Demand side management method based on 5G smart power grid daily load feature clustering analysis

Abstract

The invention relates to a demand side management method based on 5G smart power grid daily load characteristic clustering analysis, which adopts a K-means clustering analysis method to analyze user historical load data, classifies user loads with different characteristics, obtains different power utilization habits and characteristics of different types of users, and excavates demand side response potentials of different users. A demand side management model is constructed based on user load characteristic clustering analysis, and users with different load characteristics are considered to select different time-of-use electricity price strategies to improve user satisfaction; when bidirectional interaction is carried out between the smart grid and a demand side, a smart grid and demand side interaction management model based on the game theory is adopted to realize benefit consideration of the smart grid and the demand side; aiming at the influence of inaccurate demand baseline prediction on interaction, an enhanced interaction strategy is adopted to deal with uncertainty; the method has the advantages of establishing the demand management model based on the load characteristics, optimizing the resource configuration of the power grid, maximizing the net profit of the power grid, reducing the demand fluctuation and improving the stability.

Description

Technical Field

[0001] The present invention belongs to the technical field of smart grids, and specifically relates to a demand side management method based on cluster analysis of daily load characteristics of a 5G smart grid. Background Art

[0002] Since the urbanization of my country, the urban infrastructure and people's living standards have been continuously improved, and the electricity load has been growing. In addition, due to the regularity of the residential load, this type of load usually occurs during the peak period of system electricity consumption, which will increase the peak-to-valley difference of the power grid load. The traditional solution is to increase the installed capacity of the system power supply to meet the peak load demand. However, this solution can solve the problem of system load during peak periods, but the utilization rate is limited at other times, especially during the load valley period, which causes resource waste. In addition, the frequent start and stop of the unit will also increase the coal consumption, increase the operating cost of the system, and waste social resources. The number of residential users and industrial and commercial users is large and the distribution is relatively scattered, and the electricity load is relatively concentrated, mainly concentrated in the peak load period. Without affecting commercial operations and life, the load reliability requirements are not high, and there is a great demand-side response potential. Demand response (Demand Dynamic Response (DR) management is a key feature of smart grids. Its characteristic is that end users can adjust their electricity consumption patterns at any time according to changes in electricity prices, reduce electricity costs, and improve the reliability and safety of real-time operation of the power grid. With the development of the times, the proportion of electricity consumption of large buildings in the total national electricity consumption is increasing. To reduce the energy consumption of buildings, it is necessary to actively change their load distribution during peak hours through DR management; however, the load transfer of buildings will lead to changes in the total electricity consumption on the demand side, which in turn affects the operation status and dynamic electricity prices of smart grids. Therefore, how to maximize the benefits of both smart grids and demand sides in the process of interaction has been a problem that has attracted much attention in recent years; therefore, it is very necessary to provide a demand side management method based on 5G smart grid daily load feature clustering analysis, which can establish a demand management model based on load characteristics, optimize the resource allocation of the power grid, maximize the net profit of the power grid, reduce demand fluctuations and improve stability. Summary of the invention

[0003] 1. Technical issues

[0004] In view of the above-mentioned existing technical status, this application mainly addresses the following technical problems:

[0005] 1. How to solve the problem that the traditional method of increasing the installed capacity of the system power supply to meet the peak load demand has limited utilization, wastes resources, increases the operating cost of the system, and wastes social resources;

[0006] 2. How to maximize the benefits of both smart grid and demand side during the interaction process.

[0007] (II) Technical solution

[0008] The purpose of the present invention is to overcome the shortcomings of the prior art and to provide a demand-side management method based on 5G smart grid daily load characteristic clustering analysis, which establishes a demand management model based on load characteristics, optimizes the resource allocation of the power grid, maximizes the net profit of the power grid, reduces demand fluctuations and improves stability.

[0009] The object of the present invention is achieved by: a demand side management method based on 5G smart grid daily load characteristics clustering analysis, the method comprising the following steps:

[0010] Step 1: Use the K-means clustering analysis method to analyze the user's historical load data, classify the user loads with different characteristics, obtain the different electricity consumption habits and characteristics of different types of users, and explore the demand-side response potential of different users;

[0011] Step 2: Based on the cluster analysis of user load characteristics mined in the previous step, a demand-side management model is constructed, and different time-of-use electricity price strategies are selected to improve user satisfaction by considering users with different load characteristics;

[0012] Step 3: When the smart grid and the demand side have two-way interactions, a smart grid and demand side interaction management model based on game theory is adopted to achieve a balanced benefit of the smart grid and the demand side;

[0013] Step 4: In view of the impact of inaccurate demand baseline forecast on interaction, adopt an enhanced interaction strategy to deal with uncertainty.

