A method and system for controlling power consumption of thermal storage electric heating based on Pso-CNN

By clustering and dynamic model optimization of thermal storage electric heating loads based on Pso-CNN, combined with particle swarm and convolutional neural network, the problems of large control dimensions and reduced prediction accuracy of thermal storage electric heating loads are solved, and the problem of large control dimensions and reduced prediction accuracy is realized. Efficient new energy consumption and load fluctuations have stabilized, reducing electricity consumption costs.

CN114742410BInactive Publication Date: 2024-11-29XINJIANG UNIVERSITY +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210380057.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-12
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing technology is difficult to effectively solve the problem of large load control dimensions, reduced prediction accuracy, and deep reinforcement learning methods in user homes, resulting in the failure to optimize power consumption strategies in real time.

Method used

Using a Pso-CNN-based method, cluster the thermal storage electric heating loads in the region, establish a dynamic model of temperature and energy storage, use particle swarm optimization algorithm and convolutional neural network to predict future electricity consumption strategies, and combine user comfort Constrain, optimize grid status and load information mapping, and maximize new energy consumption.

Benefits of technology

The thermal storage electric heating load control is achieved in the high proportion of new energy grid-connected scenarios, reducing electricity consumption costs, suppressing load fluctuations, improving prediction accuracy, and suitable for actual scenarios and avoiding the occurrence of load peaks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114742410B_ABST
    Figure CN114742410B_ABST
Patent Text Reader

Abstract

The present invention relates to a thermal storage electric heating power control decision method and system based on Pso-CNN, and the method comprises: step S1: collecting all thermal storage electric heating load data in the area, clustering them, and obtaining clustering results; step S2: establishing a dynamic mathematical model of thermal storage electric heating temperature and thermal storage electric heating energy storage device, and determining model parameters; step S3: constructing an objective function to satisfy the dynamic mathematical model and its constraints; initializing the operation strategy of each clustering cluster, and calculating the objective function value based on the particle swarm optimization algorithm, iteratively optimizing the thermal storage electric heating operation strategy of each clustering cluster; step S4: constructing a state information mapping curve graph, inputting a convolutional neural network for learning, and predicting the thermal storage electric heating operation strategy of each clustering cluster in each future period. The method provided by the present invention is based on historical data, uses the Pso algorithm to solve the optimal operation strategy of thermal storage electric heating, and uses CNN to predict future power control strategies.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of electricity control decision-making for thermal storage electric heating, and in particular to a method and system for electricity control decision-making for thermal storage electric heating based on Pso-CNN. Background Art

[0002] At present, with the further reform of the power system, the infrastructure and policies such as "coal to electricity" are gradually improved, and the interaction between the user-side load and the grid-side is enhanced. As a new type of load, thermal storage electric heating can use energy storage devices to store electricity when the load demand is small at night and the wind and solar output is large, and use energy storage devices to release electricity when the wind and solar load demand is large during the day and the split-light output is relatively insufficient, thereby achieving the purpose of peak shaving and valley filling, smoothing load fluctuations, and promoting the consumption of new energy.

[0003] Thermal storage electric heating is different from other energy storage devices. The first thing to do when participating in regulation is to meet the user's comfort level. Secondly, because it is distributed in various user households, it is also necessary to consider the size of its regulation capacity and the issue of optimizing its control strategy. In addition, factors such as the uncertainty of wind and solar power output and the volatility of the load side must also be considered.

[0004] Under actual working conditions, the accuracy of power forecasting and load forecasting decreases as the forecasting time increases. Therefore, the formulation of the day-ahead control strategy based on the forecast data will have a certain deviation from the actual situation. The two-stage optimization method of the day-ahead and day-intraday requires a lot of calculations and cannot meet the real-time nature of the control strategy. In recent years, many scholars have adopted data-driven deep reinforcement learning methods to solve complex optimization problems, and do not require prior knowledge of the system to obtain the optimal charging and discharging plan based on the actual state. However, thermal storage electric heating is distributed within the user's home, with a large number and different types. When dealing with such problems, the deep reinforcement learning method will produce a dimensional explosion, resulting in the inability to solve such optimization problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for controlling power consumption of thermal storage electric heating based on Pso-CNN.

