Time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elastic modeling

Abstract

The invention relates to the technical field of power system and demand side management, in particular to a time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elastic modeling, and the method comprises the steps: collecting the electricity consumption data of a user, carrying out the preprocessing of the data, and extracting a key index as a clustering feature; the method comprises the following steps of: performing clustering on indexes based on clustering algorithm initialization and a K-means clustering algorithm, dividing users into different groups, constructing a demand response model for each user group, establishing an electricity price demand elastic matrix model, performing modeling in combination with user willingness and psychological response characteristics, and constructing a time-of-use electricity price optimization model based on a clustering layering result. An electricity price scheme is iteratively optimized through the whale algorithm, a time-of-use electricity price strategy is globally searched and optimized through the whale optimization algorithm, the whale optimization algorithm improves the search optimization efficiency by simulating whale surrounding and spiral foraging behaviors, the user behavior is combined with a power grid target, and the user experience is improved. And the time-of-use electricity price is optimally designed from the system level.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems and demand-side management, and in particular to a time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling. Background Art

[0002] Time-of-use electricity pricing, a key price adjustment mechanism in the power system, focuses on setting differentiated electricity prices based on power supply and demand conditions at different times. Specifically, a day is typically divided into peak, flat, and off-peak periods, with different electricity prices set for each period. This guides users to adjust their electricity usage, shifting some of their peak load to off-peak hours, thereby achieving the goal of peak load shifting.

[0003] To optimize time-of-use electricity pricing, some research is using cluster analysis to categorize user electricity usage patterns and establish demand elasticity models to model demand response. Cluster analysis can classify user load curves and identify groups of users with similar electricity usage patterns. Demand elasticity models can quantify the sensitivity of user electricity demand to changes in electricity prices.

[0004] However, most existing methods are single-perspective modeling, focusing only on user electricity consumption patterns or demand elasticity analysis, failing to organically combine user behavior with grid goals, and making it difficult to achieve optimal design of time-of-use electricity prices from a system level.

[0005] Therefore, it is necessary to provide a new time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling. Summary of the Invention

[0006] In view of the above problems existing in the prior art, an embodiment of the present invention aims to provide a time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling.

[0007] To achieve the above objectives, the present invention adopts a technical solution: a time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling, comprising: S1, collects user electricity consumption data and preprocesses the data to extract key indicators as clustering features; S2, clustering indicators based on clustering algorithm initialization and K-means clustering algorithm to divide users into different groups; S3, build a demand response model for each user group, establish an electricity price demand elasticity matrix model and combine user willingness and psychological response characteristics to model; S4, based on the clustering and stratification results, builds a time-of-use electricity price optimization model and iteratively optimizes the electricity price scheme through the whale algorithm.

[0008] Furthermore, in S1, user electricity load data is collected from smart meters and energy management systems to obtain user electricity load curves and user behavior data (including appliance on-time periods, controllable load equipment lists, electricity usage time preferences, and historical response records) to obtain the initial data set.

[0009] Furthermore, the abnormal values ​​in S1 are recorded; Specifically, outliers refer to observations in a data set that deviate significantly from the majority of data points and are significantly different from other data. The existence of outliers will distort statistics such as the mean and variance, and interfere with the accuracy of algorithms such as regression and clustering.

[0010] Furthermore, in S2, clustering the indicators based on clustering algorithm initialization and K-means clustering algorithm includes: The clustering algorithm is used to quickly generate the initial centroid and improve the convergence speed, and then the K-means clustering algorithm is used to divide the users into several groups in order to distinguish the response potential of different user groups.

[0011] Furthermore, the K-means clustering algorithm is used to divide users into several groups, including clustering the feature data preprocessed by the K-means clustering algorithm to divide users into several groups with similar load patterns; the clustering results are labeled by combining with domain knowledge.

[0012] Furthermore, in S3, an electricity price demand elasticity matrix model is established, including: The demand elasticity coefficient of each time period is solved to quantify the user's response potential for load reduction and load transfer under different electricity price schemes. An improved S-shaped function is used as the user behavior response function, and the relationship between price difference and user load transfer rate is mapped into a saturated "S" curve.

[0013] Furthermore, in S4, based on the clustering and stratification results, a time-of-use electricity price optimization model is constructed, including: Set the objective function such as "minimizing the system peak-valley load difference" or "minimizing the overall cost", and the constraints include non-negative electricity price difference and grid balance. Use intelligent optimization algorithms such as particle swarm optimization to search and optimize the electricity price strategies for peak, flat and valley periods.

