A low-carbon operation optimization method and management system for a household energy system based on PVT photovoltaic thermal energy

By establishing a multi-objective planning model in PVT photovoltaic photothermal system and using LSTM prediction model, combining particle swarm algorithm and TOPSIS algorithm, the problems of energy fluctuations and insufficient prediction in the new rural residential energy system are solved, and low-carbon operation and energy are achieved Improved efficiency.

CN118657067BActive Publication Date: 2024-11-26NANJING NORMAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411141441.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-26
Estimated Expiration
2044-08-20

AI Technical Summary

Technical Problem

New rural residential energy systems equipped with PVT systems have large fluctuations in energy electricity/thermal loads and photovoltaic photothermal output, lack of load and photovoltaic photothermal prediction models, resulting in energy losses, making it difficult for users to make decisions on minimum carbon emissions and operating costs, resulting in Energy wasted.

Method used

Using the low-carbon operation optimization method of household energy systems based on PVT photovoltaic solar thermal, a multi-objective planning model is established through mathematical modeling and particle swarm algorithm, and optimizing operating costs and carbon emissions, combining LSTM prediction model to predict energy demand to achieve accurate Adjust and use the TOPSIS algorithm to select the optimal operating solution.

Benefits of technology

Through precise prediction and optimization control, energy waste is reduced, energy utilization efficiency is improved, operating costs and carbon emissions are reduced, and the purpose of energy conservation and emission reduction is achieved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118657067B_ABST
    Figure CN118657067B_ABST
Patent Text Reader

Abstract

The present invention discloses a low-carbon operation optimization method and management system for a household energy system based on PVT photovoltaic and thermal energy. The optimization method comprises the following steps: selecting the operation cost and carbon emission of the system as optimization targets; taking the actual power generation of PVT, the actual heat supply of PVT, the transmission power of the battery, the transmission power of the power grid, the heat transmission of the heat storage tank and the heat transmission of the natural gas grid as decision variables, mathematically modeling the battery and the heat storage tank; taking the power balance constraint, the transmission power constraint of the main grid, the PVT power constraint, the battery constraint and the heat storage tank constraint as constraint conditions, establishing a multi-objective planning model with the operation cost and carbon emission as optimization targets; using an improved particle swarm algorithm to optimize the multi-objective planning model and obtain the optimal solution set; using the superior and inferior solution distance method to process the optimal solution set and obtain the optimal operation plan. The present invention can accurately adjust the energy supply and improve the energy utilization efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of energy-carbon management systems, and in particular to a low-carbon operation optimization method and management system for a household energy system based on PVT photovoltaic thermal energy. Background Art

[0002] Compared with traditional power generation systems, home energy systems have the characteristics of less environmental pollution and low operating investment costs, and can improve the flexibility and reliability of the power grid. Solar photovoltaic thermal (PVT) combines power generation and heating in one component, which not only improves the efficiency of photovoltaic power generation, but also provides heating and domestic hot water. It is a product that increases value by improving efficiency. Its core is to recover and utilize excess heat energy while generating solar photovoltaic power. This not only has a cooling effect on the battery, but also can improve the power generation efficiency and life, greatly improve the comprehensive utilization efficiency of solar energy, and reduce the cost of separate power and heat supply. In addition, the heat and electricity generated by solar photovoltaic thermal PVT is at least three times that of photovoltaic power generation of the same area. As a solution suitable for near-zero energy buildings, it is gaining more and more attention. However, the new rural residential energy system equipped with PVT system currently has the following problems: (1) The building energy electricity / heat load and photovoltaic thermal output fluctuate greatly, and there is a lack of load and photovoltaic thermal prediction model, resulting in large energy loss. (2) Most users control “source, storage, and network” based on experience alone, and often fail to make the best decisions to minimize carbon emissions and operating costs. This also results in energy waste, which goes against the user’s original intention. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a low-carbon operation optimization method and management system for a home energy system based on PVT photovoltaic thermal energy, so as to reduce the consumption of traditional energy in the residence and increase the overall benefits, and ultimately achieve the purpose of energy conservation and emission reduction.

[0004] Technical solution: A low-carbon operation optimization method for a household energy system based on PVT photovoltaic thermal energy, comprising the following steps:

[0005] S1, select the system's operating cost and carbon emissions as optimization targets;

[0006] S2, with the actual power generation of PVT, the actual heat supply of PVT, the battery transmission power, the grid transmission power, the heat transmitted by the hot water storage tank and the heat transmitted by the natural gas grid as decision variables, mathematical modeling of the battery and the hot water storage tank is carried out;

[0007] S3, with power balance constraint, main grid transmission power constraint, PVT power constraint, battery constraint and hot water storage tank constraint as constraint conditions, establish a multi-objective planning model with operation cost and carbon emission as optimization objectives;

[0008] S4, using the improved particle swarm algorithm to optimize the multi-objective programming model and obtain the optimal solution set;

[0009] S5, the obtained optimal solution set is processed by using the superior and inferior solution distance method to obtain the optimal operation plan.

[0010] Further, in step S2, mathematical modeling is performed on the battery and the hot water storage tank:

[0011] , , where represents the total amount of electrical energy in the household battery at time t; , Respectively represent the charging and discharging efficiency of household batteries; Indicates the charging or discharging power of the battery. Represents charging, when Represents discharge; represents the total amount of heat energy in the household water storage tank at time t; , They represent the heat storage and heat release efficiencies of the household water storage tank respectively; It represents the heat storage or heat release power of the hot water storage tank at time t. represents heat storage, represents heat release; represents the heat dissipation loss of the household water storage tank at time t; where the heat dissipation loss of the water storage tank is The calculation formula is as follows: , where represents the surface area of ​​the jth hot water storage tank; Indicates the heat loss coefficient of the hot water storage tank, in units of ; It indicates the temperature of the working fluid in the hot water storage tank at that moment. It indicates the ambient temperature at that moment, in °C; m indicates the total number of water tanks.

[0012] In step S3, the mathematical models of the power balance constraint, main network transmission power constraint, PVT power constraint, battery constraint and hot water storage tank constraint are as follows:

[0013] A1, power balance constraint: ,

[0014] A2, main network transmission power constraints: ,

[0015] A3, PVT power constraints: , ,

[0016] A4, Battery constraints: , ,

[0017] A5, hot water storage tank constraints: , , where Indicates the actual power generation of photovoltaic cells, Indicates the amount of electricity purchased from the main network. Indicates the charge and discharge capacity of the battery. represents the predicted value of household electricity load, Indicates the power consumption of the circulating water pump. Indicates the power consumption of the gas water heater. Indicates the actual heat supply of the collector plate. Indicates the calorific value of gas purchased from the main grid, Indicates the charging and discharging heat of the hot water storage tank. It represents the predicted value of household heat load in kWh; , They represent the upper and lower limits of the power transmission of the power grid respectively; , They represent the upper and lower limits of the natural gas pipeline flow respectively; q represents the calorific value of natural gas; represents the predicted power generation of the photovoltaic cell at time t, kWh; It represents the predicted heating supply of the collector at time t; Indicates the state of charge of the battery at time t; represents the total amount of electrical energy in the household battery at time t; Indicates the rated capacity of the battery; , Respectively represent the upper and lower limits of the battery's power storage capacity; , Respectively represent the upper and lower limits of the battery transmission power; Indicates the heat storage state of the hot water storage tank at time t; represents the total amount of heat energy in the household water storage tank at time t; Indicates the rated capacity of the hot water storage tank; , They represent the upper and lower limits of the heat storage capacity of the hot water storage tank respectively; , They represent the upper and lower limits of the power transmitted by the hot water storage tank respectively.

