CN111598612B - Time-sharing electricity price making method - Google Patents

Abstract

The invention discloses a time-sharing electricity price making method, which comprises the following steps: predicting a wind power interval at a moment to be predicted; randomly sampling in a wind power interval at a moment to be predicted to generate 100 groups of wind power scenes; reducing the scenes into 10 groups of wind power scenes; adding a wind power scene into a user load to generate a net load scene graph; dividing a peak-valley period in a net load scene into H time periods, and then dividing the H time periods into a peak period, a flat period and a valley period; calculating the wind power scene expectation, and generating an equivalent net load curve; constructing a time-of-use electricity price model according to the equivalent net load; and (4) performing multi-objective solution on the time-sharing electricity price model by adopting an NSGA-II algorithm. Compared with a method without considering wind power uncertainty, the method has the advantages that the randomness of the wind power is fully considered, the user is effectively guided to respond to the time-of-use electricity price, the goals of balancing system load and load shifting are achieved, and meanwhile the electricity utilization habits of the user are not influenced.

Description

A method for formulating time-of-use electricity prices

Technical Field

The invention belongs to the technical field of new energy and relates to a time-of-use electricity price formulation method.

Background Art

Time-of-use electricity pricing is a price-based demand response measure. It can set different electricity prices for users at different times of electricity consumption, guide users to use electricity rationally, and effectively adjust users' electricity consumption behavior to achieve the purpose of peak shaving and valley filling and optimizing the load curve.

At present, China has abundant wind power reserves, and the government encourages new energy to access the grid first. Wind power can be connected to the system as a negative load, but due to its strong volatility, poor regularity, and difficulty in prediction, it has great uncertainty. Therefore, its access will increase the operating pressure of the system, increase the peak-to-valley difference of the original system, and affect the safe and stable operation of the system. The study of time-of-use electricity prices considering the uncertainty of wind power as a user-side demand response can make full use of demand-side resources, promote the consumption of new energy, effectively adjust the peak-to-valley difference of the system, and improve the economic and stable operation ability of the system, which has become an important research direction at present. Many researchers regard wind power output as a random variable, and establish chance constraint models, random fuzzy models, and random probability models to describe wind power. Some literature assumes that wind speed obeys Rayleigh distribution or Weibull distribution, and decomposes the random process of uncertainty into several typical discrete probability scenarios, and calculates the probability of each scenario in combination with the power characteristics of wind turbines. The scenario method is an important method for dealing with wind power uncertainty. Most studies use Monte Carlo sampling, Latin hypercube sampling and other methods to generate wind power scenarios after obtaining simulated wind power, and use backward reduction method (BR method), fast forward selection method (FFS method), scenario tree construction method (STC method), clustering partitioning method and other methods to reduce uncertainty scenarios to generate typical wind power scenarios. As a means of prediction, interval prediction can accurately predict the upper and lower limits of wind power, thereby providing more accurate information for scheduling and operation. It can be used as a method to deal with wind power uncertainty in time-of-use electricity prices. However, wind power has countless values within the interval, which is difficult to connect with existing loads, affecting the formulation of time-of-use electricity prices. It is necessary to convert the uncertainty of interval prediction into deterministic scenario research.

Although the above studies have discussed time-of-use electricity prices and handling wind power uncertainty, there are two types of problems in the current methods of handling wind power uncertainty time-of-use electricity prices:

(1) Existing studies on time-of-use electricity prices that incorporate wind power generally use theoretical probability models to simulate wind power, but fail to take into account the complexity of the real environment, resulting in a certain error between the simulated wind power and the actual wind farm output;

(2) The current scene reduction method can easily solve the problem of small-scale scene reduction, but as time increases, the scene scale grows exponentially, the accuracy of the scene reduction results decreases, and there is a disadvantage that it cannot give the best clustering results.

The access of wind power can greatly alleviate the pressure on power supply. As an important demand-side response, time-of-use electricity price can effectively improve resource utilization. Combining the two can make fuller use of resources and optimize resource allocation. However, due to the randomness and volatility of wind power, it is difficult to calculate the time-of-use electricity price with wind power.

Summary of the invention

The purpose of the present invention is to provide a time-of-use electricity price formulation method, which solves the problem of error between simulated wind power and actual wind farm output existing in the prior art.