[0014] Furthermore, search optimization includes: In each data iteration, the elasticity model and response function are used to predict the load distribution of each group under the new electricity price, calculate the overall load curve, and update the electricity price parameters accordingly until the preset target is met.

[0015] The embodiment of the present invention further provides a network-side server, comprising: At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so as to enable the at least one processor to execute the above-mentioned time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling.

[0016] An embodiment of the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling.

[0017] The beneficial effects of the present invention are as follows: by collecting user electricity consumption data and preprocessing the data, key indicators are extracted as clustering features, the indicators are clustered based on clustering algorithm initialization and K-means clustering algorithm, users are divided into different groups, a demand response model is constructed for each user group, an electricity price demand elasticity matrix model is established and modeled in combination with user willingness and psychological response characteristics, a time-of-use electricity price optimization model is constructed based on the clustering stratification results, and the electricity price scheme is iteratively optimized through the whale algorithm, thereby achieving the effect of combining user behavior with power grid goals and optimizing the design of time-of-use electricity prices from a system level. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below with reference to the accompanying drawings and examples.

[0019] In the picture: Figure 1 A flowchart of a time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling provided in the first embodiment of the present invention; Figure 2 for Figure 1 Flowchart of clustering processing; Figure 3 for Figure 1 User response flow chart in the middle; Figure 4 for Figure 1 Whale algorithm processing flow chart; Figure 5 It is a structural diagram of a network-side server provided according to a second embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0021] The first implementation method is to collect user electricity consumption data and pre-process the data, extract key indicators as clustering features, cluster the indicators based on clustering algorithm initialization and K-means clustering algorithm, divide users into different groups, build a demand response model for each user group, establish an electricity price demand elasticity matrix model and combine user willingness and psychological response characteristics to model, build a time-of-use electricity price optimization model based on the clustering stratification results, and iteratively optimize the electricity price plan through the whale algorithm, so as to achieve the effect of combining user behavior with power grid goals and optimizing the design of time-of-use electricity prices from the system level.

[0022] A first embodiment of the present invention provides The following is a detailed description of the implementation details of the time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling in this embodiment. The following content is only for the convenience of understanding the implementation details and is not necessary for the implementation of this solution. The specific process of this embodiment is as follows: Figure 1 As shown, S1, collects user electricity consumption data and preprocesses the data to extract key indicators as clustering features; Specifically, user electricity load data is collected from smart meters and energy management systems to obtain user electricity load curves and user behavior data (including appliance startup time, controllable load equipment list, electricity usage time preference, and historical response records) to obtain the initial data set.

[0023] Record the abnormal values ​​in S1; Specifically, outliers refer to observations in a data set that deviate significantly from the majority of data points and are significantly different from other data. The existence of outliers will distort statistics such as the mean and variance, and interfere with the accuracy of algorithms such as regression and clustering.

[0024] Preprocess the user's electricity load data set. The specific steps are as follows: Step S11: Collect user electricity load data from smart meters and energy management systems to obtain user electricity load curves and user behavior data (including appliance on-time periods, controllable load equipment lists, electricity usage time preferences, and historical response records) to obtain an initial data set.

[0025] Step S12: Process outliers. Outliers refer to observations in a data set that significantly deviate from the majority of data points and are significantly different from other data points. The presence of outliers can distort statistics such as mean and variance, interfering with the accuracy of algorithms such as regression and clustering.

[0026] First, detect abnormal data: the 3σ principle should be followed, that is, eliminate or correct data that meets the | - |>3σ data points. If the sample data and the overall mean If the absolute deviation exceeds 3 times the standard deviation (3σ), the data point is considered an outlier and can be considered for removal or correction.

[0027] For n sample data, the formulas for the mean and standard deviation are:

[0028] When | - When |>3σ, Considered as an outlier, is the Greek letter representing the population mean, n is the sample size, that is, the total number of data points, Represents the standard deviation of the data.

[0029] The data of outliers are corrected, and the corrected data expression is as follows:

[0030] α and β are the weight coefficients of the load data of the node before and after, respectively, and the sum of the weight coefficients is equal to 1.