[0014] Furthermore, in step 1, cluster analysis is performed on the user's historical load data, specifically: data collection is performed at intervals of every hour in a day to obtain the daily load data of each user as each sample point of a data sample set, generate raw data, and preprocess the raw data before performing cluster analysis.

[0015] Furthermore, the demand side management model in step 2 includes the following steps:

[0016] Step 2.1: Model for selecting residential time-of-use electricity prices in demand-side management;

[0017] Step 2.2: Residents’ response process to time-of-use electricity prices.

[0018] Furthermore, the residential time-of-use electricity price selection model in the demand side management in step 2.1 includes the following steps:

[0019] Step 2.11: Division of load peak and valley periods: Select the highest peak point and the lowest valley point in the load as the reference points. The probability of the peak point belonging to the peak period is 100%, and the probability of belonging to the valley period is 0%. On the contrary, the probability of the valley point belonging to the valley period is 100%, and the distance to the peak period is 0%. The peak and valley periods of other points are determined by the direct distance from the peak point and the valley point. The specific formula is as follows: In the formula, the load value of each point every day is b and a are the maximum and minimum values ​​of daily load, respectively; are the membership values ​​of the peak value and valley value of the load in the i-th period respectively;

[0020] Step 2.12: Use the peak-valley price ratio to determine the time-of-use electricity price: Assume that the total load curve of all users in a distribution network has been divided into three periods: peak, flat and valley, corresponding to load q p ,q f ,q v , the electricity price is expressed as p p 、p f 、p v , then its peak-valley electricity price can be expressed as: p p =(1+α)p f , p v =(1-β)p f , where α and β are the price increase coefficients during peak hours and price decrease coefficients during valley hours respectively; the average electricity price after price adjustment can be expressed as: Simplifying, we can get: In the above formula, That is, the average electricity price is equal to the normal period electricity price, and we can get αq p -βq v =0, the peak-to-valley price ratio coefficient can be expressed as: The ratio of electricity consumption during peak and valley periods is used as the peak-valley price ratio coefficient. After knowing the electricity consumption during peak and valley periods, the electricity price during peak and valley periods can be determined.

[0021] Furthermore, the process of residents responding to the time-of-use electricity price in step 2.2 includes the following steps:

[0022] Step 2.21: Price elasticity coefficient: The price elasticity coefficient can be expressed as: In the formula, ρ represents the price demand elasticity coefficient; Δp and Δq are the increment of electricity price and the increment of user load respectively. Since the electricity demand of users is not independent, the cross elasticity coefficient is introduced: In the formula, i and j represent different time periods respectively; the user's demand elasticity matrix is ​​formed by combining the self-elasticity coefficient and the cross-elasticity coefficient to express the comprehensive impact of electricity prices on user electricity consumption. First, the elasticity coefficient matrix can be expressed as: The electricity price elasticity coefficient can be expressed in the form of a matrix:

[0023]

[0024] Step 2.22: Residents’ satisfaction with time-of-use electricity prices: The expression of satisfaction is: U = U0-U cost θ1-U behavior θ2, where U0 is the user satisfaction without using time-of-use electricity price; U cost , θ1 are the satisfaction with electricity cost change caused by electricity price and the coefficient of satisfaction with cost change; U behavior , θ2 are the satisfaction of users’ electricity consumption behavior changes caused by electricity prices and their satisfaction coefficients; among them, the satisfaction of cost changes can be expressed as: U cost =P-P0, where P is the daily electricity cost when the time-of-use electricity price is implemented; P0 is the user's electricity cost when the time-of-use electricity price is not implemented; the satisfaction with the change in electricity consumption behavior can be expressed as: Where R is the influence coefficient of time-of-use unit price on the change of user's satisfaction with electricity consumption behavior; Q h , Q h0 They are respectively the electricity consumption of the user in the hth period under the time-of-use electricity price and the electricity consumption of the user in the hth period under the time-of-use electricity price.

[0025] Furthermore, the interaction management model in step 3 includes the following steps:

[0026] Step 3.1: Grid-building interaction model;

[0027] Step 3.2: Grid-Building Interaction Strategy.

[0028] Furthermore, the grid-building interaction model in step 3.1 comprises the following steps:

[0029] Step 3.11: Grid benefit model: The grid profit refers to the net profit obtained from selling electricity to buildings after deducting the corresponding electricity purchase costs and network losses. In time period i, the grid net profit is determined by the electricity sales and the corresponding power generation costs, and its calculation formula is as follows: Among them, P i is the net profit of the power grid in period i; is the total electricity purchased by the building in period i, i.e., the electricity sold by the grid; for Function of power generation cost; α, β, γ are constant coefficients; f(D f ) is the dynamic electricity price in period i, which is the fluctuation of electricity demand D f is a function of the power demand The average value of the entire discussion range The deviation between them is calculated as:

[0030] Step 3.12: Building benefit model: User dissatisfaction refers to the user's dissatisfaction when electricity demand is restricted and deviates from the normal demand level. The calculation formula is as follows: in for user dissatisfaction; is the power demand of building i in time period k; is the original power demand of building k in period i; k , k is a positive constant, indicating the preference of building i to change its electricity demand;

[0031] The transfer cost refers to the loss or cost caused by demand transfer under the economic incentive of the smart grid. The transfer cost of building k The actual power demand of building i in period k and baseline It is related to the deviation between them, that is, it is related to user dissatisfaction, and its calculation formula is as follows: Where ρ i is a preset positive factor describing the cost level of demand transfer for building i;

[0032] The electricity fee is the fee paid by the building when it purchases a certain amount of electricity, minus the corresponding transfer costs. The calculation formula is as follows: Where F i k For electricity charges.

[0033] Furthermore, the grid-building interaction strategy in step 3.2 is an interaction strategy based on the Stackelberg game, specifically: the Nash equilibrium of the preset Stackelberg game is the optimal strategy combination of the grid and the building: In the formula The optimal dynamic electricity price for the day; For building electricity needs; The optimal power demand of the building; g ,I b is the benefit function of the grid and buildings;

[0034] The Stackelberg game is solved by reverse induction to determine the optimal dynamic electricity price of the power grid and the optimal power demand of each building: First, I b Maximize, and get according to the electricity price given by the power grid Then, determine Substitute I g The optimal dynamic electricity price is obtained, that is, The superscript “-1” in the formula indicates the inverse function.

[0035] Furthermore, the grid-building interaction strategy in step 3.2 specifically includes the following processes:

[0036] Step 3.21: The power grid publishes the day-ahead dynamic electricity price;

[0037] Step 3.22: Building demand management selects the optimal power demand from its own strategy set according to its own benefit function And feed back to the grid side;

[0038] Step 3.23: The grid determines the optimal electricity price based on the optimal response strategy fed back by the building and its own benefit function.

[0039] Step 3.24: Repeat steps 3.21 to 3.23 until Nash equilibrium is reached.

[0040] Furthermore, the enhanced interactive strategy in step 4 considering the inaccurate baseline forecast of demand is specifically as follows: assuming that the baseline forecast uncertainty of building power demand follows a normal distribution: The expected value of the baseline prediction uncertainty is used as a correction term to improve the robustness of the proposed interaction strategy: By modifying the Stackelberg game solution, the optimal power demand of the building and the optimal dynamic price of the grid determined by the enhanced interactive strategy are: Among them, rev represents the revised situation.

[0041] (III) Beneficial effects

[0042] 1. The present invention applies the results of cluster analysis mining to the demand side management of smart grids, and establishes a demand side management model that takes into account user satisfaction. The model uses the load characteristic curve to determine the peak and valley periods and the time-of-use electricity price. The electricity price is adjusted through the user's response process to the time-of-use electricity price, thereby improving the user's satisfaction with the time-of-use electricity price, increasing the user's enthusiasm for participating in the demand side response, and improving the level of demand side management of smart grids;

[0043] 2. Aiming at the problem that it is difficult to take into account both the benefits of the smart grid and the demand side when the smart grid and the demand side interact in a two-way manner, the present invention proposes a smart grid and demand side interaction management strategy based on game theory. The strategy uses the Stackelberg game to ensure that the confrontational strategies between the grid and the demand side are always in a Nash equilibrium state, thereby ensuring the maximization of the interests of both parties. Aiming at the impact of inaccurate demand baseline forecasts on the interaction strategy, an enhanced interaction strategy is proposed to deal with uncertainty. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a structural diagram of the interaction model between the smart grid and the responsive building of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described below in conjunction with embodiments and / or drawings.

[0046] Example 1

[0047] like Figure 1 As shown, a demand side management method based on cluster analysis of daily load characteristics of 5G smart grid includes the following steps:

[0048] Step 1: Use the K-means clustering analysis method to analyze the user's historical load data, classify the user loads with different characteristics, obtain the different electricity consumption habits and characteristics of different types of users, and explore the demand-side response potential of different users;

[0049] In the present invention, cluster analysis is performed on the user's historical load data, specifically: data collection is performed at intervals of every hour in a day to obtain the daily load data of each user as each sample point of a data sample set, generate raw data, and before performing cluster analysis, the raw data needs to be preprocessed, such as missing value processing and outlier processing.