[0006] The thermal storage electric heating electricity consumption control decision method provided by the present invention is applicable to the scenario of a high proportion of new energy grid-connected. All thermal storage electric heating in a certain area is regarded as an adjustable load, and the remaining loads (such as televisions, refrigerators, etc.) are regarded as background loads. When the wind and solar power output, that is, the new energy output, is in excess of the power demanded by users, electric heating absorbs energy from the power grid, and uses a heat storage device to store electric energy in addition to normal heating. When the wind and solar power output is insufficient compared to the power demanded by users, electric heating stops absorbing energy from the power grid and uses a heat storage device to release electric energy. This method promotes the consumption of new energy and smoothes load fluctuations while ensuring the comfort of users in winter. At the same time, users participating in the regulation will also reduce electricity costs.

[0007] The technical solution of the present invention is: a method for controlling power consumption of thermal storage electric heating based on Pso-CNN, comprising:

[0008] Step S1: Collect all thermal storage electric heating load data in the area, determine the clustering characteristics of the thermal electric heating load, cluster the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtain a clustering result;

[0009] Step S2: establishing a thermal storage electric heating temperature dynamic model and a thermal storage electric heating energy storage device dynamic mathematical model, and determining the model parameters of each cluster according to the clustering results;

[0010] Step S3: construct an objective function to satisfy the dynamic model of thermal storage electric heating temperature and the dynamic mathematical model of the thermal storage electric heating energy storage device and their constraints; initialize the operation strategy of each clustering cluster, calculate the objective function value based on the particle swarm optimization algorithm, and iteratively optimize the thermal storage electric heating operation strategy of each clustering cluster;

[0011] Step S4: construct curve graphs of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data respectively, and convert them into corresponding images, input the images into a convolutional neural network, and predict the thermal storage electric heating operation strategies of each of the clustering clusters in each future time period.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] 1. The present invention discloses a power consumption control decision method for thermal storage electric heating based on Pso-CNN. In order to solve the problem of excessive dimension of residential load control, it is proposed to cluster thermal storage electric heating with similar adjustment capabilities according to different characteristics, determine the model parameters of each cluster according to the clustering results, and adopt the same power consumption control strategy for the load within each cluster.

[0014] 2. The present invention uses user comfort as a constraint condition and is based on an optimization method in which all states are known to solve the optimal operation strategy of thermal storage electric heating.

[0015] 3. The present invention maps the power grid status and cluster thermal storage electric heating load status information of each time period onto an image, and then provides the image to a 3-channel convolutional neural network for learning the control strategy; in addition, the present invention also introduces a prediction error, which solves the problem that the prediction accuracy decreases as the prediction time becomes longer during the current power prediction and load prediction, and is more suitable for actual scenarios; the present invention starts from the perspective of maximizing the consumption of new energy, and uses the electricity bills saved by users as the evaluation indicator of the future electricity consumption control strategy predicted by the convolutional neural network, which solves the problem that the demand response method based on electricity price information will generate new load peaks during valley periods. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of a method for controlling power consumption of thermal storage electric heating based on Pso-CNN in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a mapping picture of new energy output and background load demand in an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram of a thermal storage electric heating control capability mapping image in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a thermal storage electric heating state mapping picture in an embodiment of the present invention;

[0020] Figure 5 Schematic diagram of an input picture of a CNN convolutional neural network in an embodiment of the present invention;

[0021] Figure 6 The present invention is a block diagram of a thermal storage electric heating power control decision system based on Pso-CNN in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] The present invention provides a thermal storage electric heating electricity consumption control decision method based on Pso-CNN. Based on historical data, the particle swarm optimization Pso algorithm is used to solve the optimal operation strategy of thermal storage electric heating, and the CNN convolutional neural network is used to learn and predict the future electricity consumption control strategy.

[0023] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below through specific implementations and in conjunction with the accompanying drawings.