[0031] Step S13: Fill in the missing values ​​in the data set. Missing values ​​refer to the observation values ​​of certain fields or time points in the data set that are not recorded or stored, and are expressed as null values. The linear interpolation method of adjacent time periods is used:

[0032] in, Indicates at a point in time The missing values ​​at Indicates the point before the missing value The observed value of Indicates the time point after the missing value The observed value of Indicates the time point corresponding to the missing value, Indicates the time point corresponding to the previous valid observation value of the missing value, Indicates the time point corresponding to the next valid observation value after the missing value.

[0033] Step S14: Processing of noise points in the data set. Noise points are random errors or deviations in the data, manifesting as erroneous values ​​or isolated points that deviate from expectations. This paper uses the DBSCAN algorithm to process noise points. The steps are as follows: The neighborhood radius is calculated using the nearest neighbor distance method, and the formula is:

[0034] in, represents the neighborhood radius, Each sample is The average distance to the nearest neighbors, is the total number of samples.

[0035] The minimum number of neighborhood points is calculated using the mathematical expectation method. The formula is:

[0036] in, represents the minimum number of neighborhood points, It is Objects in The number of objects in the neighborhood.

[0037] Different applications in DBSCAN value, and observe the number of noise points as When the number of noise points tends to be stable, the corresponding The value is the optimal parameter.

[0038] Use cosine similarity to measure the distance between samples. The formula is:

[0039] in, are the feature vectors of the two samples, Transpose of a vector.

[0040] After DBSCAN clustering, the data marked as noise points will be filtered out to obtain a cleaner load data set.

[0041] Step S15, extracting and analyzing characteristic indicators of user electricity consumption: the characteristic indicators of user electricity consumption are mainly extracted from load curves and user behavior data.

[0042] The main characteristic indicators for extracting load curves include average daily charge, maximum demand, minimum demand, load rate, load fluctuation rate, peak-valley load rate and peak load ratio.

[0043] The average daily charge can reflect the user's overall electricity consumption scale and the continuity of the user's electricity consumption behavior. The formula is as follows:

[0044] in, represents the average daily charge, For users at all times The power load value, The number of time.

[0045] The maximum demand among the characteristic indicators reflects the user's instantaneous maximum pressure on the power grid, and the minimum demand indicates the lower limit of the user's basic power demand:

[0046] in, Indicates the maximum demand. Indicates the user at time The maximum value of the power load value, Indicates the minimum required quantity, Indicates the user at time The minimum value of the power load value.

[0047] The load rate refers to the electricity consumption pattern of users during a specific statistical period. The formula is as follows:

[0048] in, Indicates the total number of times the demand is counted within a specific period of time. Indicates the demand per unit time within a specific time period. Indicates the maximum demand per unit time.

[0049] The load fluctuation rate reflects the stability of electricity consumption, and its formula is as follows:

[0050] in, represents the load fluctuation rate, For users at all times The power load value, is the average daily charge, The duration of electricity usage.

[0051] The peak-valley load rate reflects the difference between peak and valley power consumption. The formula is:

[0052] in, Indicates the peak-valley load rate, Indicates the maximum demand. Indicates the minimum required quantity, is the average daily charge.

[0053] Peak load ratio indicates the time concentration of electricity consumption, and the formula is as follows:

[0054] in, represents the peak load ratio, is the load value at time t, is a set of predefined peak time periods, Indicates the total number of time periods, Indicates the peak load ratio.

[0055] The main characteristic indicators of user behavior data include: historical response frequency, concentration of active time periods, and regularity of work and rest.

[0056] Historical response frequency is the percentage of times users participated in previous demand response activities. , which can reflect the reliability of users and predict their behavior.

[0057] By analyzing the concentration of active time periods, users’ electricity usage patterns can be identified: The high concentration of electricity consumption is shown, that is, electricity consumption is highly concentrated in a few periods (such as the peak time at noon in shopping malls). , it shows a low concentration of electricity consumption, that is, the electricity load is uniform (such as a continuous production factory).

[0058]

[0059] in, Indicates the concentration of active time periods, These are the three periods with the highest electricity consumption throughout the day. It indicates the total power load value generated in the three periods with the highest power consumption throughout the day. Indicates the power load value for the whole day.

[0060] Regularity of work and rest is used to measure the stability of residents' work and rest. The formula is:

[0061] in, A quantitative value indicating a stable work and rest schedule. is the load curve for day d, is the historical average curve, and DTW is the dynamic time warping distance, which measures the similarity of the curves. The smaller the value of the expression, the more regular the residents' daily routines.