[0018] In step S3, the multi-objective programming model F is: ,in, For operating costs, for carbon emissions;

[0019] , where represents the electricity purchase price at time t, represents the operation and maintenance cost of the battery at time t, represents the amount of electricity purchased at time t, Indicates the charge and discharge amount of the battery at time t, in kWh; represents the natural gas purchase price at time t, represents the operation and maintenance cost of the hot water storage tank at time t, represents the operation and maintenance cost of the circulating water pump at time t, in units of RMB / m 3 ; represents the amount of heat purchased from the heating network at time t, represents the heat charge and discharge of the hot water storage tank at time t, represents the actual heat supply of the PVT system at time t, in kWh;

[0020] , where represents the carbon emission conversion factor of electricity; represents the carbon emission conversion factor of natural gas; Represents the carbon emission conversion factor of renewable energy.

[0021] In step S4, the improved particle swarm algorithm is implemented as follows:

[0022] S51, set the total number of particles M, the particle dimension d, the inertia weight, the individual learning factor, the social learning factor, and the maximum number of iterations ST;

[0023] S52, with decision variables , , , , and Randomly generate particles with upper and lower limits and initialize the population position;

[0024] S53, based on running costs and carbon emissions Minimum, evaluate the fitness value F of each particle, obtain the individual optimal solution and the global optimal solution, and store the individual optimal solution and the global optimal solution in the archive library;

[0025] S54, setting upper and lower limits of battery charge and discharge, and upper and lower limits of heat storage and release of the hot water storage tank, and judging whether the battery reaches the upper and lower limits of charge and discharge, and whether the hot water storage tank reaches the upper and lower limits of heat storage and release at time t by the individual optimal solution;

[0026] S55, if the maximum number of iterations SI is reached, the Pareto frontier is output; if the maximum number of iterations SI is not reached, the inertia weight and the learning factor are updated, and steps S53 and S54 are continued.

[0027] The superior and inferior solution distance method is used to process a series of non-inferior solutions provided by the Pareto frontier and select the optimal operation plan. The specific implementation steps are as follows:

[0028] S61, assuming that there are p solutions in the output Pareto solution set, the Pareto solution set of p solutions is used as the original data input, each solution represents an operation plan, and each operation plan has two indicator data, namely, operation cost and carbon emission;

[0029] The input variables are represented as a decision matrix 𝐷 with a dimension of p×2. Each element in the matrix 𝐷 Represents the value of the lth solution on the qth indicator:

[0030] , where l=1,2,...,p; q=1,2; represents the operating cost of the lth solution, represents the carbon emission of the lth option;

[0031] S62, use the vector normalization method to preprocess the data of each column of the decision matrix D to obtain the standardized decision matrix ;

[0032] S63, set the index weight vector , and the decision matrix The weighted processing is as follows:

[0033] , where Represents the standardized decision matrix Decision matrix after weighted processing;

[0034] S64, find the positive ideal solution and the negative ideal solution, the formula is as follows:

[0035] , where Represents the standardized decision matrix The positive ideal solution of the qth column; Represents the standardized decision matrix Negative ideal solution of the qth column; Represents the standardized decision matrix Elements in

[0036] S65, calculate the comprehensive evaluation index of each operation plan, the formula is as follows:

[0037] , where represents the comprehensive evaluation index of the lth scheme;

[0038] S66, the comprehensive evaluation index of each scheme Sort from large to small, and the solution corresponding to the largest number of indicators is the optimal solution.

[0039] A management system for low-carbon operation of a household energy system based on PVT photovoltaic thermal energy, used to implement any of the above-mentioned optimization methods, including a monitoring module, a prediction module and an optimization control module;

[0040] The monitoring module monitors, collects and stores information on all energy efficiency equipment in the residential building and the environmental conditions outside the residential building through intelligent information collection equipment and the Internet;

[0041] The prediction module is used to predict the user's electric heat load and the actual output value of photovoltaic thermal energy, including a PVT output prediction module and an electric heat load prediction module; the PVT output prediction module includes a first data receiving unit, a first data processing unit, a first model parameter adjustment unit and a PVT output prediction unit; the electric heat load prediction module includes a second data receiving unit, a second data processing unit, a second model parameter adjustment unit and an electric heat load prediction unit;

[0042] The first data receiving unit and the second data receiving unit are respectively used to receive the training data required for prediction from the monitoring module; the first data processing unit and the second data processing unit are respectively used for data processing and preprocessing of prediction; the first model parameter adjustment unit and the second model parameter adjustment unit are respectively used to adjust the parameters of the LSTM model to obtain the evaluation index of the prediction module under different parameters, and select the LSTM model with the best prediction effect for subsequent prediction; the PVT output prediction unit and the electric heat load prediction unit respectively predict the data of the next day through the established multi-feature input LSTM model;

[0043] The optimization control module takes the total operating cost of the energy storage unit and the carbon emissions of the entire system as the optimization targets, and the load characteristics, operation and maintenance costs, and peak and valley electricity prices of the energy storage unit as constraints, and uses the particle swarm algorithm to optimize the "source, storage, and network" for the next 24 hours through the PVT controller.

[0044] Further, the optimization control module includes a data input unit, an optimization unit and a control unit;

[0045] The data input unit is used to receive required data from the prediction module;

[0046] The optimization unit uses an improved particle swarm algorithm to optimize the electric and thermal "source, storage, and network" for the next 24 hours, and obtains the actual power generation of PVT, the actual heating supply of PVT, the battery transmission power, the grid transmission power, the hot water storage tank transmission power, and the natural gas network transmission power of the optimal solution for the next 24 hours, as well as the operating cost and carbon emission values;

[0047] The control unit is used to receive the optimization result of the optimization unit and control the "source, storage and network" of the electric heating at the hour of 24 hours in the future through the PVT controller and the Internet.

[0048] Furthermore, the PVT output prediction unit predicts the PVT output value of each period of the next 24 hours based on the PVT output value of the past month through the LSTM model; wherein the input variables of the LSTM model are the PVT output value at the hour of the previous 30 days and the local solar radiation value of the corresponding 30 days, and the output variables are the PVT power supply value at the hour of the 31st day and the PVT heating value at the hour;

[0049] The electric heating load prediction unit predicts the electric heating load for each period of the next 24 hours based on the household electric heating load value, the corresponding daily average temperature and the day type in the past month through the LSTM model; wherein the input variables of the LSTM model are the household electric load value, the household heat load value, the corresponding daily average temperature and the day type at the hour of the previous 30 days, and the output variables are the household electric load and heat load at the hour of the 31st day.