[0050] Step 2: Based on the cluster analysis of user load characteristics mined in the previous step, a demand-side management model is constructed, and different time-of-use electricity price strategies are selected to improve user satisfaction by considering users with different load characteristics;

[0051] In the present invention, the demand side management model includes the following steps:

[0052] Step 2.1: Model for selecting residential time-of-use electricity prices in demand-side management:

[0053] 2.11: Division of load peak and valley periods: The highest peak point and the lowest valley point in the load are selected as reference points. The probability of the peak point belonging to the peak period is 100%, and the probability of belonging to the valley period is 0%. On the contrary, the probability of the valley point belonging to the valley period is 100%, and the distance to the peak period is 0%. The peak and valley periods of other points are determined by the direct distance from the peak point and the valley point. The specific formula is as follows: In the formula, the load value of each point every day is b and a are the maximum and minimum values ​​of daily load, respectively; are the membership values ​​of the peak value and valley value of the load in the i-th period respectively; there are three selection strategies for the peak value membership and valley value membership values:

[0054] ①When selecting That is, the daily load point is either a peak value or a valley value, and there are only two periods of peak and valley;

[0055] ②When selecting When the membership of the peak value and the valley value is the same, which is between 50% and 100%, then the load points in each period of the day with load values ​​between this membership can be regarded as peak periods or valley periods, and the periods not in between are regarded as normal periods.

[0056] Since the first method of value selection has a large load span when dividing the time periods, it will increase the difficulty of user response and cause the efficiency of peak shaving and valley filling to decrease. In practical applications, the second method is generally selected to divide the daily time period into three periods: peak, flat and valley.

[0057] Step 2.12: Use the peak-valley price ratio to determine the time-of-use electricity price: Assume that the total load curve of all users in a distribution network has been divided into three periods: peak, flat and valley, corresponding to load q p ,q f ,q v , the electricity price is expressed as p p 、p f 、p v , then its peak-valley electricity price can be expressed as: p p =(1+α)p f , p v =(1-β)p f , where α and β are the price increase coefficients during peak hours and price decrease coefficients during valley hours respectively; the average electricity price after price adjustment can be expressed as: Simplifying, we can get: In the above formula, That is, the average electricity price is equal to the normal period electricity price, and we can get αq p -βq v =0, the peak-to-valley price ratio coefficient can be expressed as: The ratio of electricity consumption during peak and valley periods is used as the peak-valley price ratio coefficient. After knowing the electricity consumption during peak and valley periods, the electricity price during peak and valley periods can be determined.

[0058] The peak-to-valley price ratio is an important indicator for judging the effect of time-of-use electricity prices on peak shaving and valley filling on the demand side. Generally speaking, the larger the price ratio, the better the peak-shaving effect. Therefore, when determining the time-of-use electricity price, the price ratio should be increased as much as possible to increase the enthusiasm of users on the demand side to respond.

[0059] Step 2.2: Residents’ response process to time-of-use electricity prices:

[0060] 2.21: Price demand elasticity coefficient: The specific meaning of the price demand elasticity coefficient is the change in user electricity consumption caused by a percentage change in electricity price. The electricity price elasticity coefficient can be expressed as: In the formula, ρ represents the price demand elasticity coefficient; Δp and Δq are the increment of electricity price and user load respectively. However, due to the fixed characteristics of electricity demand, users still need to consider the valley period electricity price and the flat period electricity price during the peak period, so as to arrange their electricity consumption behavior to meet the daily electricity demand. That is, the electricity demand of users is not independent, so the cross elasticity coefficient is introduced: In the formula, i and j represent different time periods. When i=j, the elasticity coefficient is the self-elasticity coefficient, that is, the change in electricity demand in the current period caused by the change in electricity price in the current period; the response mechanism generated by the time-of-use electricity price enables users to respond to the high and low electricity prices in different periods, so that the user load shifts from peak to valley. The response of residential users to time-of-use electricity prices is not independent, but is affected by the comprehensive influence of electricity prices in multiple periods. Therefore, the user's demand elasticity matrix should be formed by combining the self-elasticity coefficient and the cross-elasticity coefficient to express the comprehensive impact of electricity prices on user electricity consumption. First, the elasticity coefficient matrix can be expressed as: The electricity price elasticity coefficient can be expressed in the form of a matrix: According to the above formula, by knowing the current electricity price and the change in electricity price, we can get the ratio of the change in electricity consumption in each period to the electricity consumption. Combined with the electricity consumption, we can get the change in electricity consumption. The change in electricity consumption in each period is the user's response to the change in electricity price;

[0061] 2.22: Residents’ satisfaction with time-of-use electricity prices: Satisfaction is expressed as: U = U0-U cost θ1-U behavior θ2, where U0 is the user satisfaction without using time-of-use electricity price; U cost , θ1 are the satisfaction with electricity cost change caused by electricity price and the coefficient of satisfaction with cost change; U behavior , θ2 are the satisfaction and satisfaction coefficient of the change of user electricity consumption behavior caused by electricity price; the satisfaction of cost change can be expressed as: U cost =P-P0, where P is the daily electricity cost when the time-of-use electricity price is implemented; P0 is the user's electricity cost when the time-of-use electricity price is not implemented; the satisfaction with the change in electricity consumption behavior can be expressed as: Where R is the influence coefficient of time-of-use unit price on the change of user's electricity consumption behavior, and is related to the peak electricity price p max And the off-peak electricity price p min Related; Q h , Q h0 They are respectively the electricity consumption of the user in the hth period under the time-of-use electricity price and the electricity consumption of the user in the hth period under the time-of-use electricity price.