[0062] Step S16: pre-processing and vectorizing the load curve and user behavior data to extract the user's electricity consumption characteristic index, standardize it, and construct a characteristic vector. The specific steps are as follows: Extract load curve , peak and valley load rate , load rate The standardized feature vector is constructed by using the characteristic indicators, and the formula is as follows:

[0063] in, : No. The normalized feature vector of each user, is the user load curve vector, is the peak-valley load rate, is the mean value of the load curve of all users, is the standard deviation of the daily load curve for all users, Peak and valley load rates for all users, Standard deviation of peak and valley load rates of all users, is the average load factor of all users, is the standard deviation of the load factor of all users, is the load rate, and n is the dimension of the normalized feature vector.

[0064] Save data list .

[0065] See Figure 2 ,S2, clusters the indicators based on clustering algorithm initialization and K-means clustering algorithm, and divides users into different groups; Clustering of indicators based on clustering algorithm initialization and K-means clustering algorithm includes: The clustering algorithm is used to quickly generate the initial centroid and improve the convergence speed. Then the K-means clustering algorithm is used to divide the users into several groups in order to distinguish the response potential of different user groups.

[0066] The K-means clustering algorithm is used to divide users into several groups, including clustering the feature data preprocessed by the K-means clustering algorithm to divide users into several groups with similar load patterns; the clustering results are labeled by combining with domain knowledge.

[0067] The improved Canopy clustering algorithm generates initial cluster centers, which is characterized by optimizing the initial centroid selection through an adaptive mechanism, specifically including: Step S21, setting the dual distance threshold through cross-validation: 、

[0068] in, The radius of cluster coverage is used as a loose boundary threshold, The core neighborhood radius is used as the tight boundary threshold, is the average Euclidean distance of random sample pairs.

[0069] Step S22: Calculate the fitness function such as the load curve similarity of each user, perform roulette wheel selection according to probability, and prioritize high-quality individuals. The specific steps are as follows: The fitness of each user's electricity consumption feature vector is calculated based on the similarity of the load curve. The fitness calculation formula is:

[0070] in, For the The fitness value of each user is in the range of [0,1], which represents the smoothness of the load curve. is the standard deviation of the load curve, std represents the standard deviation, because it is negatively correlated with the fitness range, so it is preferred to be selected as the initial center.

[0071] The roulette wheel algorithm is used to select the initial center point, giving priority to individuals with high fitness, and removing them from L to create the first cluster center point. The roulette wheel probability selection formula is as follows:

[0072] in, For the The probability that a user is selected as a cluster center, The total number of users, For the The fitness value of a user.

[0073] Step S23: Dynamically generate Canopy clusters, that is, divide users into different groups. The specific steps are as follows: Select from list L by probability as the initial center and remove After that, the remaining Calculate the minimum Euclidean distance to all centers , where the calculation formula of Euclidean distance is as follows:

[0074] in, is the horizontal coordinate of each n-dimensional feature vector of the data in the list L, is the horizontal coordinate of the initial center point, is the ordinate of each n-dimensional feature vector of the data in list L, is the vertical coordinate of the initial center point.

[0075] ① If , then classify it into Canopy and mark it as strongly associated, update the center to the cluster mean, and remove .

[0076] ②If , then it will be classified into Canopy and marked as weakly associated; ③If , then it will be Set up as a new center; Step S24, repeat step S23 until the list L is empty.

[0077] Step S25: Delete isolated Canopy clusters with sample numbers less than a threshold, output the final cluster number K value and center point set, and improve the load characteristic consistency of the main cluster. The specific steps are as follows: The conditions for isolated cluster removal are as follows:

[0078] in, For the The number of users in a Canopy cluster, is the threshold for determining isolated clusters, is the total number of users.

[0079] The final output cluster number K value and center point set C:

[0080] Step S26: Based on K-means fine clustering, use C output by Canopy as the initial cluster center of K-means clustering and iterate until convergence. The specific steps are as follows: Assign each user to the cluster corresponding to the nearest center, satisfying the following formula:

[0081] in, For the A user group is a cluster, which contains a set of users with similar load characteristics. is the standardized feature vector of a single user, representing its load characteristics, For the The center vector of each cluster represents the typical load characteristic pattern of the group. is the center vector of other clusters, and , representing the load characteristic patterns of different groups.

[0082] The user Assigned to the group with the most similar load characteristics , that is, select the cluster center with the smallest Euclidean distance The corresponding group.