[0050] Compared with the prior art, the present invention has the following significant effects: 1. The system of the present invention is based on the LSTM prediction model. On the one hand, it uses the solar radiation intensity, PVT power generation and PVT heat supply as input variables to predict and output the PVT power generation at the next 24 time points. and PVT heating ; On the other hand, the electric heat load value, the average temperature of the day and the type of the day are used as input variables to predict and output the household electric heat load value at the next 24 time points. By predicting the energy demand for household electricity and heat consumption and the energy supply such as PVT power generation and heat supply, the energy supply can be adjusted more accurately, energy waste can be avoided, and energy utilization efficiency can be improved; 2. In the optimization method of the present invention, the operating cost and carbon emissions are used as optimization targets, and the particle swarm algorithm and TOPSIS algorithm are used to solve the optimal solution, so as to maximize the output of PVT photovoltaic thermal energy, use the "peak and valley" electricity price difference to reduce operating costs and reduce carbon dioxide emissions during operation, thereby reducing the consumption of traditional energy in residential areas and increasing overall benefits, and ultimately achieving the purpose of energy conservation and emission reduction. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a block diagram of the management system of the present invention;

[0052] Figure 2 Construct a schematic diagram for the neural network model - long short-term memory network model;

[0053] Figure 3 Implementing processes for energy system strategies;

[0054] Figure 4 It is the flow chart of MOPSO algorithm;

[0055] Figure 5 It is a schematic diagram of the hardware device of the home energy system;

[0056] Figure 6 This is a comparison chart of the predicted value and the actual value of the PVT output prediction module;

[0057] Figure 7 It is a comparison chart of the predicted value and the actual value of the electric and thermal load prediction module;

[0058] Figure 8 (a), (b), (c), and (d) are schematic diagrams of predicted values ​​of household electric load, household heat load, PVT power generation, and PVT heat supply, respectively;

[0059] Fig. 9 This is a schematic diagram of the training curve for the household electric heating load model;

[0060] Fig.10 This is a schematic diagram of the PVT output model training curve;

[0061] Fig.11 This is a schematic diagram of the Pareto frontier solution set;

[0062] Fig.12 (a), (b), and (c) are schematic diagrams of the results of PVT power generation, PVT heating, and energy storage system operation after optimization of the household energy system;

[0063] Fig.13 (a) and (b) are schematic diagrams of household electrical load and household thermal load after optimization of household energy system operation;

[0064] Fig.14 Schematic diagram of the amount of electricity purchased from the grid to operate the optimized home energy system. DETAILED DESCRIPTION

[0065] The present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0066] Figure 1 The energy system block diagram proposed by the present invention includes: a monitoring module 100, a prediction module 200 and an optimization control module 300.

[0067] The monitoring module 100 mainly monitors, collects and stores information on all energy-efficient equipment in residential buildings and environmental conditions outside residential buildings through intelligent information collection equipment and the Internet; the prediction module 200 is mainly used to predict uncertain factors such as user electric and thermal loads and photovoltaic thermal power, so as to prepare for the optimization control module 300 to better optimize the "source, storage, and network"; the optimization control module 300 optimizes and controls the "source-storage-network" for 24 hours in the next day through the controller, which can enable the household to maximize the consumption of renewable resources, reduce operating costs and complete the carbon emission reduction plan for the next day.

[0068] The PVT output prediction module 210 in the prediction module 200 is mainly used to predict the PVT output situation at 24 hours in the future day, and includes a first data receiving unit 211, a first data processing unit 212, a first model parameter adjustment unit 213 and a PVT output prediction unit 214.

[0069] The first data receiving unit 211 and the second data receiving unit 221 are respectively used to receive the training data required for prediction from the monitoring module 100; the first data processing unit 212 and the second data processing unit 222 are respectively used for data processing and preprocessing for prediction; the first model parameter adjustment unit 213 and the second model parameter adjustment unit 223 are respectively used to adjust the parameters of the LSTM model to obtain the evaluation index of the prediction module 200 under different parameters, and select the LSTM model with the best prediction effect for subsequent prediction; the PVT output prediction unit 214 and the electric heat load prediction unit 224 respectively predict the data of the next day through the established multi-feature input LSTM model;

[0070] Preferably, the first model parameter adjustment unit 213 uses a grid search algorithm to find the optimal solution of all combinations of the Cartesian product of the learning rate set and the number of hidden units, so as to obtain the most accurate prediction value in the PVT output prediction unit 214. Among them, the intelligent information collection equipment in the monitoring module 100 mainly includes intelligent meters and PVT controllers. The training data of the electric and thermal load prediction module 220 is mainly collected from the intelligent meters and PVT controllers by time period and date for 24h household electricity / heat load data every day in the past month, and stored in the monitoring module 100; the training data of the PVT output prediction module 210 is mainly collected from the PVT controller and the Internet by time and date for 24h PVT output conditions and their corresponding local meteorological conditions every day in the past month, and stored in the PVT output prediction module 210.

[0071] In the embodiment, a photovoltaic thermal output combined prediction model based on Spearman correlation analysis and LSTM is proposed. Spearman correlation analysis is a method in statistics for evaluating the degree of correlation between two variables. Based on the determination of the main meteorological factors affecting photovoltaic thermal by using the Spearman correlation analysis method, the PVT output of the next day is predicted using meteorological factors with a positive correlation (close to 1) and the corresponding PVT output, which can greatly improve the accuracy of the model. The correlations determined by the Spearman correlation analysis method are solar radiation intensity, rainfall, total cloud cover, relative humidity and temperature from high to low. Among them, the absolute value of the Spearman correlation coefficient between PVT output and solar radiation intensity is close to 1, while the absolute values ​​of the other correlation coefficients are close to 0, so the solar radiation intensity has a greater impact on the photovoltaic thermal output, and it is selected as the key input variable of the subsequent PVT output prediction unit 214.

[0072] The first data receiving unit 211 receives the PVT output and local solar radiation intensity values ​​required by the PVT output prediction unit 214 for the past month from the monitoring module 100 for prediction or parameter adjustment. However, when some missing values ​​appear in the data received from the monitoring module 100 due to external environmental reasons or intelligent information collection equipment failure, the present invention adopts a linear interpolation method to estimate the value of the new data point between the known data points.