[0024] Embodiment 1

[0025] like Figure 1As shown, a method for controlling power consumption of thermal storage electric heating based on Pso-CNN provided by an embodiment of the present invention comprises the following steps:

[0026] Step S1: Collect all thermal storage electric heating load data in the area, determine the clustering characteristics of thermal electric heating loads, cluster the thermal storage electric heating loads according to the clustering characteristics based on the K-means algorithm, and obtain clustering results;

[0027] Step S2: Establish a dynamic model of thermal storage electric heating temperature and a dynamic mathematical model of the thermal storage electric heating energy storage device, and determine the model parameters of each cluster according to the clustering results;

[0028] Step S3: construct an objective function to satisfy the dynamic model of thermal storage electric heating temperature and the dynamic mathematical model of the thermal storage electric heating energy storage device and their constraints; initialize the operation strategy of each cluster, calculate the objective function value based on the particle swarm optimization algorithm, and iteratively optimize the thermal storage electric heating operation strategy of each cluster;

[0029] Step S4: Construct curve graphs of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data, respectively, and convert them into corresponding images. Input the images into the convolutional neural network to predict the thermal storage electric heating operation strategies of each clustering cluster in each future time period.

[0030] In one embodiment, the above step S1: collecting all thermal storage electric heating load data in the area, determining the clustering characteristics of the thermal electric heating load, clustering the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtaining the clustering cluster results, specifically includes:

[0031] Step S11: Collect all thermal storage electric heating load data in the area, and determine the clustering characteristics of the thermal storage electric heating load as follows:

[0032] Energy storage capacity of thermal storage electric heating tor , unit: kw·h;

[0033] Rated power P of thermal storage element for electric heating in , unit kw;

[0034] Thermal storage electric heating element discharge rated power P out , unit kw;

[0035] Thermal storage electric heating electric heat conversion efficiency η;

[0036] Thermal storage electric heating rated power P, unit kw;

[0037] Thermal storage electric heating energy storage element efficiency ηin ;

[0038] Step S12: Randomly initialize the cluster center and the number of center points of the thermal storage electric heating load data;

[0039] Step S13: Based on the K-means algorithm, according to the clustering characteristics, the Euclidean distance from each point to the cluster center is calculated to cluster the thermal storage electric heating load;

[0040] Step S14: Calculate the evaluation index a according to formula (1) to evaluate the clustering result;

[0041]

[0042] Among them, n is the number of cluster centers, C i is the number of thermal storage electric heating loads belonging to the i-th category in the clustering results, R i is the Euclidean distance from the data point farthest from the cluster center in the i-th class to the cluster center, and m represents the number of selected clustering features;

[0043] Step S15: Based on the elimination rule in the genetic algorithm, the clusters are sorted from large to small according to fitness, and those with fitness less than the threshold are eliminated. A new cluster is generated from the two clusters with the highest fitness, and a new cluster center and the number of center points are randomly selected;

[0044] Step S16: Determine whether the preset number of iterations has been reached. If not, repeat steps S13 to S15. Otherwise, the iteration ends and the clustering result that maximizes the evaluation index a is selected as the clustering result.

[0045] After the above step S1, the thermal storage electric heating loads in the area are analyzed and clustered, and the thermal storage electric heating loads with similar adjustment capabilities are classified into the same cluster. A unified thermal storage electric heating operation strategy will be adopted for each cluster.

[0046] In one embodiment, the above step S2: establishing a thermal storage electric heating temperature dynamic model and a thermal storage electric heating energy storage device dynamic mathematical model, and determining the model parameters of each cluster according to the clustering results, specifically includes:

[0047] Step S21: Establish a dynamic mathematical model of thermal storage electric heating temperature, as shown in formula (2):

[0048]