[0083] Recalculate the typical load pattern of each group. The new center is the arithmetic mean of the feature vectors of all users in the group. The formula is as follows:

[0084] in, is the updated cluster center vector, reflecting the average mode of user load characteristics within the group, Cluster The number of users in the group, that is, the total number of users belonging to the group, Cluster The feature vector of a single user in .

[0085] Step S27, repeat step S26 until the maximum number of iterations is reached or the center point change satisfies the following formula .

[0086] Step S28: Verify the optimal number of clusters based on the silhouette coefficient and the sum of squared errors, and perform multi-dimensional evaluation based on the Davidson-Botting index to evaluate and select the clustering results. The specific steps are as follows: Calculate the silhouette coefficient S and the sum of squared errors (SSE) corresponding to different K values ​​to determine the optimal number of clusters. The specific formula of the silhouette coefficient is as follows:

[0087] in, The overall silhouette coefficient range is 、 is the cohesion, representing the user The average distance to other users in the same cluster, is the separation degree, characterizing the user The average distance to the nearest user in a different cluster.

[0088] The specific formula for the sum of squared errors is as follows:

[0089] in, The smaller the value is, the tighter the clustering is. is the final number of clusters.

[0090] The optimal number of clusters is selected, which maximizes S and has an obvious SSE inflection point, that is, a small SSE.

[0091] Step S29: The Davidson-Boulding index used in the multi-dimensional evaluation of inter-cluster separation is as follows:

[0092] in, The Davidson-Botting Index, is the intra-cluster dispersion, characterizing the cluster The average distance from users to the center, Characterize clusters as the degree of separation between clusters With cluster Euclidean distance to the center.

[0093] When DBI < 0.8, it indicates good separation between clusters.

[0094] Step S210: Based on the characteristic indicators of S1, the divided groups with similar load patterns are classified and labeled as user groups such as "high response potential type" and "stable type". The specific steps are as follows:

[0095] in, is the peak-valley load rate, is the load rate, and the peak load ratio is the proportion of electricity consumption during the peak electricity price period. >0.7 indicates large load fluctuations, <0.4 indicates that the difference between peak and valley electricity consumption is significant.

[0096] The high response potential type has the characteristics of large load curve fluctuations and high peak-to-valley difference; the stable type has the characteristics of smooth load curve and low elasticity; the peak-sensitive type has the characteristics of daily load concentrated in the peak electricity price period.

[0097] The present invention combines clustering user load patterns with the construction of user willingness and psychological behavior response functions to form a multi-perspective comprehensive model, which can systematically optimize time-of-use electricity price strategies, achieve precise incentives for different user groups and coordinated response of the overall load, thereby significantly improving the overall load response efficiency.

[0098] See Figure 3 ,S3, build a demand response model for each user group, establish an electricity price demand elasticity matrix model and combine user willingness and psychological response characteristics to model; Establish an electricity price demand elasticity matrix model, including: The demand elasticity coefficient of each time period is solved to quantify the user's response potential for load reduction and load transfer under different electricity price schemes. An improved S-shaped function is used as the user behavior response function, and the relationship between price difference and user load transfer rate is mapped into a saturated "S" curve.

[0099] S31, solve the demand elasticity coefficient of each time period and establish the electricity price demand elasticity matrix model, as follows: Solving the period elasticity coefficient (group average elasticity × period adjustment factor)

[0100] Among them, K is the user group number, i.e., group identification, is the time period adjustment factor, It is the peak preference period of the group.

[0101]

[0102] Resilience Matrix Construction

[0103] in is the inter-period inhibition factor, is the group time sensitivity coefficient.

[0104] S32, analyzing user intentions and psychological response characteristics, as follows: Willingness-fatigue bimodal factor (willingness incentive item × response fatigue item)

[0105] in, For price sensitivity, is the fatigue decay rate, The last response time.

[0106] Anchoring Effect Correction

[0107] in, is the actual electricity price change, Anchoring electricity prices for the group benchmark, For the recent experience of electricity prices, is the anchor sensitivity coefficient, is the memory decay factor.