[0073] The first data processing unit 212 is used for processing and preprocessing the PVT output prediction data. First, data processing. In order to facilitate the subsequent tuning of the model hyperparameters by the first model parameter adjustment unit 213, the first 90% of the received data is divided into a training set, and the remaining 10% of the data is divided into a test set. The goal of the training set is to enable the LSTM model to learn the patterns in the data so that it can make accurate predictions. The performance of the LSTM model on the test set can help evaluate the accuracy, precision and generalization ability of the model. In an embodiment of the present invention, the first data receiving unit 211 receives a total of 720 data groups with a sequence number of 3 from the monitoring module 100, the first 648 data groups are divided into training sets, and the last 72 data groups are divided into test sets. Secondly, data preprocessing. Due to the large differences in the scale of PVT output data, the prediction effect of the PVT output prediction unit 214 may be unsatisfactory. Therefore, in order to improve the accuracy and generalization ability of the model, it is necessary to perform unified dimension processing on the PVT output data, that is, normalization processing. For the normalization method, the normalization of standard deviation and mean (Z-scoreNormalization) and the normalization of maximum and minimum values ​​(Min-Max Scaling) are generally used. In order to accelerate gradient descent and avoid model overfitting, the present invention selects the normalization method of standard deviation and mean.

[0074] like Figure 2 As shown in the figure, a neural network model - a long short-term memory network model (LSTM model) is built: sequence input layer, LSTM neural network layer, dropout layer, fully connected layer and regression layer. For the specific model construction diagram, see Figure 2 ; Select the root mean square error (RMSE) as the loss function of the LSTM model; The optimizer selects the stochastic gradient descent method Adam algorithm and tunes its parameters to find the optimal solution of the model. Compared with the RNN recurrent neural network, the LSTM model can alleviate the problems of gradient disappearance and gradient explosion, and the unique gating mechanism of the LSTM model allows the network to adaptively learn patterns and regularities in the data, so that it can perform better on various tasks and data types.

[0075] In this embodiment, the specific configuration of the selected Adam algorithm parameters is: the initial learning rate is 0.001, the exponential decay rate of the first-order moment estimate is is 0.9, and the second-order estimated exponential decay rate is 0.999, and segmented learning is adopted.

[0076] In the above LSTM model, solar radiation intensity, PVT power generation and PVT heat supply are selected as the input values ​​of the input layer, and the PVT power generation at the next 24 time points is output. and PVT heating Specifically, the solar radiation intensity at 24h every hour for 30 days (720 time measurement points in total) is selected. (W / m²), PVT power generation (kW) and PVT heating (kW) is used as the input value of the LSTM model prediction input layer, that is, The output value of the PVT output prediction unit 214 is the predicted PVT power generation at the next 24 time points. and PVT heating , the vector representation is .

[0077] LSTM model hyperparameters determine the performance of the prediction model. Therefore, before model training, the first model parameter adjustment unit 213 first divides the data set into a training set and a validation set, and selects a grid search algorithm (Grid Search) to select the model hyperparameters. The algorithm traverses the parameter combination of the learning rate and the number of hidden units, and cross-validates to determine the best parameter combination with the result of the evaluation function, and the evaluation index is the root mean square error function RMSE. In this embodiment, in the first model parameter adjustment unit 213, the given parameter combination is the learning rate and the number of hidden units. Based on the training set and the test set processed by the first data processing unit 212, the grid search technology is used to find the smallest RMSE value among all combinations of the Cartesian product of these two parameter sets, so as to establish a prediction model with the best prediction effect.

[0078] The PVT output prediction unit 214 uses the established multi-feature input LSTM model assisted by the optimizer Adam neural network optimization algorithm to predict the PVT output value for each period of 24 hours in the next day based on the PVT output value of the past month. That is, the input variables of the electric and thermal load prediction unit 224 are the PVT output values ​​at the hour of the previous 30 days and the local solar radiation values ​​of the corresponding 30 days, and the output variables are the PVT power supply value and the PVT heating value at the hour of the 31st day.

[0079] In the embodiment, it can be known from common sense and observation that the load is related to the daily average temperature and the day type. When the second data receiving unit 221 receives the day type data, 0 is used to represent a working day and 1 is used to represent a rest day.

[0080] The second data receiving unit 221 receives the electric heat load value at 24h every day of the past month, the average temperature of the day and the day type data required for electric heat load prediction from the monitoring module for prediction or parameter adjustment, and uses linear interpolation to fill in the missing data.

[0081] The second data processing unit 222 is used for processing and preprocessing the electric and thermal load forecasting data. The detailed steps are shown in the first data processing unit 212. The second data processing unit 222 is similar to the first data processing unit 212, except that the input value of the model selection is and output value Specifically, select the household electricity load value at 24h every hour for 30 days (720 time measurement points in total) (kW), heat load value (kW), average temperature of the day (℃) and day type As the input value of the LSTM model input layer, that is The output value of the electric heat load prediction unit 224 is the household electric load value predicted at 24 time points in the future. and heat load value , the vector representation is .

[0082] The second model parameter adjustment unit 223 for the electric and thermal load is similar to the first model parameter adjustment unit 213 for the PVT output prediction.

[0083] The electric and thermal load prediction unit 224 predicts the electric and thermal load for each period of the next 24 hours based on the household electric and thermal load values, the corresponding daily average temperature and the day type in the past month through the established multi-feature input LSTM model assisted by the optimizer Adam neural network optimization algorithm. That is, the input variables of the PVT output prediction unit 214 are the household electric load values, household thermal load values, the corresponding daily average temperature and the day type at the hour of the previous 30 days, and the output variables are the household electric load and thermal load at the hour of the 31st day.

[0084] The optimization control module 300 establishes a multi-objective planning model, taking the total operating costs such as the operation and maintenance costs of the energy storage unit and the electricity purchase costs during operation and the carbon emissions of the entire system as the optimization targets, and taking the load characteristics, operation and maintenance costs, peak and valley electricity prices of the energy storage unit as constraints, and using the particle swarm algorithm to optimize the "source, storage, and network" through the controller for the next 24 hours a day, so that the family can maximize the consumption of renewable resources, reduce operating costs, and achieve low-carbon operation of buildings. See the energy system strategy execution process for details. Figure 3 .

[0085] The optimization control module 300 includes a data input unit 310 , an optimization unit 320 and a control unit 330 .

[0086] The data input unit 310 is mainly used to receive the 24-hour PVT power supply value, the 24-hour PVT heating value, the 24-hour household power load and the 24-hour household heat load from the prediction module 200 and the monitoring module 100. These vectors are input values ​​for the optimization unit 320 to establish a model.

[0087] Preferably, the improved particle swarm algorithm used in the optimization unit 320 improves the inertia factor and the learning factor, so that the improved particle swarm algorithm is less likely to fall into a local optimal solution than the traditional particle swarm algorithm, and can find the global optimal solution faster and better.

[0088] The optimization unit 320 uses the selected solution method to improve the particle swarm algorithm to solve the established multi-objective planning model and obtain the actual power generation of the PVT of the optimal solution for 24 hours. 、The actual heat supply of PVT , battery transmission power , Grid transmission power , Heat storage tank transmission power and natural gas grid transmission power , as well as the operating cost and carbon emission values ​​of the scheme corresponding to the objective function.