[0049] Among them, T in (t) is the user's indoor temperature at time t, T out(t) is the outdoor temperature at time t, δ is the heat dissipation coefficient of the user's house, γ is the temperature rise coefficient of the user's house, and P(t) is the power absorbed by the thermal storage electric heating from the power grid at time t; P a (t) is the charging power of the thermal storage electric heating energy storage element at time t, P r (t) is the discharge power of the thermal storage electric heating energy storage element at time t; △t represents the length of a time period. For example, if a 24-hour day is divided into 96 time periods, then △t = 24*60 / 96; the λ value is 0 or 1, indicating that the thermal storage electric energy storage device can only take one action in the same time period, namely, discharge or charge action;

[0050] Step S22: Establish a dynamic mathematical model of the thermal storage electric heating energy storage device, as shown in formula (3):

[0051]

[0052] Among them, S(t) is the energy stored in the thermal storage electric heating energy storage element at time t, and μ is the heat dissipation coefficient of the thermal storage electric heating energy storage device itself;

[0053] Step S23: According to the clustering results, determine the mathematical model parameters of each cluster, as shown in formula (4):

[0054]

[0055] Among them, q is the mathematical model parameter of the cluster, b is the number of thermal storage electric heating in the cluster, Q i are the model parameters of the thermal storage electric heating load of the i-th cluster.

[0056] In one embodiment, the above step S3: constructing an objective function to satisfy the dynamic model of thermal storage electric heating temperature and the dynamic mathematical model of the thermal storage electric heating energy storage device and their constraints; initializing the operation strategies of each clustering cluster, and calculating the objective function value based on the particle swarm optimization algorithm, iteratively optimizing the thermal storage electric heating operation strategy of each clustering cluster, specifically including:

[0057] Step S31: Divide a day into k time periods, and based on historical data, obtain the difference P between the output of new energy and the demand for thermal storage electric heating load in each time period. val , as shown in formula (5):

[0058] P val (i) = P new (i)-P load (i) (5)

[0059] Where P new (i) is the output of renewable energy in the i-th period, P load(i) is the background load demand in the i-th period;

[0060] According to P val At the value of time period i, the initialization thermal storage electric heating operation strategy is as follows:

[0061] When new energy overflows, that is, P val (i)>0, let P(i)>0, P a (i) = 0.5P in , P r (i) = 0;

[0062] When new energy is lacking, that is, P val (i)<0, let P(i)=0, P a (i) = 0, P r (i) = 0.5P out ;

[0063] Where P(i) is the power absorbed by the thermal storage electric heating from the power grid in the i-th period, P a (i) is the charging power of the thermal storage electric heating energy storage device in the i-th period, P r (i) is the discharge power of the thermal storage electric heating energy storage device in the i-th period;

[0064] In the embodiment of the present invention, a day is divided into 96 time periods, that is, k=96;

[0065] Step S32: Set the constraints of the optimization formulas (2) to (3), as shown in formulas (6) to (10):

[0066] 0.1S tor <S(t)<0.9S tor (6)

[0067] 0≤P a (t) <P in (7)

[0068] 0≤P r (t) <P out (8)

[0069] T set -△t≤T in (t) <T set +△t (9)

[0070] λ=0 or λ=1 (10)

[0071] Among them, T set is the indoor temperature value set by the user;

[0072] Step S33: According to the electricity consumption strategy of each cluster thermal storage electric heating in each time period, the total demand of the background load in each time period is updated, as shown in formula (11):

[0073] P load (i) = P load (i)+P(i) (11)

[0074] Step S34: setting the objective function in the peak shaving and valley filling scenario, as shown in formula (12):

[0075]

[0076] Among them, P load (i) is the power demand of the background load in the i-th time period, and k is the number of time periods;

[0077] The objective function in the scenario of promoting new energy consumption is set as shown in formula (13):

[0078]

[0079] Among them, P new (i) is the output of new energy in the i-th time period;

[0080] Step S35: According to the electricity consumption strategy of each cluster thermal storage electric heating in each time period, calculate the indoor temperature T of each user in each stage according to formulas (2) and (3): in And the thermal storage electric heating energy storage capacity S tor If constraints (6) and (9) are not met, penalty terms M1 and M2 are added to the objective function, as shown in formulas (14) to (15):

[0081]

[0082]