The technical solution adopted by the present invention is a method for formulating time-of-use electricity prices, comprising the following steps:

Step 1: Use continuous wind power samples of the wind farm to predict the wind power range at the time to be predicted;

Step 2: randomly sample and generate 100 groups of wind power scenarios within the wind power range at the time to be predicted;

Step 3: Use the binary K-means clustering algorithm to reduce the scenarios to 10 groups of wind power scenarios;

Step 4: Add the wind power scenario obtained in step 3 to the user load to generate a net load scenario diagram;

Step 5: Use 0-1 integer programming to divide the peak and valley periods in the net load scenario into H time periods, and then divide the H time periods into peak, flat, and valley periods;

Step 6: Calculate the wind power scenario expectation and generate an equivalent net load curve;

Step 7: construct a time-of-use electricity price model based on the equivalent net load, wherein the objective function is to minimize the peak-to-valley difference of the system equivalent net load and minimize the user's electricity dissatisfaction after implementing the peak-to-valley electricity price;

Step 8: Use the NSGA-II algorithm to solve the multi-objective problem of the time-of-use electricity price model, obtain the Pareto optimal solution set, evaluate the objective function, and select the optimal solution.

The present invention is also characterized in that:

Step 1 specifically includes:

Step 1.1, set the condition number τ, the number of segmentation intervals K, and the confidence level β;

Step 1.2: Obtain the discrete probability distribution function based on the continuous wind power samples of the wind farm and the corresponding marginal distribution function

的概率Pj由大到小依次进行累加，假设当累加到j＝q时，

则在置信度水平β下的边缘分布函数位于区间

所属子区间的并集内，通过逆函数计算得到待预测时刻风电功率区间

Step 1.3: Discrete probability distribution function

The probability Pj of is accumulated from large to small. Assume that when j=q,

Then the marginal distribution function at the confidence level β is located in the interval

In the union of the sub-intervals, the wind power interval at the time to be predicted is obtained by inverse function calculation

It also includes the use of PIAW and PICP indicators to evaluate the accuracy of the wind power range at the time of prediction.

Step 3 specifically includes:

Step 3.1, first group 100 wind power scenarios into clusters;

Step 3.2, select a cluster from the cluster table;

Step 3.3: Use K-means algorithm to divide the cluster into two clusters;

Step 3.4: Repeat step 3.3 until the preset number of experiments is reached;

Step 3.5: Select two clusters with the smallest total error from the clusters obtained in step 3.4;

Step 3.6, add the two clusters obtained in step 3.5 to the cluster table;

Step 3.7. Repeat steps 3.2-3.6 until the cluster table contains 10 clusters.

Step 5 specifically includes:

Step 5.1, divide the peak and valley periods in the net load scenario into H time periods, assuming that i is the start time of a certain divided period and j is the end time, and define the distance between objects i and j as d _ij , then the matrix [d _ij ] is an N×N matrix with rows and columns labeled 0, 1, …N-1;

The objective function of the model that divides the peak and valley periods into H cluster periods is

In the above formula, z _ij is a 0-1 variable. When the start time of a period is i and the end time is j, the value of z _ij is 1, otherwise it is 0; _Dij is the sum of all _dij in the same time period:

In the above formula, mod is the modulo operation;

The constraints of the model that divides the peak and valley periods into H cluster periods are:

Step 5.2: Divide the H time periods into peak, flat, and valley periods.

Step 6 specifically includes:

The expected wind power scenario obtained in step 6.1 and step 3 is:

为每个场景的概率值，P_wind为风电出力功率，θ_w为风功率场景集；In the above formula, θ represents a possible wind power scenario.

is the probability value of each scenario, P _wind is the wind power output, and θ _w is the wind power scenario set;

Step 6.2: The equivalent net load expression before implementing the peak-valley electricity price is:

In the above formula, Q(t) is the user load.