[0089] The control unit 330 is mainly used to receive the optimization results of the optimization unit 320 and control the "source, storage, and network" of electric heating 24 hours a day in the future through the controller DDC and 5G Internet of Things technology. In the embodiment, the modeling stage in the optimization unit 320 models the home energy system optimization problem as a multi-objective planning model, and the specific steps include:

[0090] Step 1: Determine the optimization target. According to social surveys, users' primary consideration for home energy management systems is the carbon emissions of the entire system, and the second consideration is the operating cost of the entire system. The optimization targets selected by the present invention are operating cost and carbon emissions.

[0091] Step 2: Determine the decision variables. Determine the decision variables that affect the objective function. These variables are the parameters or variables that need to be optimized in the model. The decision variables selected by the present invention are the actual power generation of PVT. 、The actual heat supply of PVT , battery transmission power , Grid transmission power , Heat storage tank transfers heat Heat transfer to the natural gas grid , and mathematically model the battery and hot water storage tank:

[0092] (1)

[0093] (2)

[0094] Where: represents the total amount of electrical energy in the household battery at time t; , Respectively represent the charging and discharging efficiency of household batteries; Indicates the charging or discharging power of the battery. When it is greater than zero, it means charging; represents the total amount of heat energy in the household water storage tank at time t; , They represent the heat storage and heat release efficiencies of the household water storage tank respectively; It represents the heat storage or heat release power of the hot water storage tank at time t. Greater than zero indicates heat storage; Represents the heat dissipation loss of the household water storage tank at time t.

[0095] Among them, the heat loss of the hot water storage tank It can be calculated by the following formula:

[0096] (3)

[0097] Where: represents the surface area of ​​the jth hot water storage tank; Indicates the heat loss coefficient of the hot water storage tank [ ]; Indicates the temperature of the working fluid in the hot water storage tank at that moment (℃); represents the ambient temperature at that moment (°C); m represents the total number of hot water storage tanks.

[0098] Step 3: Establish constraints. Constraints can be linear or nonlinear equations about decision variables, or they can be restrictions on the range of decision variables. In the embodiment, three constraints are established: power balance constraints, main network transmission power constraints, and energy storage unit constraints. The specific mathematical model is as follows:

[0099] (A1) Power balance constraints

[0100] Electrical balance: (4)

[0101] Thermal balance: (5)

[0102] (A2) Main network transmission power constraints

[0103] (6)

[0104] (7)

[0105] (A3) PVT power constraints

[0106] (8)

[0107] (9)

[0108] (A4) Battery restraint

[0109] (10)

[0110] (11)

[0111] (12)

[0112] (A5) Hot water storage tank constraints

[0113] (13)

[0114] (14)

[0115] (15)

[0116] In the formula, Indicates the actual power generation of the photovoltaic cell (kWh); Indicates the amount of electricity purchased from the main grid (kWh); Indicates the charge and discharge capacity of the battery (kWh); It represents the predicted value of household electricity load (kWh); Indicates the power consumption of the circulating water pump; Indicates the power consumption of the gas water heater; Indicates the actual heat supply of the collector (kWh); Indicates the calorific value of gas purchased from the main grid (kWh); Indicates the charging and discharging heat of the hot water storage tank (kWh); represents the predicted value of household heating load (kWh); and Indicates the upper and lower limits of the power transmission of the power grid; and Indicates the upper and lower limits of the natural gas pipeline flow; q indicates the calorific value of natural gas, which is generally 35×10³( ); represents the predicted power generation of the photovoltaic cell at time t (kWh); It represents the predicted heating supply of the collector at time t; Indicates the state of charge of the battery at time t; represents the total amount of electrical energy in the household battery at time t; Indicates the rated capacity of the battery; and Indicates the upper and lower limits of the battery's power storage capacity; and Indicates the upper and lower limits of the battery transmission power; Indicates the heat storage state of the hot water storage tank at time t; represents the total amount of heat energy in the household water storage tank at time t; Indicates the rated capacity of the hot water storage tank. and Indicates the upper and lower limits of the heat storage capacity of the hot water storage tank; and Indicates the upper and lower limits of the power transmitted by the hot water storage tank.

[0117] Step 4: Formulate the objective function. Convert the determined optimization goal into mathematical form and establish the objective function of the multi-objective programming model. In this embodiment, the objective function F is formulated: operating cost and carbon emissions , its vector form is , the specific mathematical model is as follows:

[0118] Objective 1: Operating costs

[0119] (16)

[0120] In the formula, represents the electricity purchase price at time t [yuan / (kWh)]; represents the operation and maintenance cost of the battery at time t [yuan / (kWh)]; represents the amount of electricity purchased at time t (kWh); Indicates the charge and discharge amount of the battery at time t (kWh); represents the natural gas purchase price at time t (yuan / m3); represents the operation and maintenance cost of the hot water storage tank at time t (yuan / m3); represents the operation and maintenance cost of the circulating water pump at time t (yuan / m3); Indicates the heat purchased from the heating network at time t (kWh); Indicates the heat capacity of the hot water storage tank at time t (kWh); Represents the actual heating capacity of the PVT system at time t (kWh).

[0121] Target 2: Carbon Emissions

[0122] Since the current main power grid still mainly uses coal and methane for power generation and heating, only the main power grid in the entire system will produce carbon dioxide. Therefore, in the second carbon emission target, only the carbon emissions of the main grid are considered. According to the carbon accounting method provided by the IPCC, the objective function of target 2 is listed as follows:

[0123] (17)

[0124] Where: represents the carbon emission conversion coefficient of electricity, which is 0.830; represents the carbon emission conversion coefficient of natural gas, which is 0.191; Represents the carbon emission conversion coefficient of renewable energy, which is set to 0.

[0125] Step 5: Select a solution. Select an appropriate multi-objective planning solution method to solve the problem. Common methods include weighted summation method, Pareto optimal solution method, and multi-objective evolutionary algorithm. Considering that the optimization problem of home energy system has the characteristics of high dimension, discontinuity, and multiple constraints, this design adopts the improved particle swarm algorithm (MOPSO), which has better convergence efficiency and speed than general optimization algorithms and can obtain better results.

[0126] In the traditional particle swarm optimization (PSO), assuming that the dimension of the decision variable is d, the total number of initialized particle swarms is M, where the position and velocity vectors of the nth particle are respectively and The nth particle tracks its previous individual optimal solution and the global optimal solution To adjust its position and speed, the vector forms of the two optimal solutions are and . Update the velocity of particle n in the next iteration and location The formula is:

[0127] (18)

[0128] In the formula, , is a random number uniformly distributed between [0,1]; , They are individual learning factor and social learning factor; is the inertia weight.

