency - existing models rely on manual calculation and judgment when calculating and predicting energy demand index, involving a large amount of manual data processing and calculation work, which not only is tedious and inefficient, but also can lead to poor data processing accuracy and increase the risk of human error.

[0072] Second, the subjectivity problem of index selection and weight assignment: existing demand analysis models generally rely on the selection of a group of energy indicators and the assignment of their weights, which is highly subjective; different studies may choose different indicators and weight schemes based on their own preferences, resulting in differences in demand index results, lack of consistency and objectivity.

[0073] Third, the problem of energy demand identification: Insufficient sensitivity to asymmetric energy demand. Existing composite index models usually assume that energy demand is symmetric, i.e., the expansion and contraction periods are similar in length and amplitude. However, actual energy demand often exhibits asymmetric characteristics, such as expansion periods that are longer than recession periods. Existing models cannot effectively capture this asymmetry, resulting in limited accuracy and timeliness of demand indices.

[0074] Fourth, the problem of dynamic and real-time limitations: Traditional demand analysis models rely on fixed weights and historical data, making it difficult to reflect the impact of rapidly changing events in a timely manner. This characteristic makes the demand index slow to respond to changes in demand, affecting its real-time monitoring and prediction capabilities and reducing its applicability in dynamic demand changes.

[0075] Fifth, the problem of complexity and transparency: Although composite index models can provide a comprehensive indicator, the model calculation process is quite complex, making it difficult to fully evaluate its logic and effectiveness, which affects the transparency of the model and the public's trust in the model.

[0076] Sixth, the problem of insufficient trend prediction ability: Existing composite index models usually rely on leading indices for trend prediction. However, traditional leading index methods often assume that future demand trends are closely related to current leading indicators and do not fully consider complex nonlinear relationships and long-term trends in demand changes. Therefore, when dealing with complex cyclical fluctuations in energy demand, the prediction results may be inaccurate or lagging.

[0077] Seventh, the problem of limitations of manual measurement: In the process of predicting energy demand, manual calculation and judgment are usually relied upon, which involves a large amount of manual data processing and calculation work. This manual calculation method has many drawbacks. On the one hand, it is time-consuming and labor-intensive and is easily influenced by human factors, resulting in inconsistent calculation results and reducing accuracy. On the other hand, the data processing steps in the manual calculation process are complex and prone to omissions or errors, especially when dealing with large-scale complex data. Manual calculation has obvious shortcomings in efficiency, reliability, and transparency.

[0078] Based on this, this specification proposes a Mixed-Frequency Markov Dynamic Factor Demand Analysis Model (MF-MS-DFM-CI) based on the Composite Index. By introducing the Mixed-Frequency Dynamic Factor Model (MF-DFM), Markov Switching Mechanism (MS), and Composite Index Method (CI), it not only systematically solves the significant limitations of existing Dynamic Factor Model (DFM), Composite Index Model (CI), and Diffusion Index Model (DI) in processing inter-frequency data, capturing energy demand asymmetry, dynamically adjusting index weights, and improving real-time performance and forecasting capabilities, but also achieves a high degree of automation and systematization of demand forecasting by developing an automated system based on Python, Tkinter, and MySQL, significantly improving computational efficiency, accuracy, and user experience.

[0079] like Figures 1 to 3 As shown, the energy demand analysis method includes:

[0080] Step S101: Screen and group the initial indicators of the predetermined target energy demand to generate multiple indicator groups for the target energy demand.

[0081] This specification collects initial indicators that can reflect the target energy demand, such as the target total energy consumption, energy utilization efficiency, average level of industrial automation (percentage of automated equipment), energy saving and emission reduction rate, target energy output, pollutant emission concentration (smoke, sulfur dioxide, nitrogen oxides, etc.), target energy unit energy consumption, unit energy consumption, target energy intensity, carbon emission intensity, renewable energy penetration rate (percentage of renewable energy in total energy consumption), clean energy share, and energy efficiency of energy-consuming equipment.

[0082] Furthermore, based on the time-difference correlation coefficient analysis method, and according to the statistical attributes of the indicators, multiple initial indicators for the determined target energy demand are screened and then grouped. The screened target energy demand evaluation indicators are divided into multiple indicator groups, each containing multiple demand evaluation indicators. In this way, leading indicator groups, consistency indicator groups, and lagging indicator groups for the target energy demand are generated through indicator grouping.

[0083] Specifically, before selecting indicators, a benchmark indicator that reflects energy demand is first determined. In this specification, the target total energy consumption, which is highly consistent with the trend of energy demand changes, is used as the benchmark indicator. After determining the benchmark indicator, the time-difference correlation coefficient analysis method is used to calculate the time-difference correlation coefficients between multiple initial indicators of the target energy demand and the benchmark indicator. Based on the time-difference correlation coefficients, indicators are selected to identify multiple demand evaluation indicators, which are then divided into leading indicator group, consistency indicator group, and lagging indicator group.

