Industrial user adjustable potential prediction method and system based on secondary clustering

Abstract

The invention proposes an industrial user adjustable potential prediction method and system based on secondary clustering, and relates to the technical field of adjustable potential prediction, and the method comprises the steps: carrying out the first clustering grouping of a plurality of to-be-predicted industrial users through employing the original daily load curve of the industrial users, taking the daily load curve of the group as the daily load curve of the industrial users in the group; based on the difference value between the baseline load curve and the daily load curve of the industrial user, carrying out second clustering grouping on the industrial user, and taking the adjustable potential value of the group as the adjustable potential value of the industrial user in the group; taking the adjustable potential value obtained by secondary clustering as an initial adjustable potential value of the enterprise user, and inputting the initial adjustable potential value and the operation auxiliary data of the industrial user to be predicted into an adjustable potential prediction model to obtain an adjustable potential prediction value; according to the method, industrial users are grouped by using a secondary clustering algorithm, and user groups with similar load characteristics are finely identified, so that the utilization efficiency of electric power resources is improved, and energy conservation and emission reduction are promoted.

Description

Technical Field

[0001] The present invention belongs to the technical field of adjustment potential prediction, and in particular relates to an industrial user adjustable potential prediction method and system based on secondary clustering. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the transformation of the new power system, the installed capacity and power generation of new energy sources have increased year by year. However, the access of a large number of new energy sources to the power system has weakened the regulation capacity of the power generation side. It is difficult to achieve ionization balance relying solely on the power supply on the power generation side. Therefore, tapping the adjustable resources on the load side is of great significance to ensure the safe and stable operation of the power system.

[0004] The rapid development of industrialization has led to a continuous increase in the power load demand of industrial users, which is complex and changeable. Compared with residential and commercial users, industrial users consume more electricity and have a high level of automation control. They are an important part of the adjustable resources on the load side. Therefore, it is particularly important to predict the adjustable potential of industrial users and identify or tap the adjustable industrial users on the load side.

[0005] Traditional methods for predicting the adjustable potential on the power load side are often based on experience or simple classification. For example, a single clustering method is severely affected by unbalanced data, has weak adaptability, and is unable to meet the needs of stratified, refined, and personalized management of industrial user loads, resulting in inaccurate predictions of the adjustable potential of industrial users. Summary of the invention

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a method and system for predicting the adjustable potential of industrial users based on secondary clustering, which uses a secondary clustering algorithm to group industrial users and finely identify user groups with similar load characteristics, thereby improving the efficiency of power resource utilization and promoting energy conservation and emission reduction.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] A first aspect of the present invention provides an industrial user adjustable potential prediction method based on secondary clustering.

[0009] The industrial user adjustable potential prediction method based on secondary clustering includes:

[0010] Using the original daily load curve of industrial users, several industrial users to be predicted are clustered and grouped for the first time, and the daily load curve of the group is used as the daily load curve of the industrial users in the group;

[0011] The difference between the baseline load curve and the daily load curve of the industrial users is taken as the original adjustable potential value, and the industrial users are clustered and grouped for the second time, and the adjustable potential value of the group is taken as the adjustable potential value of the industrial users in the group;

[0012] The adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, and is input into the adjustable potential prediction model together with the operation auxiliary data of the industrial user to be predicted to obtain the final adjustable potential prediction value of the industrial user to be predicted.

[0013] Furthermore, the first clustering grouping adopts K-means algorithm to calculate the distance between the original daily load curves of various industrial users, so as to initially group the industrial users.

[0014] Furthermore, the daily load curve of the group is calculated by summing and averaging the daily load curves of all industrial users in the group.

[0015] Furthermore, the second clustering grouping adopts a hierarchical clustering algorithm to calculate the distance between the original adjustable potential values ​​of each industrial user, so as to further group the industrial users in the same initial grouping to obtain the final grouping of industrial users.

[0016] Furthermore, the adjustable potential value of the group is calculated by summing and averaging the adjustable potential values ​​of all industrial users in the group.

[0017] Furthermore, the operation auxiliary data includes production time proportion, equipment operation efficiency, equipment aging rate, energy price and temperature.

[0018] Furthermore, the adjustable potential prediction model is constructed based on a neural network and uses a simulated annealing algorithm to optimize the model parameters of the adjustable potential prediction model with the goal of minimizing the difference between the actual adjustable potential and the predicted adjustable potential.

[0019] A second aspect of the present invention provides an industrial user adjustable potential prediction system based on secondary clustering.