[0129] Traditional particle swarms tend to fall into local optimal solutions when solving optimization problems, and lack global exploration capabilities, resulting in slow convergence, low search accuracy, and difficulty in obtaining global optimal solutions in practical applications. In order to solve these problems, the adaptive parameters in traditional particle swarms are improved, and two methods, nonlinear dynamic adjustment of inertia weight value and nonlinear dynamic adjustment of learning factor, are used to form an improved particle swarm algorithm. The specific improvements are as follows:

[0130] In the early stage of algorithm optimization, the slow movement of particles will affect the algorithm search speed. At this time, it is necessary to increase the inertia weight value. and social learning factors In order to facilitate global search; in the later stage of optimization, the particles may skip the optimal solution due to their fast movement speed, so it is necessary to reduce the inertia weight value and individual learning factors To facilitate local search. In summary, dynamically update the inertia weight value The formula is as follows:

[0131] (19)

[0132] Dynamically update learning factors , The formula is as follows:

[0133] (20)

[0134] Where: Indicates the inertia weight value; and Represents individual learning factor and social learning factor; IT represents the current iteration number; ST represents the maximum iteration number; the superscript s represents the initial value of the value, and the superscript e represents the final value of the value.

[0135] In the embodiment, the improved particle swarm algorithm is applied in the optimization unit 320, and the MOPSO algorithm flow chart is shown in Figure 4 :

[0136] Step 51, algorithm parameter setting. Set the total number of particles M = 100, the particle dimension d = 96, and the inertia weight , are 0.5 and 0.001 respectively, and the individual learning factor , They are 2.5 and 0.5 respectively, and the social learning factor , They are 0.5 and 2.5 respectively, and the maximum number of iterations ST is 100.

[0137] Step 52: Under the condition that the constraint conditions (4)-(15) are satisfied, the decision variables , , , , and The upper and lower limits of the power are randomly generated to initialize the population position. The upper and lower limits of the decision variables are set to [0,5], [0,5], [0,4] and [0,4] respectively.

[0138] Step 53, evaluate the fitness value F of each particle according to equations (16) and (17), i.e., the operating cost and carbon emissions , to determine its position in the Pareto solution set. Obtain the individual optimal solution and the global optimal solution , the individual optimal solution and the global optimal solution Stored in the archive.

[0139] Step 54, perform boundary processing and set the battery and 0.1 and 0.9 respectively, hot water storage tank and are 0.1 and 0.9 respectively. Determine whether the battery has reached the upper and lower limits of charge and discharge at time t and whether the hot water storage tank has reached the upper and lower limits of heat storage and release at time t.

[0140] Step 55: determine whether the maximum number of iterations SI is reached. If the determination is not satisfied, the inertia weight w and the learning factor are updated using formulas (19) and (20). and , continue to execute step 53 and step 54; if satisfied, output the Pareto frontier.

[0141] Step 6: Selection of the optimal solution. The Pareto front provides a set of solutions that balance two conflicting objectives. It provides decision makers with a series of non-inferior solutions, but the final choice needs to be based on the actual situation and the decision maker's judgment. Therefore, in this embodiment, a series of non-inferior solutions are processed based on the TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) algorithm to select the optimal operation plan. TOPSIS algorithm solution process:

[0142] Step 61, inputting original data. Assume that there are p solutions in the output Pareto solution set. The Pareto solution set with p solutions is input as original data, that is, p operation plans are input, and each operation plan has two indicator data, namely, operation cost (C) and carbon emission (E).

[0143] The input variables can be represented as a decision matrix D with dimension p × 2. Each element in the matrix D (l=1,2,...,p; q=1,2) represents the value of the l-th solution on the q-th indicator. Specifically, the matrix D can be expressed as follows:

[0144] (twenty one)

[0145] Where: represents the operating cost of the lth solution, which can also be expressed as ; represents the carbon emissions of the lth option, which can also be expressed as ; p is the number of optimization solutions.

[0146] Step 62, data preprocessing. In order to apply the decision matrix D to the TOPSIS algorithm, it is necessary to standardize each column of the decision matrix 𝐷. Use the vector normalization method to preprocess each column of the decision matrix D. The standardized decision matrix It can be expressed as:

[0147] (twenty two)

[0148] Where: represents the running cost vector The 2-norm of ; Represents the carbon emission vector The 2-norm of .

[0149] Step 63: Weighted processing. The weight represents the importance of each indicator. Since these two indicators are equally important, the weights of these two indicators are Set the indicator weight vector to 0.5 , and the decision matrix The weighted processing is as follows:

[0150] (twenty three)

[0151] Where: Represents the standardized decision matrix Decision matrix after weighting.

[0152] Step 64, find the positive ideal solution and the negative ideal solution. Since both data are cost-type attributes, when finding the positive ideal solution, each indicator corresponds to the minimum value among all its solutions; and when finding the negative ideal solution, each indicator corresponds to the maximum value among all its solutions. The formula is as follows:

[0153] (twenty four)

[0154] Where: Represents the standardized decision matrix The positive ideal solution of the qth column; Represents the standardized decision matrix Negative ideal solution of the qth column; Represents the standardized decision matrix The elements in .

[0155] Step 65, obtain the comprehensive evaluation index of each scheme. The comprehensive evaluation index formula of each operation scheme is as follows:

[0156] (25)

[0157] Where: Represents the comprehensive evaluation index of the lth option.

[0158] Step 66: The comprehensive evaluation index of each scheme is Sort from large to small, and the solution corresponding to the largest number of indicators is the optimal solution.

[0159] In order to test the feasibility and effectiveness of the proposed home energy system, a small single-family residential building in Zhenjiang City, Jiangsu Province was selected to build a home energy system hardware device based on the PVT system. The schematic diagram of the home energy system hardware device is shown in Figure 5 , the construction steps are as follows:

[0160] Step D1, install the PVT system according to the area of ​​the building roof to ensure sufficient sunlight and suitable installation conditions. Select NES72-6-330P as the model of PVT solar panel, install a total of twelve units, and install the support structure, photovoltaic components and collectors at the same time. Select AL-FC-GV2.0 as the model of heat pump collector. The installation of the support structure must comply with relevant safety standards and building specifications. Select TYT-1 as the model of circulation pump, install two units. Select ZYSUN-150A as the model of heat storage tank, install a total of three units. And install an 8L gas water heater. Finally, connect the pipes between the collector, water pump, gas water heater and heat storage tank to ensure smooth heat energy transmission.

[0161] Step D2, install the PVT management system. Select XL-21 as the controller model to monitor the operating status of the PVT system and ensure the normal operation of the system. Select SUN2000-36KTL as the photovoltaic inverter model and 6-CNJ-200 as the battery model. Install six units in total. Connect photovoltaic panels, photovoltaic inverters, batteries, smart meters, controllers, household loads and other equipment. Select DDSZ6 as the model of the smart meter. The information monitored by the meter is uploaded to the software system.

[0162] Step D3: After the installation of the PVT system is completed, the system is debugged and tested to ensure that all components are working properly and the system is running stably. In addition, it is combined with the constructed software system to use the controller and 5G Internet of Things technology to monitor and optimize household energy and carbon emissions.