[0084] In a specific example, define X = {x1, x2, …, xn} is the benchmark index, and x n} is the other to-be-selected index (i.e., initial index), and r l is the time difference correlation coefficient, according to the formula:

[0085]

[0086] The time difference correlation coefficient r l between the t-lth initial index value and the tth benchmark index value is calculated; in the formula, l is the number of leading or lagging periods (a negative value is a leading index, a positive value is a lagging index, and zero is a consistent index), l = {0, ±1, ±2, …, ±L}, L is a positive integer; n l is the number of benchmark index values, i.e., the number of elements in the benchmark index, y t is the tth benchmark index value, x t-l is the data of the other to-be-selected index in the previous period, i.e., the t-lth initial index value, is the average value of the other to-be-selected index, i.e., the average value of n (n is a positive integer) initial index values, is the average value of n l (n l is a positive integer) benchmark index values.

[0087] By calculating the correlation coefficients of each candidate index with the benchmark index at different delay periods, the index with the largest time difference correlation coefficient is selected, and the number of leading or lagging periods of the index relative to the benchmark index is determined, and the index group is divided. In this specification, the index with a time difference correlation coefficient exceeding 0.5 is screened, and the screened initial index with a time difference correlation coefficient exceeding 0.5 is divided into a leading index group, a consistent index group, and a lagging index group according to the time difference of at least one quarter. Here, the index in the leading index group is defined as a leading index, the index in the consistent index group is defined as a consistent index, and the index in the lagging index group is defined as a lagging index.

[0088] In step S102, the mixed frequency data generated by the mixed frequency operation on the pre-acquired target energy demand data is subjected to mixed frequency operation, and the common factors of each quarter in each index group are extracted based on the mixed frequency dynamic factor model of the Markov switching mechanism, and the energy demand attributes corresponding to the common factors of each quarter in each index group are determined.

[0089] In this application, the mixed-frequency dynamic factor model based on the Markov switching mechanism effectively integrates the high-frequency data (monthly data) and low-frequency data (quarterly data) in the obtained target energy demand data, and uses linear interpolation and nearest neighbor interpolation methods to fill in the missing values in the monthly and quarterly data to ensure the integrity and timeliness of the data; the X-13ARIMA-SEATS method is used to perform seasonal adjustment and smoothing processing on the monthly data to further ensure the consistency and comparability of the data at different time granularities.

[0090] The mixed-frequency dynamic factor model based on the Markov switching mechanism performs mixed-frequency operations on data of different frequencies, effectively avoiding the inconsistency of traditional composite index methods when facing data of different frequencies, and more accurately capturing the asymmetric changes in energy demand. Here, among the obtained target energy demand data, the mixed-frequency data at the monthly frequency within the t-th quarter is defined as Y m,t,1 , and the mixed-frequency data at the quarterly frequency within the t-th quarter is defined as Y q,t,2 , for {Y m,t,1 , -∞ < t < +∞} and {Y q,t,2 , -∞ < t < +∞}, a single-factor observation equation is established, specifically:

[0091]

[0092] In the formula, α is the intercept term, and β is the factor loading coefficient; is the common factor of multiple demand evaluation indicators of the j-th indicator group in the t-th quarter; is the data error of the mixed-frequency data of the j-th indicator group in the t-th quarter. Both t and j are positive integers, and t > 1. Here, through the nonlinear fitting method, the mixed-frequency data Y m,t,1 , Y q,t,2 and the common factor data error are nonlinearly fitted to determine the intercept term α and the factor loading coefficient β.

[0093] A factor dynamic equation is established:

[0094]

[0095] In the formula, φ f is the lag operator, representing the lag effect of the common factor of multiple demand evaluation indicators of the j-th indicator group in the (t - 1)-th quarter on the common factor of multiple demand evaluation indicators of the j-th indicator group in the t-th quarter; υ t is the error term of the mixed-frequency data of the j-th indicator group in the t-th quarter, and υ t ∼ N(0, ∈ υ ), that is, the random error υ tFollows a pattern with mean 0 and variance ∈ v It follows a normal distribution.

[0096] S t Let S be the Markov state variable of multiple demand evaluation indicators in the j-th indicator group in the t-th quarter. t It follows a first-order Markov process and has a definite state transition probability P(S). t |S t-1 ), representing the state transition probability of the target energy in the j-th indicator group from the energy demand in the (t-1)-th quarter to the energy demand in the t-th quarter. In this way, the intercept term α and the factor loading coefficient β can be dynamically adjusted to adapt to the dynamic changes in the target energy demand and reflect the expansion or contraction of the target energy demand.

[0097] To better adapt to the periodic characteristics of target energy demand data, this specification employs a mixing dynamic factor model based on a Markov switching mechanism to capture the asymmetric fluctuation relationship between high-frequency data (monthly data) and low-frequency data (quarterly data). Specifically, the state-space equations of the mixing dynamic factor model based on the Markov switching mechanism are established as follows:

[0098]

[0099] In the formula, Y t This refers to the mixing data within the t-th quarter, including the mixing data for the monthly frequencies within the t-th quarter. m,t,1 Mixed frequency data Y of the quarterly frequency in the t-th quarter q,t,2 ;X t Let Y be the state variable matrix for the j-th indicator group in the t-th quarter, representing the state probability of energy demand corresponding to the j-th indicator group in the t-th quarter of the mixed-frequency data. Based on this, the measurement equation (i.e., Y) is used to... t =A+HX t ) describes the observed variable Y t With unobservable state variable X t The interdependence between them is shown in the diagram, where A is the constant term vector and H is the loading matrix, describing the observed variable Y. t With unobservable state variable X t The relationship between the two is analyzed using a nonlinear fitting method on the mixing data Y. t and state variable matrix X t Determined by nonlinear fitting.