Step 7 specifically includes:

Step 7.1: Construct the objective function based on the equivalent net load:

Objective function 1: After implementing the peak-valley electricity price, the difference in equivalent net load between the peak and valley periods of the system is minimized:

minC=min[maxL(t)-minL(t)] (16);

In the above formula, L(t) is the equivalent net load of the system after the implementation of peak and valley electricity prices;

Objective function 2: Minimize user dissatisfaction with electricity consumption:

表示t时段用户用电量的变化率；In the above formula,

Indicates the rate of change of user electricity consumption during period t;

Step 7.2: Calculate the load transfer rate and obtain the user response curve based on historical data fitting:

In the above formula, Δp _ab is the price difference between period a and period b, h _ab is the saturation threshold of the price difference, l _ab is the dead zone threshold of the price difference, K _ab is the slope of the linear region during the load transfer process, and λ _max is the maximum load transfer rate;

Similarly, the load transfer rate from the peak period to the normal period of the equivalent net load is obtained as follows: λ _fp , the load transfer rate from the peak period to the valley period of the equivalent net load is obtained as follows: _{λ fg ,} the load transfer rate from the normal period to the valley period of the equivalent net load is obtained as follows:

为执行分时电价前各个时段的等效净负荷平均值；步骤7.3、构建约束条件：In the above formula, _Tf , _Tp , _Tg represent the peak, flat and valley periods respectively, _Lk0 , _Lk are the loads before and after the implementation of time-of-use electricity price;

The equivalent net load average value of each period before the implementation of the time-of-use electricity price; Step 7.3, construct the constraint conditions:

Constraints 1. Time constraints:

T＝T _f +T _p +T _g =24 (20);

Constraint 2: Load constraint:

L = L _f + L _p + L _g (21);

Constraint 3: After the implementation of time-of-use electricity prices, the user's electricity expenditure is less than or equal to the expenditure cost before the implementation of time-of-use electricity prices:

U ₀ ≥ U _TOU (24);

Constraint 4: Limit the peak-to-valley price ratio after implementing time-of-use electricity prices:

In the above formula, _k1 and _k2 are constants that limit the peak-valley electricity price ratio.

The beneficial effects of the present invention are:

The invention discloses a time-of-use electricity price formulation method, which establishes a discrete probability distribution function of a time to be predicted, mines the correlation between wind power sequences in adjacent time periods to predict the wind power interval, can improve the wind power prediction accuracy, and reduce the gap with the actual wind power; by combining the prediction interval with random sampling, the uncertainty of the interval prediction is converted into a deterministic scenario, and the binary K-means is used to reduce the scenario, which is not only accurate and efficient, but also does not need to find a complex alternative model, can improve the calculation efficiency, and solves the problem of combining the interval prediction result with the time-of-use electricity price customization; in the case of considering multiple scenarios, the integer programming method is used to divide the peak, flat and valley time periods, which can effectively deal with the continuity problem of the time periods; based on the consumer psychological response, a multi-objective optimization time-of-use electricity price model is established with the minimum peak-valley difference and the minimum user satisfaction, and the NSGA-II multi-objective method is used to optimize the model parameters to obtain the non-inferior solution set of the time-of-use electricity price model; compared with the method that does not consider the uncertainty of wind power, the randomness of wind power is fully considered, and the user is effectively guided to respond to the time-of-use electricity price, so as to achieve the goal of balancing the system load and shaving the peak and filling the valley, while ensuring that the user's electricity consumption habits are not affected.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of a method for formulating a time-of-use electricity price according to the present invention;

FIG2 is a load transfer rate curve diagram of a time-of-use electricity price setting method of the present invention;

FIG3 is a flow chart of the NSGA-II algorithm used in a time-of-use electricity price setting method of the present invention.

DETAILED DESCRIPTION

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

A time-of-use electricity price formulation method, as shown in FIG1 , comprises the following steps:

Step 1: Use continuous wind power samples of the wind farm to predict the wind power range at the time to be predicted;

Step 1.1, set the condition number τ, the number of segmentation intervals K, and the confidence level β;

步骤1.2.1、假设连续t+1时刻的风功率序列为[X_T-t,X_T-t-1,…,X_T]：Step 1.2: Obtain the discrete probability distribution function based on the continuous wind power samples of the wind farm and the corresponding marginal distribution function

Step 1.2.1, assume that the wind power sequence at the consecutive time t+1 is [X _Tt ,X _Tt-1 ,…,X _T ]:

In the above formula, each row represents a sample in the X domain;

The marginal distribution function value corresponding to the historical sample of wind power is calculated as:

In the above formula, F _Tt ,…F _T are the marginal distribution functions of wind power in the Tt,…T periods respectively. Each row represents a sample in the F domain, and there are N samples in total.