[0163] The PVT system installed capacity of the building is about 5KW, and the total household power storage capacity is The battery charge and discharge efficiency is 20kWh. and is 0.89, the total household heat storage capacity The heat storage efficiency of the hot water storage tank is 20kWh. and is 0.89, and the heat loss coefficient K of the hot water storage tank is 0.5 , the surface area A of the box is 2.1㎡. The monitoring module 100 collects the PVT output data, local environmental conditions and household electric and heating loads of the building at intervals of 60 minutes from August 1 to August 30, 2022, and uses 30 days of data to input into the prediction module 200 to train the LSTM model. The detailed solution process of the model is shown in the prediction module 200. The first model parameter adjustment unit 213 uses a grid search algorithm to find the learning rate set and hidden units are 0.2 and 500 respectively; the second model parameter adjustment unit 223 uses a grid search algorithm to find the learning rate set and hidden units are 0.2 and 1000 respectively; the error function RMSE values ​​of the PVT output prediction module 210 are 0.1425 and 0.1132 respectively, and the error function RMSE values ​​of the electric and heating load prediction module 220 are 0.116 and 0.1663 respectively. The comparison between the predicted value and the true value of the PVT output prediction module 210 is shown in Figure 6 The comparison between the predicted value and the actual value of the electric and thermal load prediction module 220 is shown in Figure 7 Finally, the household electric load PL, household heat load PH, PVT power generation P and PVT heating H for 24 hours on the 31st day are output. Figure 8 (a), (b), (c), (d) in the figure. The training curve of household electric heating load model is shown in Fig. 9 , PVT output model training curve is shown in Fig.10 .

[0164] The output result of the prediction module 200 is then transmitted to the optimization control module 300. Taking the typical summer working condition of the building as an example, the solution optimization process of the model is shown in the optimization control module 300. Fig.11 The final optimization result obtained 66 non-inferior solutions. Based on the TOPSIS algorithm, the 44th non-inferior solution was selected as the optimal operation plan. The operation cost of this plan was 3.53 yuan and the carbon emissions were 2.08 kg. The results of the PVT system and energy storage system after operation optimization are shown in Fig.12 ,from Fig.12 From (a)-(b) in Figure 2, it can be seen that the household energy system absorbs the output of the PVT system as much as possible. Fig.12 It can be seen from (c) in the figure that the SOC value of the energy storage system in this operation scheme is within the upper and lower limits, which meets the system settings. The results after the optimization of the operation of the entire home energy system are shown in Fig.13 (a)-(b) in the figure. Compared with ordinary residential buildings, it can save 14.4 kWh of electricity and 0.28 m³ of natural gas every day in summer, which is equivalent to reducing carbon emissions by 14.11 kg. It is estimated that carbon emissions can be reduced by about 2,000 kg throughout the year. While saving energy and reducing emissions, the system operating fee only costs 3.54 yuan per day. Fig.14The amount of electricity purchased from the grid by the home energy system after operation optimization shows that the system avoids purchasing electricity during the “peak” period most of the time, which can effectively reduce the system operating costs and further verify the reliability of the home energy system.

Claims

1. A low-carbon operation optimization method for a household energy system based on PVT photovoltaic thermal energy, characterized in that: The steps include: S1, select the system's operating cost and carbon emissions as optimization targets; S2, with the actual power generation of PVT, the actual heat supply of PVT, the battery transmission power, the grid transmission power, the heat transmitted by the hot water storage tank and the heat transmitted by the natural gas grid as decision variables, mathematical modeling of the battery and the hot water storage tank is carried out; S3, with power balance constraint, main grid transmission power constraint, PVT power constraint, battery constraint and hot water storage tank constraint as constraint conditions, establish a multi-objective planning model with operation cost and carbon emission as optimization objectives; S4, using the improved particle swarm algorithm to optimize the multi-objective programming model and obtain the optimal solution set; S5, processing the obtained optimal solution set using the superior and inferior solution distance method to obtain the optimal operation plan; In step S2, mathematical modeling is performed on the battery and the hot water storage tank: In the formula, E b (t) represents the total amount of electrical energy in the household battery at time t; Respectively represent the charging and discharging efficiency of household batteries; P B (t) represents the charging or discharging power of the battery. B (t)>0 means charging, when P B (t)<0 represents discharge; E h (t) represents the total amount of heat energy in the household water storage tank at time t; Respectively represent the heat storage and heat release efficiency of the household water storage tank; H B (t) represents the heat storage or heat release power of the hot water storage tank at time t. B (t)>0 represents heat storage, H B (t) < 0 means exothermic; represents the heat dissipation loss of the household water storage tank at time t; Among them, the heat loss of the hot water storage tank The calculation formula is as follows: In the formula, A j represents the surface area of ​​the jth hot water storage tank; K represents the heat loss coefficient of the hot water storage tank, in kJ / (m 2 ℃); T j (t) represents the temperature of the working fluid in the hot water storage tank at that moment, T env (t) represents the ambient temperature at that moment, in °C; m represents the total number of water tanks; In step S3, the mathematical models of the power balance constraint, main network transmission power constraint, PVT power constraint, battery constraint and hot water storage tank constraint are as follows: A1, power balance constraint: P PV (t)+P G (t)+P B (t) = P' L (t)+P p (t)+P W (t) H PT (t)+H G (t)+H B (t)=H' L (t) A2, main network transmission power constraints: A3, PVT power constraint: 0≤P PV (t)≤P'(t) 0≤H PT (t)≤H'(t) A4, Battery constraints: A5, hot water storage tank constraints: Where P PV (t) represents the actual power generation of the photovoltaic cell, P G (t) represents the amount of electricity purchased from the main grid, P B (t) represents the charge and discharge capacity of the battery, P L '(t) represents the predicted value of household electricity load, P p (t) represents the power consumption of the circulating water pump, P W (t) represents the power consumption of the gas water heater, H PT (t) represents the actual heat supply of the collector plate, H G (t) represents the calorific value of gas purchased from the main grid, H B (t) represents the heat of the hot water tank, H' L (t) represents the predicted value of household heat load, in kWh; They represent the upper and lower limits of the power transmission of the power grid respectively; They represent the upper and lower limits of the natural gas pipeline flow respectively; q represents the calorific value of natural gas; P'(t) represents the predicted power generation of the photovoltaic cell at time t, kWh; H'(t) represents the predicted heat supply of the collector at time t; SOC B (t) represents the state of charge of the battery at time t; E b (t) represents the total amount of electrical energy in the household battery at time t; E b,sum Indicates the rated capacity of the battery; Respectively represent the upper and lower limits of the battery's power storage capacity; Respectively represent the upper and lower limits of the battery transmission power; SOC H (t) represents the heat storage state of the hot water storage tank at time t; E h (t) represents the total amount of heat energy in the household water storage tank at time t; E h,sum Indicates the rated capacity of the hot water storage tank; They represent the upper and lower limits of the heat storage capacity of the hot water storage tank respectively; They represent the upper and lower limits of the power transmitted by the hot water storage tank respectively; The multi-objective programming model F is: F=[f1,f2] Among them, f1 is the operating cost, f2 is the carbon emission; In the formula, C grid (t) represents the electricity purchase price at time t, C bess (t) represents the operation and maintenance cost of the battery at time t, P G (t) represents the amount of electricity purchased at time t, P B (t) represents the charge and discharge capacity of the battery at time t, in kWh; C gridH (t) represents the natural gas purchase price at time t, C bessH (t) represents the operation and maintenance cost of the hot water storage tank at time t, C pump (t) represents the operation and maintenance cost of the circulating water pump at time t, in units of RMB / m 3 ;H G (t) represents the amount of heat purchased from the heating network at time t, H B (t) represents the heat of the hot water storage tank at time t, H PT (t) represents the actual heat supply of the PVT system at time t, in kWh; In the formula, δ g Indicates the carbon emission conversion coefficient of electricity; δ hg Indicates the carbon emission conversion coefficient of natural gas; δ P represents the carbon emission conversion factor of renewable energy; In step S4, the improved particle swarm algorithm is implemented as follows: S51, set the total number of particles M, the dimension of the particles d, the inertia weight, the individual learning factor, the social learning factor, and the maximum number of iterations ST; S52, with decision variable P PV (t), P G (t), P B (t), H PT (t), H G (t) and H B (t) generates particles randomly with upper and lower limits, and initializes the population position; S53, based on the minimum operating cost f1 and carbon emission f2, evaluate the fitness value F of each particle, obtain the individual optimal solution and the global optimal solution, and store the individual optimal solution and the global optimal solution in the archive library; S54, setting upper and lower limits of battery charge and discharge, and upper and lower limits of heat storage and release of the hot water storage tank, and judging whether the battery reaches the upper and lower limits of charge and discharge, and whether the hot water storage tank reaches the upper and lower limits of heat storage and release at time t by the individual optimal solution; S55, if the maximum number of iterations SI is reached, the Pareto frontier is output; if the maximum number of iterations SI is not reached, the inertia weight and the learning factor are updated, and steps S53 and S54 are continued; The superior and inferior solution distance method is used to process a series of non-inferior solutions provided by the Pareto frontier and select the optimal operation plan. The specific implementation steps are as follows: S61, assuming that there are p solutions in the output Pareto solution set, the Pareto solution set of p solutions is used as the original data input, each solution represents an operation plan, and each operation plan has two indicator data, namely, operation cost and carbon emission; The input variable is represented as a decision matrix D, whose dimension is p×2. Each element D in the matrix D lq Represents the value of the l-th solution on the q-th indicator: Where, l = 1, 2, ..., p; q = 1, 2; C l represents the operating cost of the lth solution, E l represents the carbon emission of the lth option; S62, using a vector normalization method to perform data preprocessing on each column of the decision matrix D to obtain a standardized decision matrix D'; S63, set the index weight vector w' = [0.5, 0.5] T , and weight the decision matrix D', the formula is as follows: In the formula, represents the decision matrix after the weighted processing of the standardized decision matrix D'; S64, find the positive ideal solution and the negative ideal solution, the formula is as follows: In the formula, Represents the standardized decision matrix The positive ideal solution of the qth column; Represents the standardized decision matrix The negative ideal solution of the qth column; d lq Represents the standardized decision matrix Elements in S65, calculate the comprehensive evaluation index of each operation plan, the formula is as follows: In the formula, represents the comprehensive evaluation index of the lth scheme; S66, the comprehensive evaluation index of each scheme Sort from large to small, and the solution corresponding to the largest number of indicators is the optimal solution.