[0100] In the state-space equations, through the state equations (i.e. ) describes the state variable X t The dynamic path is controlled by a Markov mechanism to effectively capture asymmetric changes in energy demand. In the state-space equations, Ξt is the intercept vector, i.e., the vector representation of the intercept term a; is the state variable matrix X of the (t-1)th quarter of the jth index group t-1 is the state variable matrix X of the tth quarter of the jth index group t is the coefficient matrix of the lagged effect of X t is the error matrix of the mixed frequency data of the tth quarter of the jth index group, i.e., the random error t is the vector representation of the mixed frequency data of the tth quarter of the jth index group.

[0101] In the present specification, through the mixed frequency dynamic factor model, the high-frequency (monthly) and low-frequency (quarterly) data in the target energy demand data are integrated to extract common factors describing the synchronism of energy demand; the common factors follow a first-order autoregressive AR(1) process, and the Markov mechanism has two states, i.e., S t = 0 or S t = 1, S t = 0 corresponds to energy demand μ1, S t = 1 corresponds to energy demand μ2, μ1 < 0 represents the contraction of demand for the target energy; μ2 > 0 represents the expansion of demand for the target energy. Through the mixed frequency dynamic factor model based on the Markov switching mechanism, the common factors are extracted from the high-frequency and low-frequency data of the target energy demand data, so that the common factors dynamically change with the change of the Markov state variable, capture the asymmetric changes of energy demand, effectively ensure the dynamic adjustment of weight distribution in different time periods, and effectively enhance the adaptability to the dynamic changes of energy demand.

[0102] In step S103, according to the pre-constructed composite index model, the demand index of the target energy in each quarter is calculated according to the mixed frequency data and the common factors of each demand evaluation index in each quarter in each index group and the corresponding energy demand attribute.

[0103] In the present specification, the mixed frequency dynamic factor model based on the Markov switching mechanism can extract common factors synchronously describing the amount of energy demand; meanwhile, the Markov switching mechanism is introduced, and each weight in the demand index is dynamically adjusted through state transition, so as to ensure the adaptive adjustment according to different time periods (demand expansion or demand contraction).

[0104] Further, in order to more accurately calculate the demand index of the target energy, the extracted common factors The traditional composite index model is improved by introducing Markov switching mechanism, so as to realize dynamic allocation of weights. The improved model not only makes full use of the implicit information in the data, but also makes the demand index more sensitive to the periodic fluctuations of energy demand by adjusting the state-dependent weight, and then combines the dynamic common factor with the traditional composite index method to effectively enhance the sensitivity and accuracy of the demand index, so as to better reflect the dynamic changes of energy demand.

[0105] Firstly, the symmetric change rate C ij (t) of the i-th demand evaluation index in the j-th index group in the t-th quarter is calculated to determine the standardized factor A ij of the i-th demand evaluation index in the j-th index group in n quarters; wherein t, i, j and n are positive integers, and t>1. In this regard, a dynamic common factor is introduced in the calculation of the symmetric change rate C ij (t) to represent the common change of multiple demand evaluation indexes. Specifically, according to the mixed data, the symmetric change rate C ij (t) of the i-th demand evaluation index in the j-th index group in the t-th quarter is calculated according to the formula:

[0106]

[0107] ij (t) is the data matrix of the mixed data of the i-th demand evaluation index in the j-th index group in the t-th quarter; Y ij (t-1) is the data matrix of the mixed data of the i-th demand evaluation index in the j-th index group in the t-1-th quarter; and λ is the adjustment parameter of the common factor of the multiple demand evaluation indexes in the j-th index group in the t-th quarter, which is used to control the influence of the common factor in the calculation of the symmetric change rate.

[0108] After obtaining the symmetric change rate C ij (t) of the i-th demand evaluation index in the j-th index group in the t-th quarter, the standardized factor A ij of the i-th demand evaluation index in the j-th index group in n quarters is determined according to the formula:

[0109]

[0110] Further, the standardized change rate S ij (t) of the i-th demand evaluation index in the j-th index group in the t-th quarter is determined according to the symmetric change rate C ij (t) and the standardized factor A ij . Specifically, the standardized change rate S ij (t) of the i-th demand evaluation index in the j-th index group in the t-th quarter is calculated according to the formula:

[0111]

[0112] Determine the standardized rate of change S of the i-th demand evaluation indicator in the j-th indicator group in the t-th quarter. ij (t). Thus, by standardizing the symmetric rate of change, the impact of a single high-volatility indicator on the composite index can be effectively reduced.