[0020] Industrial user-adjustable potential prediction system based on secondary clustering, including:

[0021] The first clustering module is configured to: use the original daily load curve of the industrial users to perform a first clustering grouping on a plurality of industrial users to be predicted, and use the daily load curve of the group as the daily load curve of the industrial users in the group;

[0022] The second clustering module is configured to: use the difference between the baseline load curve and the daily load curve of the industrial user as the original adjustable potential value, perform a second clustering grouping on the industrial users, and use the adjustable potential value of the group as the adjustable potential value of the industrial users in the group;

[0023] The potential prediction module is configured as follows: the adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, together with the operation auxiliary data of the industrial user to be predicted, and is input into the adjustable potential prediction model to obtain the final adjustable potential prediction value of the industrial user to be predicted.

[0024] The third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the method for predicting adjustable potential of industrial users based on secondary clustering as described in the first aspect of the present invention.

[0025] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the method for predicting adjustable potential of industrial users based on secondary clustering as described in the first aspect of the present invention are implemented.

[0026] One or more of the above technical solutions have the following beneficial effects:

[0027] The present invention performs secondary clustering and grouping of industrial users based on daily load curves and original adjustable potential values, and finely identifies user groups with similar load characteristics, thereby improving the efficiency of power resource utilization and promoting energy conservation and emission reduction.

[0028] The present invention considers production plan, equipment efficiency, energy price and weather factors to construct an adjustable potential prediction model, optimizes the load adjustable potential prediction value, and thus supports the optimization of load demand response regulation for typical industrial users.

[0029] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0031] Figure 1 This is a flow chart of the method of the first embodiment.

[0032] Figure 2 This is the flow chart of the K-means algorithm of the first embodiment.

[0033] Figure 3 This is a structural diagram of the adjustable potential prediction model of the first embodiment. DETAILED DESCRIPTION

[0034] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0035] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0036] Embodiment 1

[0037] In one embodiment of the present disclosure, a method for predicting the adjustable potential of industrial users based on secondary clustering is provided, such as Figure 1 As shown, the following steps are included:

[0038] Step S1: using the original daily load curves of industrial users, a number of industrial users to be predicted are clustered and grouped for the first time, and the daily load curve of the group is used as the daily load curve of the industrial users in the group.

[0039] Furthermore, the first clustering grouping adopts K-means algorithm to calculate the distance between the original daily load curves of various industrial users, so as to initially group the industrial users.

[0040] Furthermore, the daily load curve of the group is calculated by summing and averaging the daily load curves of all industrial users in the group.

[0041] Specifically, the load data of industrial users is obtained, and the data is preprocessed by standardization or normalization to eliminate the dimensional differences of load data between different users or different time periods; the daily load curve of industrial users is constructed from the preprocessed load data. The daily load curve (daily load curve of electrical power system) is a curve that describes the change of load over time (in hours) within a day, reflecting the law of load change over time within a period of time; the daily load curve is used as a load feature to group enterprise users, and users with similar load characteristics are divided into the same group.

[0042] As the preferred algorithm for processing large amounts of load data, the K-means algorithm has many advantages. Therefore, the K-means algorithm is used to cluster industrial users. The process is as follows: Figure 2 As shown, specifically:

[0043] For the daily load curve, the daily electricity consumption time series of each user (every 15 minutes) is used as the feature vector of the user. Considering the problem of too large dimension of the time series, the maximum load value, minimum load value and peak-to-valley difference are used as feature values ​​to reduce the dimension.

[0044] First, k initial cluster centers are selected. For each data point, its distance to the k cluster centers is calculated, and it is classified into the class where the nearest cluster center is located. The Euclidean distance is used to measure the correlation in the K-means clustering process:

[0045]

[0046] Among them, dist(x i ,x j ) is x i ,x j The Euclidean distance, x i ,x j is the daily load data sequence of users i and j, x ia ,x ja is the ath feature of user i and j, P is the number of feature vectors, here it is 3, x i ,x j belong to the same data set.

[0047] Objective function OBJ of k-means clustering algorithm:

[0048]

[0049] Among them, r ik It means that when the sample x i It is 1 when it is classified into cluster k, otherwise it is 0. k Represents the mean vector of cluster k.

[0050] In general, k centroids are randomly selected from n sample data as the initial cluster centers u1, u2, u3, ..., u k ; Start the loop, calculate the distance from each sample point to the centroid, assign the sample to the centroid that the sample is closer to, and get K clusters; for each cluster, calculate the average distance of all sample points assigned to the cluster as the new centroid; until the objective function converges, that is, all clusters no longer change.