必然落入风功率序列的子空间

其中，P₁,…,P_l…P_t∈{1,2,…,K}；即F_T-t(x_T-t-1)∈((P₁-1)δ,P₁δ],…,F_T(x_T)∈((P_t+1-1)δ,P_t+1δ]；Step 1.2.2: Assume that the interval (0, 1) to which F _Tt, …F _T belongs includes K intervals, namely intervals S ₁ , …S _K , where S ₁ = [0, δ], δ = 1/K, S _j = [(j-1)δ, jδ], j = 2, …K, K is a positive integer, then the wind power sequence [X _Tt, X _Tt-1, …, X _T ] forms K ^t+1 subspaces, then any sample in the F domain

It must fall into the subspace of the wind power series

Among them, P ₁ ,…,P _l …P _t ∈{1,2,…,K}; that is, F _Tt (x _Tt-1 )∈((P ₁ -1)δ,P ₁ δ],…,F _T (x _T )∈((P _t+1 -1)δ,P _t+1 δ];

落在相同区间内，条件矩阵为：Select N ₁ samples in formula (2) to form a conditional matrix, and the first t elements of each sample are respectively

Falling in the same interval, the conditional matrix is:

Step 1.2.3: Divide the N ₁ samples in formula (3) into J categories, and classify the samples whose elements in the t+1th column fall in the same subinterval into the same category. The number of samples in each category is: M ₁ ,…M _j ,…,M _J ;

每类样本出现的概率P^j：Step 1.2.4: The interval of the t+1th element of each class of samples is the same, but the value is different; in order to obtain a same value to represent the t+1th element in each class of samples, use formula (4) to calculate the value of the t+1th column of each class of samples:

The probability of each type of sample appearing P ^j is:

作为场景，P^j为该场景出现的概率，

为以

为条件

的离散概率分布函数。If you will

As a scenario, ^Pj is the probability of the scenario occurring,

For

Condition

The discrete probability distribution function of .

的概率Pj由大到小依次进行累加，假设当累加到j＝q时，

则在置信度水平β下的边缘分布函数位于区间

所属子区间的并集内，通过逆函数计算得到待预测时刻风电功率区间

Step 1.3: Substitute the discrete probability distribution function

The probability Pj of is accumulated from large to small. Assume that when j=q,

Then the marginal distribution function at the confidence level β is located in the interval

In the union of the sub-intervals, the wind power interval at the time to be predicted is obtained by inverse function calculation

按照概率进行降序排列，记排列后离散概率分布函数为

P⁽¹⁾为最大概率，P^(J)为最小概率；从j＝1开始对P⁽¹⁾进行累加，直至累加和大于等于β停止；假设当累加到j＝q时，

则在置信度水平β下的边缘分布函数位于区间

所属子区间的并集内，假设：Specifically, the discrete probability distribution function

Arrange in descending order according to probability, and the discrete probability distribution function after arrangement is:

P ⁽¹⁾ is the maximum probability, P ^(J) is the minimum probability; start accumulating P ⁽¹⁾ from j = 1 until the accumulated sum is greater than or equal to β; suppose that when the accumulated sum reaches j = q,

Then the marginal distribution function at the confidence level β is located in the interval

In the union of the subintervals, assume that:

Then the interval of the marginal distribution function is:

计算在置信度水平β下，风电功率在第T时刻发生的区间

即得到待预测时刻风电功率区间。Through the inverse function

Calculate the interval of wind power generation at time T under the confidence level β

That is, the wind power range at the time to be predicted is obtained.

The PIAW and PICP indices are used to evaluate the accuracy of the discrete probability distribution function prediction results under different values of m and K.

Specifically, prediction interval coverage probability (PICP) and prediction interval average width (PIAW) are used as evaluation criteria to verify the prediction accuracy of the proposed algorithm. The calculation formula of interval coverage probability (PICP) is:

In the above formula, U is the total number of wind power to be predicted, u = 1, 2, ..., U; Au is the indicative function, defined as:

与V_u 分别为预测区间的上下边界，V_u为预测时刻的风电功率实际值。区间平均宽度(PIAW)的计算式为：That is, when the actual value of wind power at the time to be predicted falls within the prediction interval, Au takes the value of 1, otherwise it takes the value of 0;

and _Vu are the upper and lower boundaries of the prediction interval, respectively, and _Vu is the actual value of wind power at the prediction moment. The calculation formula for the average width of the interval (PIAW) is:

PICP indicates the number of actual wind power values falling within the prediction interval. On the basis of meeting the confidence level, the larger its value, the more actual wind power falls within the prediction interval, which means the better the prediction effect. PIAW indicates the average width of the upper and lower boundaries of the interval. When the same confidence level is met, the smaller its value, the narrower the prediction interval, which means the prediction interval is closer to the actual value.