[0113] The average rate of change of multiple demand evaluation indicators within each indicator group directly reflects the overall change of each indicator group. In this specification, the Markov state variable S is used to represent this. t Introducing the average rate of change allows for dynamic adaptation to changes in energy demand. Specifically, based on the standardized rate of change S... ij (t) Calculate k in the j-th index group j The average rate of change R of the demand evaluation indicators in the t-th quarter j (t); specifically, according to the formula:

[0114]

[0115] Determine k in the j-th indicator group j The average rate of change R of the demand evaluation indicators in the t-th quarter j (t); where w ij S is the standardized rate of change of the i-th demand evaluation indicator in the j-th indicator group in the t-th quarter. ij The weight of (t) can be determined. Here, the standardized rate of change S of the i-th demand evaluation indicator in the j-th indicator group during the t-th quarter can be determined using methods such as correlation coefficient method, factor analysis method, and entropy weight method. ij The weight w of (t) ij This is to ensure that the importance of each indicator in the composite index is reasonably reflected.

[0116] Different Markov states have different energy demands, and different energy demands correspond to different average rate of change adjustment factors. Here, based on k in the j-th index group... j The average rate of change R of the demand evaluation indicators in the t-th quarter j (t) Calculate the corresponding Markov state variable S t The standardization factor is as follows. Specifically, according to the formula:

[0117]

[0118] Determine the standardized factor F of the j-th indicator group in the n-th quarter. j In the formula, S represents the data error of the mixed-frequency data for the j-th indicator group in the t-th quarter; tLet R2(t) be the Markov state variable for multiple demand evaluation indicators in the j-th indicator group in the t-th quarter, representing the state transition probability of the target energy in the j-th indicator group from the energy demand in the (t-1)-th quarter to the energy demand in the t-th quarter; R2(t) is the k-th indicator in the second indicator group (i.e., the consistency indicator group). j The average rate of change of each demand evaluation indicator in the t-th quarter.

[0119] Then, based on the current Markov state variable S t Adjust the standardized average rate of change, i.e., based on the standardization factor F. j For the average rate of change R j (t) is corrected to obtain k in the j-th index group. j The adjusted standardized average rate of change of each demand evaluation indicator in the t-th quarter, i.e., the first standardized average rate of change V. j (t,S t Specifically, according to the formula:

[0120]

[0121] For the j-th indicator group, k j The average rate of change R of the demand evaluation indicators in the t-th quarter j (t) is corrected to obtain k in the j-th index group. j The first standardized average rate of change V of the demand evaluation indicators in the t-th quarter j (t,S t ).

[0122] In the j-th indicator group, k j After calculating the adjusted standardized average rate of change of each demand evaluation indicator in the t-th quarter, and combining it with an initial composite index determined through an index synthesis method, the average rate of change of the indicator group is trend-adjusted. Specifically, in the calculation of the initial composite index, the average growth rate of each indicator sequence in the consistent indicator group is calculated using a compound interest formula to determine the target trend G of the target energy. r Specifically, the average growth rate r of each indicator sequence in the consistency indicator group is first calculated using the compound interest formula. i That is, according to the formula:

[0123]

[0124] Calculate the average growth rate r of each indicator series in the consistency indicator group. i ; In the formula sa, These are the average values ​​of the first and last cycles of the i-th demand evaluation indicator in the consistency indicator group, respectively. respectively the first cycle period and the last cycle period of the u-th demand evaluation index in the consistency index group; m i is the cycle period of the i-th demand evaluation index in the consistency index group from the first cycle center to the last cycle center; i = 2, 3, …, k2, k2 is the number of demand evaluation indexes in the consistency index group. i (t) is the i-th demand evaluation index in the t-th quarter in the consistency index group.

[0125] Then, according to the average growth rate r of each index sequence in the consistency index group i , the average growth rate of the consistency index group, i.e. the target trend G r , is calculated.

[0126]

[0127] The average growth rate of the consistency index group, i.e. the target trend G r , is determined.

[0128] At the same time, based on the exponential synthesis method, the initial synthetic indexes of the leading index group, the consistency index group and the lagging index group are determined respectively. Specifically, according to the formula:

[0129]

[0130] The Markov state variable S t based on the t-th quarter of the j-th index group is determined. j The initial synthetic index I j (t) of the j-th index group in the t-th quarter is adjusted; in the formula, I t-1 (t-1) is the initial synthetic index of the j-th index group in the t-1-th quarter adjusted. j V t (t, S j ) is the first standardized average change rate V j (t, S t ) of the k j -th demand evaluation index in the j-th index group in the t-th quarter.

[0131] Then, the average growth rate r′ j of the initial synthetic indexes of the leading index group, the consistency index group and the lagging index group is calculated again using the compound interest formula respectively. Specifically, according to the formula:

[0132]

[0133] The average growth rate r′j In the formula, These are the average values ​​of the j-th demand evaluation indicator in the first and last cycles, respectively. , which represent the first and last cycle periods of the j-th demand evaluation indicator in the indicator group, respectively; m represents the number of cycles from the first cycle center to the last cycle center of the j-th demand evaluation indicator in the indicator group.

[0134] Next, based on the average growth rate r′ of the initial composite index of the determined multiple indicator groups... j And the target energy target trend G r For the first standardized average rate of change V j (t,S t ) Perform trend adjustments to obtain k in the j-th indicator group j The adjusted, standardized average rate of change of each demand evaluation indicator in the t-th quarter, i.e., the second standardized rate of change V′ j (t,S t Specifically, according to the formula:

[0135]

[0136] For the j-th indicator group, k j The adjusted standardized average rate of change V of the demand evaluation index in the t-th quarter. j (t,S t After trend adjustment, the second standardized rate of change V′ is obtained. j (t,S t In the formula, G r For a predetermined target energy trend, r′ j This represents the average growth rate of the initial composite index of multiple indicator groups.