[0051] The optimal number of clusters K is determined by the silhouette coefficient. The silhouette coefficient formula is as follows:

[0052]

[0053] In the formula, A and B represent the cohesion of sample points, the value range of silhouette coefficient S is [-1,1], and the larger the silhouette coefficient, the better the clustering effect. For different K values, the average silhouette coefficient of the entire data set is calculated, and the K value with the largest average silhouette coefficient is selected as the final number of clusters.

[0054] Step S2: taking the difference between the baseline load curve and the daily load curve of the industrial user as the original adjustable potential value, performing a second clustering grouping on the industrial users, and taking the adjustable potential value of the group as the adjustable potential value of the industrial users in the group.

[0055] Furthermore, the second clustering grouping adopts a hierarchical clustering algorithm to calculate the distance between the original adjustable potential values ​​of each industrial user, so as to further group the industrial users in the same initial grouping to obtain the final grouping of industrial users.

[0056] Furthermore, the adjustable potential value of the group is calculated by summing and averaging the adjustable potential values ​​of all industrial users in the group.

[0057] Specifically, the first clustering identifies industrial users with similar daily load curves. In order to group industrial users more finely, the difference between the baseline load curve and the daily load curve of the industrial user is used as the original adjustable potential value. The industrial users in the same group in the first grouping are clustered for the second time, which includes two sub-steps:

[0058] Step S201: Determination of baseline load curve

[0059] Assume that the set of enterprise users to be grouped includes n enterprise users, represented by R = {user 1, user 2, ... user k, ... user n}, and calculate the baseline load curve of each enterprise user, specifically:

[0060] (1) User baseline reference date:

[0061] The selection is divided into two situations: working days and non-working days. The specific selection method is as follows:

[0062] If the industrial user's call (i.e., the adjustment behavior based on the adjustable potential) occurs on a working day, the 5 days before the call date shall be selected as the reference day, and non-working days and call days shall be excluded. The part that is less than 5 days after the exclusion shall be selected in the forward order, and 5 days shall be supplemented;

[0063] If the industrial user's call occurs on a non-working day, the 3 days before the call day will be selected as the reference day, excluding the working day and the call day. The part that is less than 3 days after exclusion will be taken in reverse order to make up 3 days.

[0064] (2) Calculate the initial baseline load curve:

[0065] Determine the calling time T = {t1, t2, ... t i ,...t m};

[0066] Calculate user k at time t i The initial baseline load p ki , the formula is:

[0067]

[0068] Among them, p ki-j is user k at time t i The corresponding load on the jth reference day, p ki is user k at time t i The average load.

[0069] Through the above method, the initial baseline load curve P of each industrial user is finally obtained. b '={P k1 , ...P ki ...P km}.

[0070] (3) Baseline load correction:

[0071] For industrial users whose electricity load is greatly affected by factors such as climate, the initial baseline load curve should be corrected as follows:

[0072] 1) Calculate the correction coefficient K. According to the following formula, obtain K. The value range of K is limited to 0.8~1.2. If it is lower than 0.8, calculate it as 0.8, and if it exceeds 1.2, calculate it as 1.2. The formula is:

[0073]

[0074] Among them, P 2h P′ is the average load at each collection time within 2 hours before the call day, in kilowatts (kW); 2h It is the average value of the historical load value corresponding to the above collection time on the reference day, in kilowatts (kW).

[0075] 2) Result correction: use the correction factor K to correct the initial baseline load curve:

[0076] P b =K×P′ b

[0077] Among them, P b is the corrected baseline load curve.

[0078] Step S202: Secondary clustering

[0079] First, we define adjustable potential. Adjustable potential is adjustable power, and adjustable power is the difference between baseline load and actual load. Therefore, the adjustable potential of industrial users is composed of the adjustable power at each moment in a period of time, that is, the difference between the baseline load curve and the actual load curve. Among them, t i The adjustable power at the moment is expressed as:

[0080] ΔP li =P bi -P ri

[0081] Among them, P ri is the actual load data, P bi is the calculated baseline load, ΔP li For adjustable power, it is a positive value when the load is reduced and a negative value when the load is increased.

[0082] Based on the above adjustable potential, the industrial users in the same group in the first grouping are clustered for the second time. The secondary clustering adopts the hierarchical clustering method. The input is the difference between the daily load curve of each user obtained by the first clustering and the baseline load curve, that is, the adjustable potential. The hierarchical clustering algorithm is used to cluster the enterprise users in the same group to obtain a more refined load adjustable potential classification result. The specific steps are as follows:

[0083] Step 1: Data preparation and preprocessing

[0084] Data collection: Obtain daily load curve data of industrial users in the same group after the first clustering.