[0083] Step S36: solving the thermal storage electric heating operation strategy based on the particle swarm optimization algorithm, including the following steps:

[0084] 1) Initialize particles P(i), P a (i) P r (i) Obtain the initial thermal storage electric heating operation strategy and generate the initial search speeds V and V r 、V a ;

[0085] 2) Calculate the objective function value of each cluster to obtain the initial historical optimal point and the global optimal point; the historical optimal point is the optimal point within each particle group, and the global optimal point is the optimal point for all particle groups;

[0086] 3) Update the velocity and position of each particle, as shown in formulas (16) to (17):

[0087] P=P+V (16)

[0088] P=ω*P+c1*rand()*(P best -P)+c2*rand()*(G best -p) (17)

[0089] Where P represents P(i), P a (i) P r (i) V represents V, V r 、V a ; ω is the inertia factor, c1 and c2 are learning factors, P best is the historical optimal point of the particle, G best is the global optimal point; rand() is a random function;

[0090] 4) Calculate the objective function value of each particle and update the historical optimal point of each particle and the global optimal point;

[0091] 5) Determine whether the number of iterations has been reached. If the maximum number of iterations has been reached, stop the iteration and output the global optimal point as the thermal storage electric heating operation strategy; if the maximum number of iterations has not been reached, repeat steps 3) to 5).

[0092] After the above steps, the embodiment of the present invention regards the problem of solving the optimal operation strategy of thermal storage electric heating as an optimization problem rather than an interaction process with the environment based on historical data. Therefore, a particle swarm optimization algorithm is used instead of a reinforcement learning algorithm, thereby simplifying the solution process. The operation strategy of each cluster thermal storage electric heating in each period includes three parts, namely [P(i), P a (i), P r (i)], which is input into the CNN network for learning, and the thermal storage type in the future period can be predicted.

[0093] Optimal strategy for electric heating operation.

[0094] In one embodiment, the above step S4: respectively constructs a graph of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data, and converts them into corresponding images, inputs the images into a convolutional neural network, and predicts the thermal storage electric heating operation strategy of each clustering cluster in each future time period, specifically including:

[0095] S41: constructing graphs of mapping of new energy output and background load demand, mapping of thermal storage electric heating regulation capability, and mapping of thermal storage electric heating status data respectively, and converting the graphs into pictures, and dividing them into training sets and validation sets according to preset ratios, including the following steps:

[0096] 1) Construct a curve diagram of the mapping between the new energy output and the background load demand, and draw the curve P according to the following formula (18): val :

[0097] P val =P new -P load (18)

[0098] In order to fully reflect all state information in the embodiment of the present invention, the image should contain the new energy output data and background load demand data in the next 24 hours, that is, 96 time periods, and the image size is set to 96*96. The image adopts a binary method, that is, when the curve passes through the pixel point, the weight of the pixel point is 255, if the curve does not pass through the pixel point, the pixel weight is 0, and considering the problem that the prediction accuracy decreases with the increase of prediction time, the embodiment of the present invention introduces an error function f error , f error It is a randomly generated function with a mean of α and a variance of δ, where δ is a constant and α increases with the increase of the prediction time, generating two new curves As shown in formulas (19) to (20):

[0099]

[0100]

[0101] curve Enclose an area, and the weight of the pixels in the area is set to 127;

[0102] like Figure 2 As shown, it is a schematic diagram of the mapping picture of new energy output and background load demand.

[0103] 2) The regulation capability of thermal storage electric heating is related to its cluster category, so the code corresponding to the category is set as the weight of the image, such as Figure 3 shown.

[0104] 3) Thermal storage electric heating status data mainly includes user indoor temperature T in , User set temperature T set , outdoor temperature T out , the state of charge Soc of the energy storage element, divide the 96*96 image into four areas of 48*48, and map one piece of information to each area, such as Figure 4As shown;

[0105] 4) Divide the above images into a training set and a validation set in a ratio of 7:3 on a daily basis.