[0137] In this specification, the target trend G of the consistency index group is used... r As the target energy source, the target trend is used, and trend adjustments are made to the leading and lagging indicator groups respectively to effectively ensure that the composite index of all indicator groups remains consistent in the long-term trend. Furthermore, based on the adjusted and corrected standardized average rate of change (i.e., the second standardized rate of change) V′... j (t,S t Calculate the trend-adjusted composite index I′ of the j-th indicator group in the t-th quarter. j (t). Specifically, according to the formula:

[0138]

[0139] Calculate the trend-adjusted composite index I′ of the j-th indicator group in the t-th quarter. j (t); where I′j (t-1) is the jth index group after the trend adjustment in the t-1th quarter.

[0140] Finally, based on the average composite index of the jth index group after the trend adjustment in n quarters and the benchmark index I' of the jth index group determined j (1), the demand index CI j (t) corresponding to the jth index group of the target energy in the tth quarter is calculated. Specifically, according to the formula:

[0141]

[0142] The demand index CI j (t) corresponding to the jth index group of the target energy in the tth quarter is determined. In the formula, is the average composite index of the jth index group after the trend adjustment in n quarters; I' j (1) is the benchmark index of the hth index group determined in advance.

[0143] In this specification, the mixing frequency dynamic factor model and the Markov mechanism are combined to improve the traditional composite index model, so that the improved composite index model can be adjusted in real time according to data updates. Based on the Markov mechanism, the weights of the indicators are dynamically adjusted according to different stages of energy demand, the changes of different frequency data are reflected in real time through the mixing frequency dynamic factor model, and the dynamic factors are used to adjust the weights and capture the changes of energy demand. At the same time, through the automatic adjustment and transparent calculation process of the model, the human subjective interference is reduced, and the transparency and interpretability of energy prediction are effectively improved.

[0144] Further, after obtaining the demand index of the target energy in each quarter, the demand index of the target energy in the future quarter can be predicted through the Markov chain model or the long short-term memory neural network. First, according to the demand index of the target energy in each quarter, based on the preset state interval, the demand index state of the target energy in each quarter is determined. Specifically, according to the mean and the standard deviation σ of the demand index of the target energy in all quarters, the energy demand index state interval is divided, the demand index state of the interval is defined as "supercooling", the demand index state of the interval is defined as "slightly cold", the demand index state of the interval is defined as "normal", the demand index state of the interval is defined as "slightly hot", and the demand index state of the interval is defined as "overheating".

[0145] Under "normal" conditions, energy demand is relatively stable; under "slightly cold" conditions, energy demand relatively contracts; under "excessively cold" conditions, energy demand drops sharply; under "slightly hot" conditions, energy demand rises relatively slowly; and under "excessively hot" conditions, energy demand expands excessively. In this specification, 1, 2, 3, 4, and 5 are used to represent the five states of the target energy source: "excessively cold," "slightly cold," "normal," "slightly hot," and "excessively hot," respectively, and are represented by the formula:

[0146]

[0147] For the i-th state S of the target energy i The number of times r appears i Perform statistics; in the formula, n is the total number of quarters, t and n are both positive integers, and x t The demand index CI for target energy in the t-th quarter j (t); I(x) t ∈S i ) is a linear function, when the condition I(x) is satisfied t ∈S i ) = 1, otherwise I(x) t ∈S i ) = 0.

[0148] At the same time, according to the formula:

[0149]

[0150] For the target energy in the t-th quarter, the i-th state S i Transition to the j-th state S in the (t+1)-th quarter j The number of times z appears ij Perform statistics; in the formula, i, j ∈ n, and i, j, and n are all positive integers. I(x t ∈S i ;x t+1 ∈S j I(x) is a linear function, and when the condition is satisfied, I(x) t ∈S i ;x t+1 ∈S j ) = 1, otherwise I(x) t ∈S i ;x t+1 ∈S j ) = 0.

[0151] According to the i-th state S of the target energy i The number of times r appears i And the target energy in the i-th state of the t-th quarter S i Transition to the j-th state S in the (t+1)-th quarter j The number of times z appearsij Generate the state transition probability matrix of the target energy over n quarters. Specifically, according to the formula:

[0152]

[0153] Calculate the i-th state S of the target energy in the t-th quarter. i Transition to the j-th state S in the (t+1)-th quarter j State transition probability P ij Then, establish the state transition probability matrix P = (P_n) for the target energy over n quarters. ij ) 5×5 .

[0154] Simultaneously, based on the current quarter's demand index status of the target energy source, the state transition vector for the current quarter is determined. Specifically, according to the formula:

[0155] v T ={P 11 ,…,P 1i ,…,P 1n}

[0156] Determine the state transition vector v of the target energy source for the current quarter. T Where n = 5. For the current quarter's state S... i Then, in its corresponding state transition vector, P 1i =1, and the rest of the elements are 0.