[0085] Calculate the difference: Subtract each user's daily load curve from the corresponding baseline load curve point by point to obtain the load adjustable potential data series.

[0086] Step 2: Define the distance metric

[0087] The load-adjustable potential similarity between different users (or clusters) is quantified by Euclidean distance.

[0088]

[0089] Where dist(u,v) is the Euclidean distance between u and v, u and v are the user's adjustable potential data sequences, and P is the number of feature vectors.

[0090] Step 3: Perform hierarchical clustering based on AGNES: AGNES hierarchical clustering algorithm using a bottom-up aggregation strategy. The core idea of ​​the algorithm is to initially treat each object in the data set as a separate cluster, and then gradually merge these clusters according to criteria (such as inter-cluster distance) until a certain termination condition is met (such as reaching a preset number of clusters). The specific steps are:

[0091] Initialization: Treat each user's load scalability potential data as a separate cluster.

[0092] Merging process: Calculate the distances between all cluster pairs; find the two clusters with the smallest distance and merge them into a new cluster; update the inter-cluster distance matrix, considering the distances between the new cluster and other clusters; repeat steps 2 and 3 until the predetermined number of clusters is reached or the stopping condition is met (such as the inter-cluster distance exceeds a threshold).

[0093] Step S3: The adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, and is input into the adjustable potential prediction model together with the operation auxiliary data of the industrial user to be predicted to obtain the final adjustable potential prediction value of the industrial user to be predicted.

[0094] Specifically, in order to accurately predict the final adjustable potential, the factors affecting the adjustable potential are analyzed first, including:

[0095] Production plan: The production plan of industrial users will affect their electricity demand, such as the operating time of the production line, the priority of production tasks, etc.

[0096] Equipment efficiency: The operating efficiency of the equipment, its age, and its maintenance schedule all affect electricity usage.

[0097] Energy prices: Changes in energy prices may affect users' electricity usage behavior. For example, electricity prices during peak hours may affect users' electricity usage decisions.

[0098] Weather factors: For some industrial users, such as agricultural processing or users with large cooling / heating needs, weather conditions (such as temperature, humidity) may affect their electricity demand.

[0099] Since the adjustable potential value ΔP obtained by the above-mentioned secondary clustering has a large error, the production plan, equipment efficiency, energy price, weather factors can be considered, and factors such as production time proportion, equipment operating efficiency, equipment aging rate, energy price, temperature data, etc. can be input to further predict the adjustable potential. To this end, this embodiment selects production time proportion, equipment operating efficiency, equipment aging rate, energy price, and temperature from these factors as auxiliary operation data, and inputs them into the adjustable potential prediction model to obtain the final adjustable potential prediction value of the industrial user.

[0100] The adjustable potential prediction model is constructed based on a neural network and uses a simulated annealing algorithm to optimize the model parameters of the adjustable potential prediction model with the goal of minimizing the difference between the actual adjustable potential and the predicted adjustable potential.

[0101] Specifically, the adjustable potential prediction model adopts the LSTM architecture, inputs the adjustable potential value ΔP obtained by secondary clustering and auxiliary data (production time proportion, equipment operation efficiency, equipment aging rate, energy price, temperature), and outputs the adjustable potential prediction value; the model structure is as follows Figure 3 As shown in the figure, the objective function is to minimize the LSTM mean squared error (MSE), where the LSTM unit calculation model includes a forget gate, an input gate, and an output gate.

[0102] LSTM objective function:

[0103]

[0104] Where N is the number of samples; y i is the actual adjustable potential of the i-th sample; is the predicted adjustable potential of the ith sample, which may be the output of the LSTM model at a certain time step or the last time step.

[0105] LSTM unit calculation model:

[0106] The LSTM network controls the flow of information by introducing a "gate" mechanism, including a forget gate, an input gate, and an output gate. In this embodiment, the sigmoid and tanh functions (with an output range of (0, 1)) are used to control the degree of information passing through, where 0 indicates that information is not allowed to pass through at all and 1 indicates that information is allowed to pass through at all. The forget gate is used to determine how much information in the cell state at the previous moment needs to be forgotten, the input gate is used to determine how much information in the input at the current moment needs to be added to the cell state, and the output gate is used to determine how much information in the cell state at the current moment needs to be output to the hidden state.