[0062] Step 3: When the smart grid and the demand side have two-way interactions, a smart grid and demand side interaction management model based on game theory is adopted to achieve a balanced benefit of the smart grid and the demand side;

[0063] Step 4: In view of the impact of inaccurate demand baseline forecast on interaction, adopt an enhanced interaction strategy to deal with uncertainty.

[0064] The present invention applies the results of cluster analysis mining to the demand side management of the smart grid, and establishes a demand side management model that takes user satisfaction into consideration. The model uses the load characteristic curve to determine the peak and valley time periods and the time-of-use electricity price, and adjusts the electricity price through the user's response process to the time-of-use electricity price, thereby improving the user's satisfaction with the time-of-use electricity price, increasing the user's enthusiasm for participating in the demand side response, and improving the level of demand side management of the smart grid. The present invention has the advantages of establishing a demand management model based on load characteristics, optimizing the resource allocation of the power grid, maximizing the net profit of the power grid, reducing demand fluctuations, and improving stability.

[0065] Example 2

[0066] like Figure 1 As shown, a demand side management method based on cluster analysis of daily load characteristics of 5G smart grid includes the following steps:

[0067] Step 1: Use the K-means clustering analysis method to analyze the user's historical load data, classify the user loads with different characteristics, obtain the different electricity consumption habits and characteristics of different types of users, and explore the demand-side response potential of different users;

[0068] Step 2: Based on the cluster analysis of user load characteristics mined in the previous step, a demand-side management model is constructed, and different time-of-use electricity price strategies are selected to improve user satisfaction by considering users with different load characteristics;

[0069] Step 3: When the smart grid and the demand side have two-way interactions, a smart grid and demand side interaction management model based on game theory is adopted to achieve a balanced benefit of the smart grid and the demand side;

[0070] In the present invention, the interactive management model includes the following steps:

[0071] Step 3.1: Grid-building interaction model;

[0072] 3.11: Grid benefit model: In the grid-building interaction model, the grid aims to maximize profits and reduce demand fluctuations by setting electricity prices (i.e., price-based demand response programs); the grid profit refers to the net profit obtained by selling electricity to buildings after deducting the corresponding electricity purchase costs and network losses; in time period i, the grid's net profit is determined by the electricity sales and the corresponding generation costs, and its calculation formula is as follows: Among them, P i is the net profit of the power grid in period i; is the total electricity purchased by the building in period i, i.e., the electricity sold by the grid; for Function of power generation cost; α, β, γ are constant coefficients; f(D f ) is the dynamic electricity price in period i, which is the fluctuation of electricity demand D f is a function of the power demand The average value of the entire discussion range The deviation between them is calculated as: In the formula, N is set to 24, that is, a day is divided into 24 time periods, and the interaction strategy is executed once every hour;

[0073] 3.12: Building benefit model: In the grid-building interaction model, various buildings are designed to minimize electricity costs and transfer costs by responding to electricity price information and transferring demand, thereby reducing user dissatisfaction; user dissatisfaction refers to the user's dissatisfaction when electricity demand is restricted and deviates from the normal demand level. The calculation formula is as follows: in for user dissatisfaction; is the power demand of building i in time period k; is the original power demand of building k in period i (i.e., baseline); k , k is a positive constant, indicating the preference of building i to change its electricity demand; when the electricity demand is close to the baseline, the user's dissatisfaction will decrease, and when the electricity demand is equal to the baseline, the user's dissatisfaction will be minimized;

[0074] The transfer cost refers to the loss or cost caused by the demand transfer under the economic incentive of the smart grid. Usually, the transfer cost of building k is The actual power demand of building i in period k and baseline It is related to the deviation between them, that is, it is related to user dissatisfaction, and its calculation formula is as follows: Where ρ i is a preset positive factor describing the cost level of demand shifting for building i; generally, when electricity demand shifting increases, the shifting cost of the building will also increase accordingly;

[0075] The electricity fee is the fee paid by the building when it purchases a certain amount of electricity, minus the corresponding transfer costs. The calculation formula is as follows: Where F i k For electricity charges.

[0076] Step 3.2: Grid-Building Interaction Strategy.