[0157] Finally, based on the state transition probability matrix P and the target energy's current quarterly state transition vector v... T Calculate the state transition probability vector for the target energy source in the next quarter, and select the state interval with the largest element value in the state transition probability vector as the demand index state for the target energy source in the next quarter. Specifically, according to the formula:

[0158]

[0159] Determine the demand index state of the target energy source for the next quarter, that is, select the state corresponding to the highest state transition probability.

[0160] In a specific example, the demand index state for each quarter is determined based on the demand index of the target energy in each quarter, and the i-th state S of the target energy in the t-th quarter is then defined. i Transition to the j-th state S in the (t+1)-th quarter j The number of times z appears ij The statistics are shown in Table 1:

[0161] Table 1. Statistics on the number of state transitions

[0162]

[0163] Then, according to the formula:

[0164]

[0165] The state transition probability P of the state S of the target energy in the t+1 quarter i The state transition probability P of the state S of the target energy in the t+1 quarter j The state transition probability P of the state S of the target energy in the t+1 quarter ij The state transition probability matrix P of the target energy in n quarters is established, specifically:

[0166]

[0167] The state transition probability P of the state S of the target energy in the t+1 quarter

[0168]

[0169] Therefore, the probability of the energy demand index being in a normal state in this embodiment is 90.14%.

[0170] In this specification, the prediction of the most likely state of the energy demand index in the next quarter by the Markov chain model is a quantitative analysis, and the prediction of the trend of the index change on the trend of the energy demand index is a qualitative analysis. Through the combination of quantitative analysis and qualitative analysis, the subjectivity of the prediction is effectively reduced, and the prediction result is more objective and scientific.

[0171] In another specific example, when predicting the demand index state of the target energy in the future quarter, a long short-term memory neural network model is introduced to effectively capture the long short-term dependence relationship in the time series data. Through the learning of the historical data of the demand index of the target energy in each quarter, the future change of the energy demand index is predicted. Unlike traditional methods, the long short-term memory neural network model can continuously update the model weight, iteratively train, automatically extract effective features from the data, and reduce the dependence on artificial intervention. At the same time, the long short-term memory neural network can effectively capture long-term trends and complex fluctuations, better cope with nonlinear relationships and sudden changes in the periodic changes of the target energy.

[0172] In addition, the specification also provides a highly systematic and automated demand forecasting platform through the technologies of Python programming language, Tkinter interface development and MySQL database. Specifically, (1) Data processing automation: using Python data processing libraries (such as Pandas, NumPy, etc.) to clean, interpolate and preprocess different frequency target energy demand data. The automated data processing process not only improves work efficiency, but also effectively avoids human errors and calculation errors. (2) Graphical user interface (GUI) design: an interactive graphical user interface is built using Tkinter, which allows users to input data, control the calculation process and display results through a visual interface, thereby reducing the use threshold and improving the convenience of operation. (3) Database management: through the MySQL database management system, various energy index data, calculation results and historical records are stored and managed. MySQL database ensures long-term data storage, data consistency and query efficiency, supporting large-scale data processing and multiple query needs. (4) Model automatic execution: the system can automatically execute demand index calculation based on the MF-MS-DFM-CI model, update index results, and perform trend prediction through the long short-term memory neural network model. All calculation processes are automatically completed by the system, improving efficiency and accuracy. Through the mixed frequency Markov dynamic factor demand analysis model (MF-MS-DFM-CI) based on the composite index, the key problems in the prior art are solved, and the prediction ability of energy demand is significantly improved through innovative model design and automated system implementation.

[0173] As shown in Figure 4 The specification also provides an energy demand analysis system, comprising:

[0174] An index screening and grouping unit 401 is configured to screen and group the initial indexes of the target energy demand, and generate a plurality of index groups of the target energy demand; wherein each index group contains a plurality of demand evaluation indexes;

[0175] A factor extraction and attribute determination unit 402 is configured to perform a mixed frequency operation on the data of different frequencies in the pre-acquired target energy demand data, and based on a mixed frequency dynamic factor model of Markov switching mechanism, extract the common factor of each quarter of a plurality of demand evaluation indexes in each index group generated by the mixed frequency operation, and determine the energy demand attribute corresponding to the common factor of each quarter in each index group.

[0176] The demand index determination unit 403 is configured to calculate the demand index of each target energy in each quarter according to the mixed data and the common factor of each demand evaluation index in each index group and the corresponding energy demand attribute in each quarter according to the pre-constructed Markov chain-based synthetic index model.

[0177] The energy demand analysis system provided in the specification can realize the steps and processes of the energy demand analysis method of any of the above embodiments and achieve the same technical effects. Here, it will not be described one by one.

[0178] In the description of the present application, the terms "first", "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise explicitly specified.

[0179] In the present application, the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" and the like mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples.