[0108] Recent electricity prices are lower than the anchor point: users are more sensitive to price increases and tend to overreact Recent electricity prices are higher than the anchor point: users are numb to price increases and are slow to respond. Loss aversion reinforcement: (Price increase elasticity > price reduction elasticity) S33 uses an improved S-shaped function as the user behavior response function to calculate the load transfer rate, quantify the user's load reduction and load transfer response potential under different electricity price plans, and map the relationship between the price difference and the user's load transfer rate into a saturated "S" curve, as follows: Improved S-type function

[0109] in, is the maximum transfer rate (the proportion of the load that can be transferred to the group), Characterizing the elasticity matrix eigenvalue analysis for response sensitivity, is the saturation compensation amount ( )、 is the saturation rate ( ).

[0110] Load transfer rate:

[0111] in, is the time period coupling factor, To transfer out of peak hours, It is time to enter the trough period.

[0112]

[0113] Quantification of loads that can be reduced or transferred:

[0114]

[0115] S34: Build a demand response model for each user group, construct a corresponding reinforcement elasticity matrix, combine user willingness and psychological response characteristics, and list the response function, as follows: High response potential group model ( >0.7, η<0.4) Dynamically reinforced elastic matrix:

[0116] Elastic modulus:

[0117] Fourth-order S-shaped response function:

[0118] Stationary population model ( <0.3, η>0.8) Constant elastic matrix:

[0119] Only retain the autoelastic coefficient

[0120] No time period correction

[0121] Linear response function:

[0122] Peak-sensitive population model (peak load ratio > 60%) Time period partition elasticity matrix:

[0123] Elastic modulus:

[0124] Time period definition: peak=[17,18,19,20] valley=[0,1,2,3,4,5] Time period correction:

[0125] Transfer efficiency model:

[0126] Where, transfer distance = target period - original period.

[0127] Third-order S-shaped response function:

[0128] The electricity price plans of the three are shown in Table 1: Table 1 Electricity price plans for three types of electricity users

[0129] This paper introduces an improved S-shaped (Logistic) function to describe the relationship between electricity price and load response. Compared with traditional linear or piecewise models, it can more realistically simulate the nonlinear response behavior of users. This S-shaped function has strong adaptability and can flexibly match the user response characteristics under different electricity pricing strategies by adjusting the function parameters. It can also accurately depict the saturation effect of user response: when the electricity price difference exceeds a certain threshold, the user load response tends to stabilize, which is more in line with the actual situation.

[0130] See Figure 4 ,S4, based on the clustering and stratification results, a time-of-use electricity price optimization model is constructed, and the electricity price scheme is iteratively optimized through the whale algorithm.

[0131] Based on the clustering and stratification results, the time-of-use electricity price optimization model is constructed, including: Set the objective function such as "minimizing the system peak-valley load difference" or "minimizing the overall cost", and the constraints include non-negative electricity price difference and grid balance. Use intelligent optimization algorithms such as particle swarm optimization to search and optimize the electricity price strategies for peak, flat and valley periods.

[0132] Search optimization includes: In each data iteration, the elasticity model and response function are used to predict the load distribution of each group under the new electricity price, calculate the overall load curve, and update the electricity price parameters accordingly until the preset target is met.

[0133] S41, determine the objective function: For time-of-use electricity price optimization, the model of the present invention sets two objective functions, the formulas are as follows:

[0134]

[0135] in, is the price difference of electricity in different time periods; To minimize the peak and valley load rates of the system; To maximize user response benefits; Indicates peak and valley load rate; Indicates the maximum demand. Indicates minimum required quantity; is the average daily charge; For transferable loads.

[0136] S42, constraints include: ① Electricity price structure constraints: The peak-period electricity price is required to be higher than the normal-period electricity price, the normal-period electricity price is required to be higher than the valley-period electricity price, and the valley-period electricity price is not lower than the marginal cost of power supply.

[0137] ② Load balance constraint: The total power consumption throughout the day remains unchanged, and the integral area of ​​the optimized load curve is equal to the integral area of ​​the original load curve.

[0138] ③ User response behavior boundary: The load transfer rate does not exceed the maximum transfer capacity of the group, that is, the high response type does not exceed 85%; the stable type does not exceed 35%; the peak sensitive type does not exceed 70%.

[0139] ④ Time period coupling restriction: The time difference of load transfer across time periods is ≤ 4 hours. For example, the load at 19:00 can only be transferred to 15:00-23:00.

[0140] S43, the present invention is based on two objective functions and assigns weight coefficients thereto respectively. and , using the linear weighted fusion method, at the same time, the multi-objective optimization problem with constraints is converted into a single-objective optimization problem. The fitness value optimization function in the present invention can be obtained:

[0141] in, is the constraint violation penalty term introduced in the objective function, and is the weighting coefficient.