[0077] The grid-building interaction strategy is an interaction strategy based on the Stackelberg game. Specifically, the Nash equilibrium of the preset Stackelberg game is the optimal strategy combination of the grid and the building: In the formula The optimal dynamic electricity price for the day; For building electricity needs; The optimal power demand of the building; g ,I b is the benefit function of the grid and buildings;

[0078] The Stackelberg game is solved by reverse induction to determine the optimal dynamic electricity price of the power grid and the optimal power demand of each building: First, I b Maximize, and get according to the electricity price given by the power grid Then, determine Substitute I g The optimal dynamic electricity price is obtained, that is, The superscript “-1” in the formula indicates an inverse function.

[0079] The grid-building interaction strategy in step 3.2, based on the benefit set and strategy set of grid and building demand management, the operation process of the interaction management strategy is as follows:

[0080] Step 3.21: The power grid publishes the day-ahead dynamic electricity price;

[0081] Step 3.22: Building demand management selects the optimal power demand from its own strategy set according to its own benefit function And feed back to the grid side;

[0082] Step 3.23: The grid determines the optimal electricity price based on the optimal response strategy fed back by the building and its own benefit function.

[0083] Step 3.24: Repeat steps 3.21 to 3.23 until Nash equilibrium is reached.

[0084] The optimization strategies based on Nash equilibrium are as follows: ① Building demand management optimization based on Nash equilibrium: Faced with the electricity price information given by the power grid, each building hopes to reduce its benefit loss by adjusting the power demand at different times of the day. Therefore, the optimization problem of the building in the power grid-building interaction model is transformed into a minimization problem. The benefit loss of building electricity consumption mainly depends on three aspects, namely electricity charges, transfer costs and user dissatisfaction. The benefit loss of building k can be minimized by optimizing the power demand, and the total daily power demand remains unchanged, that is: In the formula is the power demand of building k at each time period in a day; Represents the electricity cost sensitivity coefficient of building i in time period k; introduces the Lagrange multiplier λ k To deal with the constraints in the above formula, the optimization problem of building i is as follows: Assuming that the electricity price f(D f ) It is known that in order to obtain the optimal power demand of building k in the demand management process, the first-order optimal condition for minimizing the building benefit loss is that the derivative is equal to 0, and its calculation formula is as follows: The solution of this equation is the equilibrium solution of Nash equilibrium; calculate The second-order derivative of is used to obtain the Hessian matrix to prove the existence of the Nash equilibrium of the proposed Stackelberg game strategy. The calculation formula is as follows: It can be seen from the above formula that the diagonal elements of the Hessian matrix are all positive values, and the non-diagonal factors are all zero. Therefore, the Hessian matrix is ​​a positive definite matrix, which means that when the optimal power demand of the building is in the Nash equilibrium state, it can effectively reduce its benefit loss in the process of grid-building interaction and minimize its benefit loss, that is, the existence and uniqueness of the building's Nash equilibrium are confirmed;

[0085] ② Optimization of dynamic pricing of power grid based on Nash equilibrium: In the process of grid-building interaction, the grid aims to reduce power demand fluctuations while maximizing net profit by optimizing dynamic electricity prices. Therefore, the calculation formula of grid benefits is as follows: in is the weight factor; are the minimum and maximum values ​​of the power generation of the power grid in time period k, respectively, and their values ​​are ±30% of the total power demand of the building in time period k;

[0086] According to the principle of reverse induction, the daily optimized dynamic electricity price of the power grid is determined according to the optimal electricity demand of each building; Substituting the optimal electricity demand of each building into the formula The total power demand of all buildings in each period can be obtained from the following formula. Substituting it into the benefit function of the power grid, it can be found that f(D f ) is the only variable that needs to be optimized, not f(D f )and Two variables; Finally, the optimization problem is transformed into a single-objective optimization problem, and the optimal solution to maximize the power grid benefit is Therefore, the existence of Nash equilibrium in the power grid is confirmed.

[0087] Step 4: In view of the impact of inaccurate demand baseline forecast on interaction, adopt an enhanced interaction strategy to deal with uncertainty.

[0088] In the present invention, the enhanced interactive strategy considering the inaccurate demand baseline prediction in step 4 is specifically as follows: in the grid-building interaction model, the original power demand of the building (i.e., the baseline) will affect the optimized dynamic electricity price and the building's power demand; however, in actual situations, the building's demand baseline will be predicted one day in advance, and due to various uncertainties or unexpected factors, this prediction is not completely accurate; the prediction error will cause the game solution to deviate from the Nash equilibrium, resulting in poor performance of the interactive strategy; in response to this problem, the present invention proposes an improved solution considering the uncertainty of the building demand baseline prediction, i.e., the enhanced grid-building interaction strategy; assuming that the uncertainty of the baseline prediction of the building's power demand follows a normal distribution: The expected value of the baseline prediction uncertainty is used as a correction term to improve the robustness of the proposed interaction strategy: It is worth noting that this correction does not affect the existence of Nash equilibrium; by correcting the Stackelberg game solution, the optimal power demand of the building and the optimal dynamic price of the grid determined by the enhanced interaction strategy are:

[0089]

[0090] Among them, rev represents the revised situation.