[0180] The above is only the preferred embodiment of the present application, and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of analyzing energy demand, characterized by, Comprise: Step S101, screening and grouping the initial indicators of the predetermined target energy demand, generating multiple indicator groups of target energy demand; wherein each of the indicator groups contains multiple demand evaluation indicators; Step S102, mixing the data of different frequencies in the pre-acquired target energy demand data, and based on the mixing dynamic factor model of Markov switching mechanism, extracting the common factors of each of the demand evaluation indicators in each of the indicator groups every quarter respectively, and determining the energy demand attributes corresponding to the common factors of each quarter in each of the indicator groups; wherein the mixing dynamic factor model comprises: a single factor observation equation, a factor dynamic equation and a state space equation; The single factor observation equation is: In the formula, is the mixed frequency data of the monthly frequency in the first quarter, is the mixed frequency data of the quarterly frequency in the first quarter; is an intercept term, is a factor loading coefficient; is the common factor of a plurality of demand evaluation indexes of the first index group in the first quarter; is the data error of the mixed frequency data of the first index group in the first quarter time period; are positive integers, and ; The factor dynamic equation is: In the formula, For lag operators, characterizing the first... Multiple demand evaluation indicators of the aforementioned indicator group in the first Common factors of each quarter For the Multiple demand evaluation indicators of the aforementioned indicator group in the first Common factors of each quarter The lagged effects; For the first Multiple demand evaluation indicators of the aforementioned indicator group in the first The Markov state variables of the first quarter represent the first quarter. The target energy in the aforementioned indicator group is from the first Demand in the first quarter to The probability of demand shift in the demand volume for each quarter; For the first The first of the aforementioned indicator groups The random error of the mixing data for each quarter follows a mean of 0 and a variance of . The normal distribution; The state space equation is: In the formula, For the first The mixing data within the specified quarter, including the first quarter... Mixing data for monthly frequencies within a quarter and the Mixing data of quarterly frequencies within a quarter ; A vector of constant terms. For the load matrix, For the first The first of the aforementioned indicator groups The state variable matrix for each quarter It is the intercept vector; For the first The first of the aforementioned indicator groups State variable matrix for each quarter For the The first of the aforementioned indicator groups State variable matrix for each quarter The coefficient matrix of the lag effect; For the first The first of the aforementioned indicator groups Error matrix of the mixing data for each quarter; Step S103, according to the mixing data and the common factors of each of the demand evaluation indicators in each of the indicator groups every quarter and the corresponding energy demand attributes, calculating the demand index of target energy every quarter according to the synthetic index model based on Markov chain constructed in advance.

2. The energy demand amount analysis method according to claim 1, characterized by, In step S101, Based on the time difference correlation coefficient analysis method, the multiple initial indicators of the predetermined target energy demand are screened and grouped to generate the leading indicator group, the consistency indicator group and the lagging indicator group of the target energy demand.

3. The energy demand amount analysis method according to claim 1, characterized by, Step S103 includes: calculating the symmetric change rate of the jth demand evaluation index in the ith index group in the ith quarter ;​​​​​​​​​​​​​​​ According to the standardized rate of change Calculate the first In the aforementioned indicator groups The aforementioned demand evaluation indicators in the first Average rate of change for each quarter To determine the first The aforementioned indicator groups in Standardization factor for each quarter ;in, It is a positive integer; based on the standardization factor , the average change rate is corrected to obtain the first standardization average change rate of the demand evaluation index in the first quarter of the first quarter of the first quarter ; According to the average growth rate of the initial composite index of the plurality of predetermined index groups and the target demand trend of the predetermined target energy , the first standardized average change rate is trend-adjusted to obtain the second standardized change rate of the th demand evaluation index in the th quarter of the th index group ; According to the second standardized rate of change Calculate the first The first of the aforementioned indicator groups The composite index after trend adjustment for the quarter and based on the first The aforementioned indicator groups in The quarterly trend-adjusted average composite index and the predetermined first The benchmark index of the aforementioned indicator group Calculate the target energy level The first of the aforementioned indicator groups Demand index corresponding to each quarter .

4. The energy demand amount analysis method according to claim 3, characterized by, In step S103, according to the mixing data, according to the formula: Calculate the first The first of the aforementioned indicator groups The demand evaluation index in the first The symmetrical rate of change in each quarter In the formula, For the first The first of the aforementioned indicator groups The demand evaluation index in the first A data matrix of mixed frequency data for each quarter; For the first The first of the aforementioned indicator groups The demand evaluation index in the first A data matrix of mixed frequency data for each quarter; For the first Multiple demand evaluation indicators within the aforementioned indicator group are in the first... Common factors of each quarter Adjustment parameters; All are positive integers, and ; According to the formula: determining the standardized factor of the first demand evaluation index in the first index group in the first quarter ; wherein, ; wherein, ; wherein, ; wherein, is a positive integer. According to the formula: determining the standardized change rate of the first demand evaluation index in the first quarter of the first index group of the first index group of the first index group .