[0142] The Whale Optimization Algorithm simulates the three hunting behaviors of a whale group: random search, encircling prey, and attacking prey. These three population update mechanisms are independent of each other, allowing the global exploration and local exploitation processes to be run and controlled separately during the optimization phase. In the algorithm, each humpback whale position represents a potential solution, and the global optimal solution is found by continuously updating the whales' positions.

[0143] Compared to other swarm intelligence optimization algorithms, the Whale Algorithm has a novel structure, fewer control parameters, and demonstrates superior optimization performance in solving many numerical optimization and engineering problems. Therefore, this step uses the Whale Algorithm to optimize time-of-use electricity prices.

[0144] The optimization steps of the whale algorithm are: S44, calculate the fitness of all individual positions according to the fitness value calculation function, and record the optimal individual. The formula is as follows:

[0145] S45, Update , and according to the current Get the new coefficient vector and , the formula is as follows:

[0146]

[0147] in, is a vector that decreases linearly with the number of iterations, and are random vectors distributed between [0,1]; the coefficient vector Used to control the range of movement of the individual toward or away from the selected target; coefficient vector Used to determine the individual's path and direction around prey.

[0148] S46, Whale Algorithm based on probability and To decide which method to use to update the position, generate a random number between [0, 1] , and and Make a judgment: ① Calculation of attacking prey behavior: ≥0.5, When is any value, in order to approach the optimal solution, the individual adopts a movement strategy that combines randomness and determinism, gradually narrowing the search range and concentrating on the most promising area.

[0149] This process approaches the final optimal solution by increasing deterministic moves. The whale algorithm implements this process through the logarithmic spiral equation. The updated position formula is as follows:

[0150] in, is the position vector of the current best individual in the whale group at the Nth iteration; is the logarithmic spiral coefficient; A random number between [-1, 1].

[0151] ②Calculate random search behavior: When <0.5, When >1, in order to fully search in the space, the whale algorithm updates the position according to the distance between individuals, and the searching individual will swim towards the random whale to achieve the purpose of random search.

[0152] This approach is similar to how whales randomly swim in the ocean in search of food sources. The position update of each solution is unpredictable, but the goal is to explore as much area as possible to increase the chance of finding a high-quality solution. The position update formula is as follows:

[0153] in, is the number of algorithm iterations; is the position vector of a random individual in the whale group at the Nth iteration.

[0154] ③Calculate the behavior of surrounding prey: <0.5, When ≤1, the whale algorithm assumes that the current best candidate solution is the target prey or is close to the optimal solution, and other search individuals will try to move towards the best search individual, which is similar to a whale discovering a school of fish and starting to swim around them in preparation for prey.

[0155] The position updates of the solutions become more precise and focused at this stage, and they adjust their direction and position according to the position of the current best solution.

[0156] The purpose of this phase is to narrow the search range and focus on finding the optimal solution in the most promising area. The position update formula is as follows:

[0157] S47, the position update is completed, the fitness of each individual is calculated, and compared with the position of the previously retained optimal whale. If it is better, the new optimal solution is used to replace it.

[0158] S48, determine the number of algorithm iterations Whether the maximum number of iterations has been reached : when hour, Self-increment, and return to step S42 to continue calculation.

[0159] when , output the current optimal solution and end the algorithm.

[0160] This paper uses the Whale Optimization Algorithm (WOA) to perform global search and optimization of time-of-use electricity pricing strategies. By simulating the encircling and spiraling foraging behavior of whales, WOA achieves a good balance between global search and local exploitation during the iterative process, significantly improving optimization search efficiency. The WOA algorithm has a simple structure, a small number of parameters, rapid convergence, and higher solution accuracy, making the optimization process for time-of-use electricity pricing strategies more efficient and the results more accurate.

[0161] This embodiment collects user electricity consumption data and preprocesses the data, extracts key indicators as clustering features, clusters the indicators based on clustering algorithm initialization and K-means clustering algorithm, divides users into different groups, builds a demand response model for each user group, establishes an electricity price demand elasticity matrix model and combines user willingness and psychological response characteristics to model, builds a time-of-use electricity price optimization model based on the clustering and stratification results, and iteratively optimizes the electricity price plan through the whale algorithm, thereby achieving the effect of combining user behavior with power grid goals and optimizing the design of time-of-use electricity prices at the system level.