Step 2: randomly sample and generate 100 groups of wind power scenarios within the wind power range at the time to be predicted;

Step 3: Use the binary K-means clustering algorithm to reduce the scenarios to 10 groups of wind power scenarios;

Step 3.1, after initializing 100 groups of wind power scenarios, group the 100 groups of wind power scenarios into clusters;

Step 3.2, select a cluster from the cluster table;

Step 3.3: Use K-means algorithm to divide the cluster into two clusters;

Step 3.4: Repeat step 3.3 until the preset number of experiments is reached;

Step 3.5: Select two clusters with the smallest total error from the clusters obtained in step 3.4;

Step 3.6, add the two clusters obtained in step 3.5 to the cluster table;

Step 3.7. Repeat steps 3.2-3.6 until the cluster table contains 10 clusters.

Step 4: Add the wind power scenario obtained in step 3 to the user load to generate a net load scenario diagram;

Step 5: Divide the net load scenario into time periods;

First, the net load scenario is divided into H time periods using 0-1 integer programming. The total number of peak and valley time periods is T, and H can be set to any integer from 1 to 24 hours according to the user load situation. Then, the H time periods are divided into peak, flat, and valley time periods according to the load.

Step 5.1: First, use 0-1 integer programming to divide the net load scenario into H time periods, assuming that i is the start time of a certain divided time period and j is the end time. The continuity of the time period needs to be considered in the division of the time period. If the start time i of a certain divided time period is less than the end time j, any time r (i＜r＜j) should belong to the same time period; if the start time i is greater than the end time j, the time period range is i≤r≤T and 1≤r≤j.

In this embodiment, the object to be analyzed is the T moments to be divided, and the parameter of a certain moment is the net load value vector under different scenarios at that moment. In practical applications, the net load value should be normalized. Define the distance between moments i and j as d _ij , then the matrix [d _ij ] is an N×N matrix, with rows and columns labeled 0, 1, ... N-1; the objective function of the model that divides the peak and valley periods into H cluster periods is

In the above formula, z _ij is a 0-1 variable. When the start time of a period is i and the end time is j, the value of z _ij is 1, otherwise it is 0; _Dij is the sum of all _dij in the same period:

In the above formula, mod is the modulo operation;

The constraints of the model that divides the peak and valley periods into H cluster periods are:

The constraint condition (12) indicates that time w can only be included in a single time period, and its left side is the sum of all possible partitions that may include time w. In the time period that includes w, it can be divided into the following three cases: 0≤i≤w≤j≤N-1, 0≤j＜i≤w≤N-1 and 0≤w≤j＜i≤N-1; in the constraint condition (13), the total number of time periods obtained after the partition is H;

Step 5.2: Divide the H time periods into peak, flat, and valley periods.

Step 6: Calculate the wind power scenario expectation and generate an equivalent net load curve;

The wind power scenario obtained in step 3 is expected to be:

为每个场景的概率值，P_wind为风电出力功率，θ_w为风功率场景集；In the above formula, θ represents a possible wind power scenario.

is the probability value of each scenario, P _wind is the wind power output, and θ _w is the wind power scenario set;

The equivalent net load expression of the system before implementing the peak-valley electricity price is:

In the above formula, Q(t) is the user load.

Step 7: construct a time-of-use electricity price model based on the equivalent net load, wherein the objective function is to minimize the peak-to-valley difference of the system equivalent net load and minimize the user's electricity dissatisfaction after implementing the peak-to-valley electricity price;

Step 7.1: Construct the objective function:

Objective function 1: After implementing the peak-valley electricity price, the difference in equivalent net load between the peak and valley periods of the system is minimized:

minC=min[maxL(t)-minL(t)] (16);

In the above formula, L(t) is the equivalent net load of the system after the implementation of peak-valley electricity prices, and C represents the difference between peak and valley.