[0106] S42: Use the training set to train the convolutional neural network, and optimize the convolutional neural network based on the genetic algorithm to predict the thermal storage electric heating operation strategy of each cluster in each future period, and verify the output results on the verification set based on the additional consumption of new energy and the electricity bill saved by the user as indicators, including the following steps:

[0107] 1) On the training set, the three mapping images are superimposed to obtain Figure 5 The input picture is shown in the figure. A 3-channel CNN convolutional neural network is input. The CNN convolutional neural network structure is optimized by genetic algorithm. The optimized hyperparameters include the number of convolution kernels, activation function type, number of hidden layers, and number of neurons in the hidden layer. The optimized objective function is the loss value.

[0108] 2) Evaluate the prediction results of the CNN network output based on the validation set, where the evaluation indicators include: the additional consumption of new energy W new And the electricity cost saved by the user side M elec , as shown in formulas (21) to (22):

[0109] W new =W reg -W nor (twenty one)

[0110]

[0111] Among them, W reg represents the amount of new energy consumed after the thermal storage electric heating participates in the regulation, W nor Indicates that thermal storage electric heating does not participate in regulating the consumption of new energy, M nor (i) represents the electricity cost required for the thermal storage electric heating cluster i not participating in the regulation, M reg (i) represents the electricity cost required for the thermal storage electric heating cluster i to participate in the regulation.

[0112] The present invention discloses a thermal storage electric heating power control decision method based on Pso-CNN. In order to solve the problem of excessive dimensionality of residential load control, it is proposed to cluster thermal storage electric heating with similar adjustment capabilities according to different characteristics, determine the model parameters of each cluster cluster according to the clustering results, and adopt the same power control strategy for the load within each cluster cluster. The present invention solves the optimal operation strategy of thermal storage electric heating based on the user comfort as a constraint condition and based on the optimization method with all known states. The present invention maps the power grid state and cluster thermal storage electric heating load state information of each time period to an image, and then provides the image to a 3-channel convolutional neural network for learning control strategies; in addition, the present invention also introduces prediction errors, solves the problem that the prediction accuracy will decrease as the prediction time becomes longer when performing power prediction and load prediction, and is more suitable for actual scenarios; the present invention takes the user's saved electricity bill as the evaluation indicator of the future power control strategy predicted by the convolutional neural network from the perspective of maximizing the consumption of new energy, and solves the problem that the demand response method based on electricity price information will generate new load peaks during valley periods.

[0113] Embodiment 2

[0114] like Figure 6 As shown, the embodiment of the present invention provides a thermal storage electric heating power control decision system based on Pso-CNN, including the following modules:

[0115] The load clustering module 51 is used to collect all thermal storage electric heating load data in the area, determine the clustering characteristics of the thermal electric heating load, cluster the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtain the clustering result;

[0116] Establishing a dynamic model module 52, for establishing a thermal storage electric heating temperature dynamic model and a thermal storage electric heating energy storage device dynamic mathematical model, and determining the model parameters of each cluster according to the clustering results;

[0117] The operation strategy acquisition module 53 is used to construct an objective function to satisfy the dynamic model of thermal storage electric heating temperature and the dynamic mathematical model of the thermal storage electric heating energy storage device and their constraints; initialize the operation strategies of each cluster, calculate the objective function value based on the particle swarm optimization algorithm, and iteratively optimize the thermal storage electric heating operation strategies of each cluster;

[0118] The module 54 for predicting future operation strategies is used to construct graphs of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data, and convert them into corresponding images. The images are input into a convolutional neural network to predict the thermal storage electric heating operation strategies of each clustering cluster in each future time period.

[0119] The above embodiments are provided only for the purpose of describing the present invention, and are not intended to limit the scope of the present invention. The scope of the present invention is defined by the appended claims. Various equivalent substitutions and modifications made without departing from the spirit and principles of the present invention should all be included within the scope of the present invention.