[0162] It is worth noting that all modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, to highlight the innovations of the present invention, this embodiment does not include units that are not closely related to solving the technical problems proposed by the present invention. However, this does not mean that other units do not exist in this embodiment.

[0163] The second embodiment of the present invention relates to a network side server, such as Figure 5 As shown, it includes at least one processor 302; and a memory 301 that is communicatively connected to the at least one processor 302; wherein the memory 301 stores instructions that can be executed by the at least one processor 302, and the instructions are executed by the at least one processor 302 to enable the at least one processor 302 to execute the above-mentioned data processing method.

[0164] Memory 301 and processor 302 are connected using a bus. The bus can include any number of interconnected buses and bridges, connecting various circuits of one or more processors 302 and memory 301. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits. These are all well known in the art and are therefore not described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 302 is transmitted over a wireless medium via an antenna. Furthermore, the antenna receives data and transmits it to processor 302.

[0165] The processor 302 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. The memory 301 can be used to store data used by the processor 302 when performing operations.

[0166] A third embodiment of the present invention relates to a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling in the first embodiment.

[0167] That is, those skilled in the art will understand that all or part of the steps in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a program. The program is stored in a storage medium and includes a number of instructions for causing a device (which may be a single-chip microcomputer, chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes: a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc., various media that can store program code.

[0168] The above is only an embodiment of the present invention. Common knowledge such as the known specific structures and characteristics in the scheme is not described in detail here. Ordinary technicians in the field are aware of all common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the field can improve and implement this scheme in combination with their own abilities under the inspiration given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

[0169] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling, characterized in that: include: S1, collects user electricity consumption data and preprocesses the data to extract key indicators as clustering features; S2, clustering indicators based on clustering algorithm initialization and K-means clustering algorithm to divide users into different groups; S3, build a demand response model for each user group, establish an electricity price demand elasticity matrix model and combine user willingness and psychological response characteristics to model; S4, based on the clustering and stratification results, builds a time-of-use electricity price optimization model and iteratively optimizes the electricity price scheme through the whale algorithm.

2. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 1 is characterized in that: In S1, user electricity load data is collected from smart meters and energy management systems to obtain user electricity load curves and user behavior data (including appliance on-time, controllable load equipment list, electricity time preference, and historical response records) to obtain the initial data set.

3. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 2 is characterized in that: Record the abnormal values ​​in S1; Specifically, outliers refer to observations in a data set that deviate significantly from the majority of data points and are significantly different from other data. The existence of outliers will distort statistics such as the mean and variance, and interfere with the accuracy of algorithms such as regression and clustering.

4. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 1 is characterized in that: In S2, clustering of indicators based on clustering algorithm initialization and K-means clustering algorithm includes: The clustering algorithm is used to quickly generate the initial centroid and improve the convergence speed. Then the K-means clustering algorithm is used to divide the users into several groups in order to distinguish the response potential of different user groups.

5. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 4 is characterized in that: Using the K-means clustering algorithm to divide users into several groups, including clustering the feature data pre-processed by the K-means clustering algorithm to divide users into several groups with similar load patterns; The clustering results are annotated by combining with domain knowledge.

6. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 1 is characterized in that: In S3, the electricity price demand elasticity matrix model is established, including: The demand elasticity coefficient for each time period is solved to quantify the user's response potential for load reduction and load transfer under different electricity price schemes. An improved S-shaped function is used as the user behavior response function, and the relationship between price difference and user load transfer rate is mapped into a saturated "S" curve.

7. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 1 is characterized in that: In S4, based on the clustering and stratification results, the time-of-use electricity price optimization model is constructed, including: Set the objective function such as "minimizing the system peak-valley load difference" or "minimizing the overall cost", and the constraints include the non-negative electricity price difference and grid balance. Use intelligent optimization algorithms such as particle swarm optimization to search and optimize the electricity price strategies for peak, flat and valley periods.

8. The time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling according to claim 7 is characterized in that: Search optimization includes: In each data iteration, the elasticity model and response function are used to predict the load distribution of each group under the new electricity price, calculate the overall load curve, and update the electricity price parameters accordingly until the preset target is met.

9. A network-side server, characterized in that: include: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the time-of-use electricity price optimization method based on electricity consumption data clustering and demand response elasticity modeling as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for optimizing time-of-use electricity prices based on electricity consumption data clustering and demand response elasticity modeling according to any one of claims 1 to 8 is implemented.

Images