Objective function 2: Minimize user dissatisfaction with electricity consumption:

表示t时段用户用电量的变化率；In the above formula,

Indicates the rate of change of user electricity consumption during period t;

Step 7.2: In the process of solving the objective function, the user's responsiveness should be considered. According to the theory of consumer psychology, the present invention considers that users have different degrees of response to different electricity prices, which will also trigger different consumption behaviors. The electricity price is a stimulus to the user. The response to this stimulus is not infinite, but there is a difference threshold. When the stimulus is too large and exceeds the upper limit of the threshold, the user will reach saturation and basically no longer transfer the load. The electricity price range greater than the upper limit of the threshold is called the response saturation zone; similarly, when the stimulus is too small and below the lower limit of the threshold, the user will not be able to perceive and will not respond. This electricity price range below the lower limit of the threshold is called the dead zone. Within the threshold range, the user will respond to the stimulus, and the degree of response is approximately linear with the size of the stimulus, which is called the linear zone. From this, an approximate piecewise function can be obtained, which is determined by three parameters: the dead zone threshold, the linear segment slope, and the saturation zone threshold. Different types of user responses will be reflected in different parameters.

The concept of load transfer rate is introduced here: the load transfer rate is defined as the ratio of the amount of user load transferred from the high electricity price period to the low electricity price period after the implementation of the peak-valley electricity price to the load in the high electricity price period. It is assumed that the load transfer rate is proportional to the price difference between peak and flat, peak and valley, and flat and valley. Based on a large amount of historical data, the price difference and load transfer rate of different peak, flat and valley periods are calculated, and a piecewise function curve that approximates the actual response curve of the user can be fitted, as shown in Figure 2. The horizontal axis represents the price difference of different periods, and the vertical axis represents the load transfer rate:

In the above formula, Δp _ab is the price difference between period a and period b, h _ab is the saturation threshold of the price difference, l _ab is the dead zone threshold of the price difference, K _ab is the slope of the linear region during the load transfer process, and λ _max is the maximum load transfer rate;

Similarly, we can obtain the load transfer rate from peak period to normal period of equivalent net load λ _fp , the load transfer rate from peak period to valley period of equivalent net load λ _fg , and the load transfer rate from normal period to valley period λ _pg , and thus the load of each period after the implementation of time-of-use electricity price is obtained as follows:

为执行分时电价前各个时段的等效净负荷平均值。In the above formula, _Tf , _Tp , _Tg represent the peak, flat and valley periods respectively, _Lk0 , _Lk are the loads before and after the implementation of time-of-use electricity price;

It is the average value of equivalent net load in each period before the implementation of time-of-use electricity price.

Step 7.3: Construct constraints:

Constraints 1. Time constraints:

T＝Tf+ _Tp + _Tg ＝24 (20)

Constraint 2: Power (load) constraint:

L＝ _Lf + _Lp + _Lg (21)

Constraint 3: In order to enable users to actively transfer loads from peak hours to off-peak hours, the electricity expenditure of users after the implementation of time-of-use electricity prices is equal to or less than the expenditure cost before the implementation of time-of-use electricity prices:

U ₀ ≥ U _TOU (24);

Constraint 4: To avoid the situation where the difference between peak and valley electricity prices is too large after the implementation of time-of-use electricity prices, resulting in excessive user response and peak-valley inversion; or the peak and valley electricity prices are not obvious, the user response is insufficient, and the expected target cannot be achieved, the range of the peak-valley electricity price ratio is limited:

In the above formula, _k1 and _k2 are constants that limit the peak-to-valley electricity price ratio. In China, the value of the peak-to-valley electricity price ratio is generally between 2 and 5.

Step 8: Use the NSGA-II algorithm to solve the multi-objective problem of the time-of-use electricity price model, as shown in Figure 3, to obtain the Pareto optimal solution set, evaluate the objective function, and select the optimal solution.