5. The energy demand amount analysis method according to claim 3, characterized by, In step S103, according to the formula: Determine the first In the aforementioned indicator groups The aforementioned demand evaluation indicators in the first Average rate of change for each quarter In the formula, For the first The first of the aforementioned indicator groups The demand evaluation index in the first Standardized rate of change for each quarter The weights; where, All are positive integers, and ; According to the formula: determining a standardized factor of the first index group in the first quarter is a data error of the mixed data of the first index group in the first quarter; is a Markov state variable of the plurality of demand evaluation indexes of the first index group in the first quarter, representing a demand transition probability of a target energy in the first index group from a demand amount in the first quarter to a demand amount in the second quarter; is an average change rate of the first demand evaluation index in the first quarter; and is a positive integer;​​​​​​​​​​​​ According to the formula: The first standardized average change rate of the demand evaluation index in the first quarter of the first index group is obtained by correcting the average change rate of the demand evaluation index in the first quarter of the first index group. ​​​​​​​​ 6. The energy demand amount analysis method according to claim 3, characterized by, The indicator group includes: the leading indicator group, the consistency indicator group and the lagging indicator group; calculating an average growth rate for each index sequence in the set of consistency indices via a compound interest formula to determine a target demand trend for the target energy source ; After determining the initial composite indices of the preceding indicator group, the consistency indicator group and the lagging indicator group respectively based on the exponential composite method, the average growth rate of the initial composite indices of the preceding indicator group, the consistency indicator group and the lagging indicator group is calculated through the compound interest formula ; Wherein, according to the formula: Determine based on the first Multiple demand evaluation indicators within the aforementioned indicator group Markov state variables for each quarter The adjustment of the first The first of the aforementioned indicator groups The initial composite index for the quarter In the formula, For the first Multiple demand evaluation indicators within the aforementioned indicator group Markov state variables for each quarter The adjustment of the first The indicator group mentioned in the first The initial composite index for each quarter; For the first In the aforementioned indicator groups The aforementioned demand evaluation indicators in the first The first standardized average rate of change for the quarter ; All are positive integers, and .

7. The energy demand amount analysis method according to claim 3, characterized by, In step S103, according to the formula: For the In the aforementioned indicator groups The first of the aforementioned demand evaluation indicators The first standardized average rate of change for the quarter After trend adjustment, the second standardized rate of change is obtained. In the formula, For the predetermined target energy demand trend, The average growth rate of the initial composite index of the multiple index groups; According to the formula: computing target energy one of the index groups corresponding to the demand index of the first quarter ; in the formula, is the trend-adjusted composite index of the first quarter of the one of the index groups; is the trend-adjusted composite index of the first quarter of the one of the index groups; is the trend-adjusted average composite index of the first quarter of the one of the index groups; is the predetermined reference index of the first one of the index groups; wherein, are all positive integers, and .​​​ 8. The energy demand amount analysis method according to claim 1, characterized by, Also includes: According to the demand index of the target energy source in each quarter, a demand index dynamic of the target energy source in each quarter is determined based on a preset index dynamic interval, to generate a demand transfer probability matrix of the target energy source in each quarter. According to the demand index dynamics of the target energy in the current quarter, determine the demand dynamic transfer vector of the target energy in the current quarter; According to the demand transfer probability matrix and the demand dynamic transfer vector of the target energy in the current quarter, calculate the demand transfer probability vector of the target energy in the next quarter, and select the index dynamic interval of the maximum element value in the demand transfer probability vector as the demand index dynamic of the target energy in the next quarter.

9. An energy demand quantity analysis system characterized by comprising: Comprise: The index screening and grouping unit is configured to screen and group the initial indicators of the predetermined target energy demand, generating multiple indicator groups of target energy demand; wherein each of the indicator groups contains multiple demand evaluation indicators; The factor extraction and attribute determination unit is configured to mix the data of different frequencies in the pre-acquired target energy demand data, and based on the mixing dynamic factor model of Markov switching mechanism, extracting the common factors of each of the demand evaluation indicators in each of the indicator groups every quarter respectively, and determining the energy demand attributes corresponding to the common factors of each quarter in each of the indicator groups; wherein the mixing dynamic factor model comprises: a single factor observation equation, a factor dynamic equation and a state space equation; The single factor observation equation is: In the formula, is the mixed frequency data of the monthly frequency in the first quarter, is the mixed frequency data of the quarterly frequency in the first quarter; is an intercept term, is a factor loading coefficient; is the common factor of a plurality of demand evaluation indexes of the first index group in the first quarter; is the data error of the mixed frequency data of the first index group in the first quarter time period; are all positive integers, and ; The factor dynamic equation is: In the formula, For lag operators, characterizing the first... Multiple demand evaluation indicators of the aforementioned indicator group in the first Common factors of each quarter For the Multiple demand evaluation indicators of the aforementioned indicator group in the first Common factors of each quarter The lagged effects; For the first Multiple demand evaluation indicators of the aforementioned indicator group in the first The Markov state variables of the first quarter represent the first quarter. The target energy in the aforementioned indicator group is from the first Demand in the first quarter to The probability of demand shift in the demand volume for each quarter; For the first The first of the aforementioned indicator groups The random error of the mixing data for each quarter follows a mean of 0 and a variance of . The normal distribution; The state space equation is: In the formula, For the first The mixing data within the specified quarter, including the first quarter... Mixing data for monthly frequencies within a quarter and the Mixing data of quarterly frequencies within a quarter ; A vector of constant terms. For the load matrix, For the first The first of the aforementioned indicator groups The state variable matrix for each quarter It is the intercept vector; For the first The first of the aforementioned indicator groups State variable matrix for each quarter For the The first of the aforementioned indicator groups State variable matrix for each quarter The coefficient matrix of the lag effect; For the first The first of the aforementioned indicator groups Error matrix of the mixing data for each quarter; The demand index determination unit is configured to calculate the demand index of the target energy for each quarter according to the mixed data and the common factor of each demand evaluation index in each index group and the corresponding energy demand attribute for each quarter according to a Markov chain-based synthetic index model constructed in advance.

Images