[0107] Forget gate, based on the hidden state h of the previous moment t-1 and the current input x t (LSTM model input value at time t), and use the sigmoid function to output a number between 0 and 1 to calculate the retention ratio of cell state information. The calculation formula is as follows:

[0108] f t =σ(W f [x t ,h t-1 ]+b f)

[0109] Input gate, according to the hidden state h at the previous moment t-1 and the current input x t , calculate the information of the updated cell state at the current moment And use the sigmode function to output a number between 0 and 1 to calculate the proportion of update information to be added to the cell state at the current moment. The calculation formula is as follows:

[0110] i t =σ(W i [x t ,h t-1 ]+b i )

[0111]

[0112] Get the current cell state Combined with the cell state information filtered by the forget gate at the previous moment, the updated cell state C at the current moment can be obtained. t :

[0113]

[0114] The output gate determines the output information based on the current cell state information; based on the hidden state h at the previous moment t-1 and the current input x t , calculate the proportion of output information at the current moment; and use the tanh function to change the cell state so that its value is between -1 and 1. The calculation formula is as follows:

[0115] o t =σ(W o [x t ,h t-1 ]+b o )

[0116] h t =o t ⊙tanh(C t )

[0117] W f , W i , W c , W o is the trainable weight matrix of the LSTM model, b f 、b i 、b c 、b o is the bias term. σ is the Sigmoid activation function, […] represents the vector concatenation operation, and ⊙ is the vector bitwise product.

[0118] LSTM network structure:

[0119] A network with two LSTM layers is used. Each LSTM layer contains 100 units. Sigmoid and tanh are used as activation functions inside the LSTM layer. To prevent overfitting, a dropout layer is added after each LSTM layer with a dropout rate of 0.2.

[0120] LSTM initialization weight and bias item settings:

[0121] The weights are sampled from a normal distribution with a mean of 0 and a standard deviation of 0.01, and the bias is initialized to 10. -6 .

[0122] The data input of the adjustable potential prediction model is the initial adjustable potential value ΔP, operation auxiliary data (including production time proportion, equipment operation efficiency, equipment aging rate, energy price, temperature), and the final adjustable potential prediction value is the output.

[0123] The accuracy of the prediction of the adjustable potential prediction model depends on the model parameters. This embodiment adopts a simulated annealing algorithm to optimize the model parameters of the adjustable potential prediction model with the goal of minimizing the difference between the actual adjustable potential and the predicted adjustable potential. The objective function is expressed as:

[0124]

[0125] Where θ is the model parameters (weights, bias terms), T is the number of time periods, and ΔP actual,t is the actual adjustable potential, ΔP predicted,t (θ,O 辅助 ) is the adjustable potential predicted based on the model parameters θ and the auxiliary operation data of the tth time period (such as production time proportion, equipment operation efficiency, equipment aging rate, energy price, temperature).

[0126] In the optimization process, the additional constraint condition is that the adjustable potential is non-negative, which can be expressed as:

[0127] ΔP predicted,t (θ)≥0

[0128] The simulated annealing algorithm is used to optimize the model parameters. The temperature in the simulated annealing algorithm is a parameter used to simulate the temperature change of the system during the physical annealing process. It guides the algorithm's search process by affecting the search range, the probability of accepting new solutions, the annealing speed, and the transition probability. The specific steps are:

[0129] (1) Initialization: Select the initial parameter θ0 of LSTM, set the initial temperature T0 to 1000°C and the cooling coefficient α to 0.9; set the maximum number of iterations N max , take 200.

[0130] (2) Iteration process:

[0131] For each temperature Ti (i=0,1,2,…,N max ):

[0132] Repeat the following steps until the termination condition is met (such as reaching a preset number of iterations or the temperature is below a certain threshold):

[0133] Generate a random parameter change Δθ, which is the change in the model parameters.

[0134] Calculate the new parameter θ′=θ i +Δθ.

[0135] Calculate the value of the objective function under the new parameters f(θ′).

[0136] If f(θ′) <f(θ i ), then accept the new parameter: θ i+1 =θ′.

[0137] Otherwise, with probability Accept a worse solution: If the random number r∈[0,1) is less than this probability, then θ i+1 =θ′, otherwise θ i+1 =θ i .

[0138] Lower the temperature T i+1 =αT i , where is the cooling coefficient.

[0139] (2) Termination condition: reaching the maximum number of iterations N max .

[0140] The temperature Ti is lower than a preset threshold.