Step 8.1: Read in equivalent net load data, set algorithm parameters and variable ranges, including optimal individual coefficient, population size, maximum evolutionary generation, and fitness function deviation; variable range refers to decision variables (the allowable range of peak, flat, and valley electricity prices);

Step 8.2: Randomly generate an initial population P ₀ and perform non-dominated sorting;

Step 8.3: After non-dominated sorting, the first generation of offspring population is obtained through three basic operations: selection, crossover, and mutation;

Step 8.3: Merge the parent population with the child population, perform fast non-dominated sorting, and calculate the crowding degree of individuals in each non-dominated layer;

Step 8.4: Select appropriate individuals to form a new parent population based on the non-dominated relationship and the crowding degree of the individuals;

Step 8.5: Obtain the next generation population through the three basic operations of selection, crossover, and mutation, and execute step 8.3 until the maximum number of evolutionary generations is reached.

Step 8.6: After the iteration, the Pareto optimal solution set is obtained, and the optimal solution is selected according to the following method; first, the peak-to-valley difference and user dissatisfaction data in the Pareto optimal solution set are standardized using formula (26); where _xi is the value to be standardized, _xj is the value after standardization, and MIN = min( _xi ).

After data standardization, the weights of the peak-to-valley difference and user dissatisfaction are set according to the importance of the two accuracy indicators. Finally, the two evaluation indicators of all points in the Pareto optimal solution are weightedly calculated, and the point with the smallest weighted value is the optimal solution.

Through the above methods, a time-of-use electricity price formulation method of the present invention establishes a discrete probability distribution function of the time to be predicted, mines the correlation between wind power sequences in adjacent time periods to predict the wind power interval, can improve the wind power prediction accuracy, and reduce the gap with the actual wind power; by combining the prediction interval with random sampling, the uncertainty of the interval prediction is converted into a deterministic scenario, and the binary K-means is used to reduce the scenario, which is not only accurate and efficient, but also does not need to find a complex alternative model, can improve the calculation efficiency, and solves the problem of combining the interval prediction result with the time-of-use electricity price customization; when considering multiple scenarios, the integer programming method is used to divide the peak, flat and valley time periods, which can effectively deal with the continuity problem of the time period; based on the consumer psychology response, a multi-objective optimization time-of-use electricity price model is established with the minimum peak-valley difference and the minimum user satisfaction, and the NSGA-II multi-objective method is used to optimize the model parameters to obtain a non-inferior solution set of the time-of-use electricity price model; compared with the method that does not consider the uncertainty of wind power, the randomness of wind power is fully considered, and the user is effectively guided to respond to the time-of-use electricity price, so as to achieve the goal of balancing the system load and shaving the peak and filling the valley, while ensuring that the user's electricity consumption habits are not affected.

Claims (7)

1. A time-sharing electricity price making method is characterized by comprising the following steps:

step 1, predicting a wind power interval at a time to be predicted by using a continuous wind power sample of a wind power plant;

step 2, randomly sampling and generating 100 groups of wind power scenes in the wind power interval at the moment to be predicted;

step 3, adopting a binary K-means clustering algorithm to reduce the scenes into 10 groups of wind power scenes;

step 4, adding the wind power scene obtained in the step 3 into the user load to generate a net load scene graph;

step 5, dividing a peak-valley period in a net load scene into H time periods by adopting 0-1 integer programming, and then dividing the H time periods into a peak period, a flat period and a valley period;

step 6, calculating the wind power scene expectation, and generating an equivalent net load curve;

step 7, constructing a time-of-use electricity price model according to the equivalent net load, wherein the target function is that the peak-to-valley difference of the equivalent net load of the system is minimum and the electricity utilization dissatisfaction of a user is minimum after the peak-to-valley electricity price is implemented;

and 8, performing multi-target solution on the time-of-use electricity price model by adopting an NSGA-II algorithm to obtain a Pareto optimal solution set, evaluating a target function and selecting an optimal solution.

2. The time-sharing electricity price making method according to claim 1, wherein the step 1 specifically comprises:

1.1, setting a condition number tau, a segmentation interval number K and a confidence level beta;

step 1.2, obtaining a discrete probability distribution function according to a continuous wind power sample and a corresponding edge distribution function of a wind power plant

Step 1.3, the discrete probability distribution function

Probability P of ^j Accumulation is performed in sequence from large to small, assuming that when j = q is accumulated, a combination is selected>

The edge distribution function at the confidence level β is located in the interval ∑ er>

In the union of the sub-intervals, the wind power interval and the time to be predicted are calculated through an inverse function to obtain the wind power interval>

3. The time-sharing electricity price making method according to claim 1 or 2, characterized by further comprising the step of performing precision evaluation on the wind power interval at the time to be predicted by using PIAW and PICP indexes.