Claims

1. A method for controlling power consumption of thermal storage electric heating based on Pso-CNN, characterized in that: include: Step S1: Collect all thermal storage electric heating load data in the area, determine the clustering characteristics of the thermal electric heating load, cluster the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtain a clustering result; Step S2: Establish a thermal storage electric heating temperature dynamic model and a thermal storage electric heating energy storage device dynamic mathematical model, and determine the model parameters of each cluster according to the clustering results, specifically including: Step S21: Establish a dynamic mathematical model of thermal storage electric heating temperature, as shown in formula (2): (2) in, for The user's indoor temperature at all times, for Outdoor temperature at any time, is the heat dissipation coefficient of the user's house, is the temperature rise coefficient of the user's house, is the power absorbed by the thermal storage electric heating from the grid at time t; Energy storage element for thermal electric heating Charging power at all times, Energy storage element for thermal electric heating Discharge power at all times; Represents the length of a time period; The value is 0 or 1, indicating that the thermal energy storage device can only take one action in the same time period; Step S22: Establish a dynamic mathematical model of the thermal storage electric heating energy storage device, as shown in formula (3): (3) in, for The energy stored in the thermal storage electric heating energy storage element at all times, is the heat dissipation coefficient of the thermal storage electric heating energy storage device itself; Step S23: According to the clustering results, the mathematical model parameters of each cluster are determined, as shown in formula (4): (4) Wherein, q is the mathematical model parameter of the cluster, n is the number of thermal storage electric heating in the cluster, Q i is the model parameter of the thermal storage electric heating load of the i-th cluster; Step S3: construct an objective function to satisfy the dynamic model of thermal storage electric heating temperature and the dynamic mathematical model of the thermal storage electric heating energy storage device and their constraints; initialize the operation strategy of each clustering cluster, calculate the objective function value based on the particle swarm optimization algorithm, and iteratively optimize the thermal storage electric heating operation strategy of each clustering cluster, specifically including: Step S31: Divide a day into k time periods, and obtain the difference between the output of new energy and the demand of the thermal storage electric heating load in each time period based on historical data. , as shown in formula (5): (5) in For the The output of new energy during the period, For the Background load demand during the period; according to In the period The value of initializes the thermal storage electric heating operation strategy as follows: like ,make , , ; like ,make , , ; in, No. The power absorbed from the power grid by the period thermal storage electric heating, For the The charging power of the thermal storage electric heating energy storage device during the period, For the The discharge power of the thermal storage electric heating energy storage device during the time period; Step S32: Set the constraints of the optimization formulas (2) to (3), as shown in formulas (6) to (10): (6) (7) (8) (9) =0 or =1 (10) in, is the indoor temperature value set by the user; Step S33: According to the electricity consumption strategy of each cluster thermal storage electric heating in each time period, the total demand of the background load in each time period is updated, as shown in formula (11): (11) Step S34: setting the objective function in the peak shaving and valley filling scenario, as shown in formula (12): (12) in, is the power demand of the background load in the i-th time period, k is the number of time periods; The objective function in the scenario of promoting new energy consumption is set as shown in formula (13): (13) in, is the output of new energy in the i-th time period; Step S35: According to the electricity consumption strategy of each cluster thermal storage electric heating in each time period, calculate the indoor temperature T of the user in each stage according to formulas (2) and (3): in And the thermal storage electric heating energy storage capacity S tor , if the constraints (6) and (9) are not met, penalty terms M1 and M2 are added to the objective function, as shown in formulas (14) to (15): (14) (15) Step S36: solving the thermal storage electric heating operation strategy based on the particle swarm optimization algorithm, including the following steps: 1) Initialization , , Get the initial thermal storage electric heating operation strategy and generate the initial search speed , , ; 2) Calculate the objective function value of each cluster to obtain the initial historical optimal point and the global optimal point; 3) Update the velocity and position of each particle, as shown in formulas (16) to (17): (16) (17) in, represent , , , represent , , ; is the inertia factor, , is the learning factor, is the historical optimal point of the particle, is the global optimal point; is a random function; 4) Calculate the objective function value of each particle and update the historical optimal point of each particle and the global optimal point; 5) Determine whether the number of iterations has been reached. If the maximum number of iterations has been reached, stop the iteration and output the global optimal point as the operation strategy of the thermal storage electric heating; if the maximum number of iterations has not been reached, repeat steps 3) to 5); Step S4: constructing graphs of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data respectively, and converting them into corresponding images, inputting the images into a convolutional neural network, and predicting the thermal storage electric heating operation strategies of each of the clusters in each future period, specifically including: S41: constructing graphs of mapping of new energy output and background load demand, mapping of thermal storage electric heating regulation capability, and mapping of thermal storage electric heating status data respectively, converting the graphs into pictures, and dividing them into a training set and a validation set according to a preset ratio; S42: Using the training set to train a convolutional neural network, and optimizing the convolutional neural network based on a genetic algorithm, predicting the thermal storage electric heating operation strategy of each of the clusters in each future time period, and verifying the output results on the verification set based on the additional consumption of new energy and the electricity bill saved by the user as indicators.