[0141] The objective function value f(θ i ) does not change much in multiple consecutive iterations, that is, it converges.

[0142] (3) Output: Return the final optimized parameter θ final .

[0143] After obtaining the optimal model parameter θ final Finally, the adjustable potential prediction model for prediction is built. The final adjustable potential of industrial users can be predicted through the initial adjustable potential value and operating auxiliary data.

[0144] Embodiment 2

[0145] In one embodiment of the present disclosure, an industrial user adjustable potential prediction system based on secondary clustering is provided, comprising:

[0146] The first clustering module is configured to: use the original daily load curve of the industrial users to perform a first clustering grouping on a plurality of industrial users to be predicted, and use the daily load curve of the group as the daily load curve of the industrial users in the group;

[0147] The second clustering module is configured to: use the difference between the baseline load curve and the daily load curve of the industrial user as the original adjustable potential value, perform a second clustering grouping on the industrial users, and use the adjustable potential value of the group as the adjustable potential value of the industrial users in the group;

[0148] The potential prediction module is configured as follows: the adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, together with the operation auxiliary data of the industrial user to be predicted, and is input into the adjustable potential prediction model to obtain the final adjustable potential prediction value of the industrial user to be predicted.

[0149] Embodiment 3

[0150] The purpose of this embodiment is to provide a computer-readable storage medium.

[0151] A computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps in the method for predicting adjustable potential of industrial users based on secondary clustering as described in the first embodiment of the present disclosure.

[0152] Embodiment 4

[0153] The purpose of this embodiment is to provide an electronic device.

[0154] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the method for predicting adjustable potential of industrial users based on secondary clustering as described in the first embodiment of the present disclosure are implemented.

[0155] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the adjustable potential of industrial users based on secondary clustering, characterized in that: include: Using the original daily load curve of industrial users, several industrial users to be predicted are clustered and grouped for the first time, and the daily load curve of the group is used as the daily load curve of the industrial users in the group; The difference between the baseline load curve and the daily load curve of the industrial users is taken as the original adjustable potential value, and the industrial users are clustered and grouped for the second time, and the adjustable potential value of the group is taken as the adjustable potential value of the industrial users in the group; The adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, and is input into the adjustable potential prediction model together with the operation auxiliary data of the industrial user to be predicted to obtain the final adjustable potential prediction value of the industrial user to be predicted.

2. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The first clustering grouping adopts K-means algorithm to calculate the distance between the original daily load curves of various industrial users, so as to initially group the industrial users.

3. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The daily load curve of the group is calculated by summing and averaging the daily load curves of all industrial users in the group.

4. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The second clustering grouping adopts a hierarchical clustering algorithm to calculate the distance between the original adjustable potential values ​​of each industrial user, so as to further group the industrial users in the same initial grouping to obtain the final grouping of industrial users.

5. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The adjustable potential value of the group is calculated by summing and averaging the adjustable potential values ​​of all industrial users in the group.

6. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The auxiliary operation data include production time proportion, equipment operation efficiency, equipment aging rate, energy price and temperature.

7. The method for predicting the adjustable potential of industrial users based on secondary clustering according to claim 1, characterized in that: The adjustable potential prediction model is constructed based on a neural network and uses a simulated annealing algorithm to optimize the model parameters of the adjustable potential prediction model with the goal of minimizing the difference between the actual adjustable potential and the predicted adjustable potential.

8. Industrial user adjustable potential prediction system based on secondary clustering, characterized by: include: The first clustering module is configured to: use the original daily load curve of the industrial users to perform a first clustering grouping on a plurality of industrial users to be predicted, and use the daily load curve of the group as the daily load curve of the industrial users in the group; The second clustering module is configured to: use the difference between the baseline load curve and the daily load curve of the industrial user as the original adjustable potential value, perform a second clustering grouping on the industrial users, and use the adjustable potential value of the group as the adjustable potential value of the industrial users in the group; The potential prediction module is configured as follows: the adjustable potential value obtained by secondary clustering is used as the initial adjustable potential value of the enterprise user, together with the operation auxiliary data of the industrial user to be predicted, and is input into the adjustable potential prediction model to obtain the final adjustable potential prediction value of the industrial user to be predicted.

9. An electronic device, comprising: a memory for non-transitory storage of computer readable instructions; a processor for executing the computer-readable instructions; Wherein, when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 7 is executed.

10. A storage medium, characterized in that: The computer-readable instructions are non-transitorily stored, wherein when the computer-readable instructions are executed by a computer, the method of any one of claims 1 to 7 is performed.

Images