4. The method for formulating a time-of-use electricity price according to claim 1, wherein the step 3 specifically comprises:

step 3.1, firstly, forming 100 groups of wind power scenes into clusters;

3.2, selecting the cluster from a cluster table;

3.3, dividing the cluster into two clusters by adopting a K-means algorithm;

step 3.4, repeating the step 3.3 until the preset experiment times are reached;

3.5, selecting two clusters with the minimum total error from the clusters obtained in the step 3.4;

step 3.6, adding the two clusters obtained in the step 3.5 into a cluster table;

and 3.7, repeating the steps 3.2-3.6 until the cluster table contains 10 clusters.

5. The time-sharing electricity price making method according to claim 1, wherein the step 5 specifically comprises:

step 5.1, dividing the peak-valley period in the net load scene into H time periods, assuming that i is the starting time of a certain divided period, j is the ending time, and defining the distance between the objects i and j as d _ij Then matrix [ d _ij ]Is an NxN matrix, and the row and column marks are 0,1, \ 8230and N-1;

the objective function of the model divided into H clustering periods in the peak-valley period is

In the above formula, z _ij Is a variable of 0 to 1When the starting time of a certain period is i and the ending time is j, z _ij The value is 1, otherwise 0; d _ij For all d in the same time period _ij The sum of (1):

in the above formula, mod is a modulo operation;

the constraint conditions of the model for dividing the peak-valley period into H clustering periods are as follows:

and 5.2, dividing the H time periods into peak time periods, flat time periods and valley time periods.

6. The method for formulating a time-of-use electricity price according to claim 1, wherein the step 6 specifically comprises:

the wind power scene obtained in step 6.1 and step 3 is expected to be:

in the above equation, theta represents a certain possible wind power scenario,

for each scene probability value, P _wind For wind power output power, theta _w A set of wind power scenarios;

step 6.2, executing an equivalent net load expression before peak-valley electricity price as follows:

in the above equation, Q (t) is the user load.

7. The method for formulating a time-of-use electricity price according to claim 1, wherein the step 7 specifically includes:

step 7.1, constructing an objective function according to the equivalent net load:

the target function 1, the minimum equivalent net load difference of the system in the peak and valley period after the peak-valley electricity price is implemented:

min C＝min[max L(t)-min L(t)] (16)；

in the above formula, L (t) is the system equivalent payload after peak-to-valley electricity prices are implemented;

objective function 2, minimum user electricity dissatisfaction:

in the above formula, the first and second carbon atoms are,

representing the change rate of the electricity consumption of the user in the t period;

and 7.2, calculating the load transfer rate, and fitting according to historical data to obtain a user response curve:

in the above formula,. DELTA.p _ab Is the difference in electricity prices of the time period a and the time period b, h _ab Saturation region threshold for electricity price difference,/ _ab Dead zone threshold for electrovalence difference, K _ab Is the slope of the linear region, λ, during load transfer _max The maximum load transfer rate;

the load from the peak period to the ordinary period of the equivalent net load can be obtained by the same methodTransfer rate lambda _fp Load transfer rate lambda from peak period to valley period of equivalent net load _fg Load transfer rate lambda from flat period to valley period of equivalent net load _pg The load of each period after the time-of-use electricity price is executed is thus obtained as:

in the above formula, T _f 、T _p 、T _g Respectively represent three periods of peak-to-valley, L _k0 ,L _k Loads before and after the time-of-use electricity price are executed;

the equivalent net load average value of each time interval before the time-of-use electricity price is executed;

and 7.3, constructing constraint conditions:

constraint 1, time constraint:

T＝T _f +T _p +T _g ＝24 (20)；

constraint 2, load constraint:

L＝L _f +L _p +L _g (21)；

constraint condition 3, the expenditure of the user electricity rate after the implementation of the time-of-use electricity rate is less than or equal to the expenditure cost before the implementation of the time-of-use electricity rate:

U ₀ ≥U _TOU (24)；

constraint condition 4, limiting the peak-to-valley electrovalence ratio after implementation of time-of-use electrovalence:

in the above formula, k ₁ And k ₂ Is a constant that limits the peak-to-valley electrovalence ratio.

Images