2. A management system for low-carbon operation of a household energy system based on PVT photovoltaic thermal energy, used to implement the optimization method as claimed in claim 1, characterized in that: It includes a monitoring module (100), a prediction module (200) and an optimization control module (300); The monitoring module (100) monitors, collects and stores information on all energy efficiency devices in the residential building and the environmental conditions outside the residential building through intelligent information collection equipment and the Internet; The prediction module (200) is used to predict the user's electric heat load and the actual output value of photovoltaic thermal energy, and comprises a PVT output prediction module (210) and an electric heat load prediction module (220); the PVT output prediction module (210) comprises a first data receiving unit (211), a first data processing unit (212), a first model parameter adjustment unit (213) and a PVT output prediction unit (214); the electric heat load prediction module (220) comprises a second data receiving unit (221), a second data processing unit (222), a second model parameter adjustment unit (223) and an electric heat load prediction unit (224); The first data receiving unit (211) and the second data receiving unit (221) are respectively used to receive training data required for prediction from the monitoring module (100); the first data processing unit (212) and the second data processing unit (222) are respectively used for data processing and preprocessing for prediction; the first model parameter adjustment unit (213) and the second model parameter adjustment unit (223) are respectively used to adjust the parameters of the LSTM model to obtain evaluation indicators of the prediction module (200) under different parameters, and select the LSTM model with the best prediction effect for subsequent prediction; the PVT output prediction unit (214) and the electric heat load prediction unit (224) respectively predict the data of the next day by using the established multi-feature input LSTM model; The optimization control module (300) takes the total operating cost of the energy storage unit and the carbon emissions of the entire system during operation as optimization targets, takes the load characteristics, operation and maintenance costs, and peak and valley electricity prices of the energy storage unit as constraints, and uses a particle swarm algorithm to optimize and control the "source, storage, and network" for the next 24 hours through a PVT controller.

3. According to claim 2, the management system for low-carbon operation of a household energy system based on PVT photovoltaic thermal energy is characterized in that: The optimization control module (300) comprises a data input unit (310), an optimization unit (320) and a control unit (330); The data input unit (310) is used to receive required data from the prediction module (200); The optimization unit (320) optimizes the electric and thermal "source, storage, and network" for the next 24 hours by using an improved particle swarm algorithm to obtain the actual power generation of PVT, the actual heat supply of PVT, the battery transmission power, the grid transmission power, the hot water storage tank transmission power, and the natural gas network transmission power of the optimal solution for the next 24 hours, as well as the operating cost and carbon emission value; The control unit (330) is used to receive the optimization result of the optimization unit (320) and control the "source, storage, and network" of electric heating at the hour of the next 24 hours through the PVT controller and the Internet.

4. According to claim 2, the management system for low-carbon operation of a household energy system based on PVT photovoltaic thermal energy is characterized in that: The PVT output prediction unit (214) predicts the PVT output value of each period of the next 24 hours based on the PVT output value of the past month through the LSTM model; wherein the input variables of the LSTM model are the PVT output values ​​at the hour of the previous 30 days and the local solar radiation values ​​of the corresponding 30 days, and the output variables are the PVT power supply value at the hour of the 31st day and the PVT heating value at the hour; The electric heat load prediction unit (224) predicts the electric heat load in each period of the next 24 hours based on the household electric heat load value, the corresponding daily average temperature and the day type in the past month through the LSTM model; wherein the input variables of the LSTM model are the household electric load value, the household heat load value, the corresponding daily average temperature and the day type at the hour of the previous 30 days, and the output variables are the household electric load and heat load at the hour of the 31st day.