[0091] In the present invention, ① the basic interaction strategy based on game theory can effectively improve the efficiency of the power grid, reduce demand fluctuations and maintain the stability of the power grid by optimizing the dynamic electricity price of the power grid; at the same time, buildings transfer demand by responding to the optimal dynamic price, thereby reducing power loss and electricity costs; ② the enhanced interaction strategy based on game theory can effectively alleviate the negative impact of inaccurate demand baseline forecasts on the interaction strategy; in addition, the enhanced interaction strategy is better than the basic interaction strategy in improving the benefits of the power grid and buildings.

[0092] The present invention proposes a smart grid and building interaction management strategy based on game theory, establishes a Stackelberg game grid-building interaction model, determines the Nash equilibrium, and calculates the optimal dynamic electricity price of the grid and the optimal electricity demand of the building, so as to maximize the net profit of the grid, reduce demand fluctuations, and improve the stability of grid operation; in addition, the present invention also proposes an enhanced interaction strategy to solve the negative impact of inaccurate baseline forecasts of building demand on the strategy; the present invention has the advantages of establishing a demand management model based on load characteristics, optimizing the resource allocation of the grid, maximizing the net profit of the grid, reducing demand fluctuations and improving stability.

Claims

1. A demand side management method based on cluster analysis of daily load characteristics of 5G smart grid, characterized by: The method comprises the following steps: Step 1: Use the K-means clustering analysis method to analyze the user's historical load data, classify the user loads with different characteristics, obtain the different electricity consumption habits and characteristics of different types of users, and explore the demand-side response potential of different users; Step 2: Based on the cluster analysis of user load characteristics mined in the previous step, a demand-side management model is constructed, and different time-of-use electricity price strategies are selected to improve user satisfaction by considering users with different load characteristics; Step 3: When the smart grid and the demand side have two-way interactions, a smart grid and demand side interaction management model based on game theory is adopted to achieve a balanced benefit of the smart grid and the demand side; Step 4: In view of the impact of inaccurate demand baseline forecast on interaction, adopt an enhanced interaction strategy to deal with uncertainty.

2. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 1, characterized in that: In step 1, cluster analysis is performed on the user's historical load data. Specifically, data collection is performed at intervals of every hour in a day to obtain the daily load data of each user as each sample point of a data sample set, generate raw data, and preprocess the raw data before performing cluster analysis.

3. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 1, characterized in that: The demand side management model in step 2 includes the following steps: Step 2.1: Model for selecting residential time-of-use electricity prices in demand-side management; Step 2.2: Residents’ response process to time-of-use electricity prices.

4. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 3 is characterized by: The residential time-of-use electricity price selection model in the demand side management in step 2.1 includes the following steps: Step 2.11: Division of load peak and valley periods: Select the highest peak point and the lowest valley point in the load as the reference points. The probability of the peak point belonging to the peak period is 100%, and the probability of belonging to the valley period is 0%. On the contrary, the probability of the valley point belonging to the valley period is 100%, and the distance to the peak period is 0%. The peak and valley periods of other points are determined by the direct distance from the peak point and the valley point. The specific formula is as follows: In the formula, the load value of each point every day is b and a are the maximum and minimum values ​​of daily load, respectively; are the membership values ​​of the peak value and valley value of the load in the i-th period respectively; Step 2.12: Use the peak-valley price ratio to determine the time-of-use electricity price: Assume that the total load curve of all users in a distribution network has been divided into three periods: peak, flat and valley, corresponding to load q p ,q f ,q v , the electricity price is expressed as p p 、p f 、p v , then its peak-valley electricity price can be expressed as: p p =(1+α)p f , p v =(1-β)p f , where α and β are the price increase coefficients during peak hours and price decrease coefficients during valley hours respectively; the average electricity price after price adjustment can be expressed as: Simplifying, we can get: In the above formula, That is, the average electricity price is equal to the normal period electricity price, and we can get αq p -βq v =0, the peak-to-valley price ratio coefficient can be expressed as: The ratio of electricity consumption during peak and valley periods is used as the peak-valley price ratio coefficient. After knowing the electricity consumption during peak and valley periods, the electricity price during peak and valley periods can be determined.

5. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 3, characterized in that: The process of residents responding to the time-of-use electricity price in step 2.2 includes the following steps: Step 2.21: Price elasticity coefficient: The price elasticity coefficient can be expressed as: In the formula, ρ represents the price demand elasticity coefficient; Δp and Δq are the increment of electricity price and the increment of user load respectively. Since the electricity demand of users is not independent, the cross elasticity coefficient is introduced: In the formula, i and j represent different time periods respectively; the user's demand elasticity matrix is ​​formed by combining the self-elasticity coefficient and the cross-elasticity coefficient to express the comprehensive impact of electricity prices on user electricity consumption. First, the elasticity coefficient matrix can be expressed as: The electricity price elasticity coefficient can be expressed in the form of a matrix: Step 2.22: Residents’ satisfaction with time-of-use electricity prices: The expression of satisfaction is: U = U0-U cost θ1-U behavior θ2, where U0 is the user satisfaction without using time-of-use electricity price; U cost , θ1 are the satisfaction with electricity cost change caused by electricity price and the coefficient of satisfaction with cost change; U behavior , θ2 are the satisfaction of users’ electricity consumption behavior changes caused by electricity prices and their satisfaction coefficients; among them, the satisfaction of cost changes can be expressed as: U cost =P-P0, where P is the daily electricity cost when the time-of-use electricity price is implemented; P0 is the user's electricity cost when the time-of-use electricity price is not implemented; the satisfaction with the change in electricity consumption behavior can be expressed as: Where R is the influence coefficient of time-of-use unit price on the change of user's satisfaction with electricity consumption behavior; Q h , Q h0 They are respectively the electricity consumption of the user in the hth period under the time-of-use electricity price and the electricity consumption of the user in the hth period under the time-of-use electricity price.

6. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 1, characterized in that: The interaction management model in step 3 includes the following steps: Step 3.1: Grid-building interaction model; Step 3.2: Grid-Building Interaction Strategy.

7. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 6, characterized in that: The grid-building interaction model in step 3.1 includes the following steps: Step 3.11: Grid benefit model: The grid profit refers to the net profit obtained from selling electricity to buildings after deducting the corresponding electricity purchase costs and network losses. In time period i, the grid net profit is determined by the electricity sales and the corresponding power generation costs, and its calculation formula is as follows: Among them, P i is the net profit of the power grid in period i; is the total electricity purchased by the building in period i, i.e., the electricity sold by the grid; for Function of power generation cost; α, β, γ are constant coefficients; f(D f ) is the dynamic electricity price in period i, which is the fluctuation of electricity demand D f is a function of the power demand The average value of the entire discussion range The deviation between them is calculated as: Step 3.12: Building benefit model: User dissatisfaction refers to the user's dissatisfaction when electricity demand is restricted and deviates from the normal demand level. The calculation formula is as follows: in for user dissatisfaction; is the power demand of building i in time period k; is the original power demand of building k in period i; k , k is a positive constant, indicating the preference of building i to change its electricity demand; The transfer cost refers to the loss or cost caused by demand transfer under the economic incentive of the smart grid. The transfer cost of building k The actual power demand of building i in period k and baseline It is related to the deviation between them, that is, it is related to user dissatisfaction, and its calculation formula is as follows: Where ρ i is a preset positive factor describing the cost level of demand transfer for building i; The electricity fee is the fee paid by the building when it purchases a certain amount of electricity, minus the corresponding transfer costs. The calculation formula is as follows: In the formula For electricity charges.

8. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 6, characterized in that: The grid-building interaction strategy in step 3.2 is an interaction strategy based on the Stackelberg game, specifically: the Nash equilibrium of the preset Stackelberg game is the optimal strategy combination of the grid and the building: In the formula The optimal dynamic electricity price for the day; For building electricity needs; The optimal power demand of the building; g ,I b is the benefit function of the grid and buildings; The Stackelberg game is solved by reverse induction to determine the optimal dynamic electricity price of the power grid and the optimal power demand of each building: First, I b Maximize, and get according to the electricity price given by the power grid Then, determine Substitute I g The optimal dynamic electricity price is obtained, that is, The superscript "-1" in the formula represents the inverse function.

9. The demand side management method based on 5G smart grid daily load characteristics clustering analysis as claimed in claim 2, characterized in that: The grid-building interaction strategy in step 3.2 specifically includes the following processes: Step 3.21: The power grid publishes the day-ahead dynamic electricity price; Step 3.22: Building demand management selects the optimal power demand from its own strategy set according to its own benefit function And feed back to the grid side; Step 3.23: The grid determines the optimal electricity price based on the optimal response strategy fed back by the building and its own benefit function. Step 3.24: Repeat steps 3.21 to 3.23 until Nash equilibrium is reached.

10. The demand side management method based on 5G smart grid daily load characteristics clustering analysis according to claim 1, characterized in that: The enhanced interactive strategy in step 4 considering the inaccurate baseline forecast of demand is specifically as follows: assuming that the uncertainty of the baseline forecast of building power demand follows a normal distribution: The expected value of the baseline prediction uncertainty is used as a correction term to improve the robustness of the proposed interaction strategy: By modifying the Stackelberg game solution, the optimal power demand of the building and the optimal dynamic price of the grid determined by the enhanced interactive strategy are: Among them, rev represents the revised situation.

Images