2. The method for controlling power consumption of thermal storage electric heating based on Pso-CNN according to claim 1 is characterized in that: The step S1: collecting all thermal storage electric heating load data in the area, determining the clustering characteristics of the thermal electric heating load, clustering the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtaining the clustering result, specifically includes: Step S11: Collect all thermal storage electric heating load data in the area, and determine the clustering characteristics of the thermal storage electric heating load as follows: Energy storage capacity of thermal storage electric heating , unit: kw·h; Rated power of charging of thermal storage element for thermal storage type electric heating , unit kw; Rated power of thermal storage element for electric heating , unit kw; Thermal storage electric heating electric heat conversion efficiency ; Thermal storage electric heating rated power , unit kw; Thermal storage electric heating energy storage element efficiency ; Step S12: randomly initializing the cluster center and the number of center points of the thermal storage electric heating load data; Step S13: Based on the K-means algorithm, according to the clustering characteristics, the Euclidean distance from each point to the cluster center is calculated to cluster the thermal storage electric heating load; Step S14: Calculate the evaluation index a according to formula (1) to evaluate the clustering result; (1) Where n is the number of cluster centers. The clustering result belongs to The number of thermal storage electric heating loads of the type, For the The Euclidean distance from the data point farthest from the cluster center to the cluster center in the cluster, Indicates the number of selected clustering features; Step S15: Based on the elimination rule in the genetic algorithm, the clusters are sorted from large to small according to fitness, and those with fitness less than a threshold are eliminated. A new cluster is generated from the two clusters with the highest fitness, and a new cluster center and the number of center points are randomly selected; Step S16: Determine whether the preset number of iterations has been reached. If not, repeat steps S13 to S15. Otherwise, the iteration ends and the clustering result that maximizes the evaluation index a is selected as the clustering result.

3. A thermal storage electric heating power control decision system based on Pso-CNN, used to implement the thermal storage electric heating power control decision method based on Pso-CNN as claimed in claim 1, characterized in that: Includes the following modules: A load clustering module is used to collect all thermal storage electric heating load data in the area, determine the clustering characteristics of the thermal electric heating load, cluster the thermal storage electric heating load according to the clustering characteristics based on the K-means algorithm, and obtain a clustering result; Establish a dynamic model module for establishing a dynamic model of thermal storage electric heating temperature and a dynamic mathematical model of a thermal storage electric heating energy storage device, and determine the model parameters of each cluster according to the clustering results; Obtaining an operation strategy module, used to construct an objective function so that it satisfies the thermal storage electric heating temperature dynamic model and the thermal storage electric heating energy storage device dynamic mathematical model and its constraints; initializing the operation strategy of each clustering cluster, and calculating the objective function value based on the particle swarm optimization algorithm, and iteratively optimizing the thermal storage electric heating operation strategy of each clustering cluster; The module for predicting future operation strategies is used to construct curve graphs of the mapping of new energy output and background load demand, the mapping of thermal storage electric heating control capability, and the mapping of thermal storage electric heating status data, and convert them into corresponding images. The images are input into a convolutional neural network to predict the thermal storage electric heating operation strategies of each of the clustering clusters in each future time period.