CN-121998294-A - Energy prediction and intelligent control method for residence comprehensive energy system based on artificial intelligence

CN121998294ACN 121998294 ACN121998294 ACN 121998294ACN-121998294-A

Abstract

The invention discloses an artificial intelligence-based residential comprehensive energy system energy prediction and intelligent control method, which comprises the steps of comprehensively considering weather characteristics such as ambient temperature, relative humidity, solar radiation and the like and time sequence characteristic influences through a deep neural network, realizing multi-characteristic fusion load prediction, establishing a load feasible region based on a multi-order RC model by combining load prediction with building thermal inertia and thermal comfort demands, and comprehensively regulating and controlling residential energy thermal inertia and thermal comfort individualized demands and multi-source coupling optimization regulation and control by adopting reinforcement learning and mathematical planning/heuristic algorithm. The invention can meet the personalized thermal comfort and energy consumption requirements of users, realize optimization of real-time performance and robustness under high-precision load supply, and can adapt to the personalized requirements of different types of households and equipment combinations.

Inventors

LI MINGJIA
SHANG QINGYUAN
WANG RUILONG
HUANG ZHAOBIN

Assignees

北京理工大学

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. The energy prediction and intelligent control method for the residence comprehensive energy system based on artificial intelligence is characterized by comprising the following steps: the method comprises the steps of carrying out multidimensional association modeling and mode extraction on meteorological features and time sequence features, and obtaining prediction results of different load demands of users based on correlation screening and multi-scale feature fusion; Adopting a multi-order RC building thermodynamic model, and generating a dynamic feasible load domain representing personalized indoor state constraint by combining building thermal inertia, user thermal comfort requirements and the prediction result; The heat comfort of the user and the individual requirements of energy conservation and cost saving are considered, a double-layer collaborative optimization algorithm is constructed, an energy conservation and cost saving strategy which takes the heat inertia of the building, the heat comfort of the user and the coupling control of multi-source equipment into consideration is constructed, and the energy consumption optimization management is realized.
2. The method for predicting and intelligently controlling the comprehensive energy system of the residence based on artificial intelligence according to claim 1, wherein the multi-dimensional association modeling and pattern extraction of the meteorological element features and the time sequence features comprises the following steps: acquiring various historical meteorological data and historical load data at least comprising ambient temperature, relative humidity, wind direction, wind speed, solar radiation and atmospheric pressure; The method comprises the steps of obtaining pearson correlation coefficients among all weather features, historical load data and other weather features, and selecting the weather features and time sequence features as input parameters according to the correlation degree; And carrying out time sequence and weather multivariable feature correlation screening and mapping relation extraction according to the historical load data and various historical weather data, and constructing a deep neural network prediction model integrating the time sequence, the weather multivariable features and the user energy consumption habit demand data features so as to output a short-term load demand curve.
3. The method for predicting and intelligently controlling the comprehensive energy system of the residence based on artificial intelligence according to claim 2, wherein the correlation screening and mapping relation extraction comprises the following steps: and when the weather multi-feature meets the condition at the same time, selecting the one with higher relevance to the load to inhibit redundancy and promote robustness.
4. The method for predicting and intelligently controlling the residential integrated energy system based on artificial intelligence according to claim 2, wherein the data demand characteristics of the deep neural network prediction model comprise time sequence characteristics and depth weather characteristic association characteristics; The deep neural network prediction model comprises a shallow layer feature extraction module, a deep layer feature extraction module, a time sequence feature extraction module, a cross-modal feature fusion module and an output module; The shallow feature extraction module and the deep feature extraction module capture shallow and deep dependency relationship of meteorological features on load change, the time sequence feature extraction module extracts time sequence feature periods and trend information, the cross-modal feature fusion module carries out multi-scale and cross-feature fusion according to different feature types, and the output module maps the fused features to load demand values.
5. The method for predicting and intelligently controlling the energy consumption of the residence comprehensive energy system based on artificial intelligence according to claim 1, wherein the method is characterized in that a multi-order RC building thermodynamic model is adopted, and a dynamic feasible load domain representing personalized indoor state constraint is generated by combining building thermal inertia, user thermal comfort requirements and the prediction result, and is realized as follows: acquiring heat transfer characteristic parameters of a building, wherein the heat transfer characteristic parameters comprise building envelope patterns, heat transfer coefficients, window wall ratios and building thermal resistances, and acquiring heat storage characteristic parameters of the building, wherein the heat capacity of building materials, building areas and heat storage coefficients; Constructing a multilevel RC building thermodynamic model based on the parameters, and calibrating the model by utilizing the prediction result under the condition of setting indoor temperature so as to characterize the thermal inertia characteristic of the building under the given indoor temperature constraint; And describing thermal comfort demand data of building thermal inertia and a load feasible domain on the basis of a prediction result and a multi-order RC building thermodynamic model, wherein the load feasible domain comprises a user thermal comfort maximum interval, a boundary crossing punishment weight and a comfort boundary distance.
6. The method for predicting and intelligently controlling the comprehensive energy system of the residence based on artificial intelligence according to claim 5 is characterized in that the personalized setting of the user is carried out, the setting of the PMV interval value of the thermal comfort of the user is included, the system operates with energy conservation or cost saving as an optimization target, and the user selects according to the self requirement; The PMV interval value of the user thermal comfort is set to be +/-1, +/-0.5 and stable 0, and the optimization target is that each device is energy-saving as a whole, namely, the energy consumption is minimum, or the whole is cost-saving, namely, the lowest-cost optimization operation is realized.
7. The method for predicting and intelligently controlling the comprehensive energy system of the residence based on artificial intelligence according to claim 1, wherein the double-layer collaborative optimization algorithm comprehensively considers the individual requirements of building thermal inertia and user thermal comfort and the energy-saving or cost-saving optimization control of coupling control of multi-source equipment comprises the following steps: Constructing a user energy consumption equipment scene, acquiring equipment operation parameters, and constructing a mapping relation between real-time dynamic equipment operation input parameters and load supply, energy consumption and cost based on MLP; the user sets a PMV value interval of user thermal comfort and sets an energy-saving or cost-saving target in a personalized way; and establishing a Markov chain double-layer optimization model constrained by the equipment operation boundary, the building thermal inertia and the thermal comfort, solving and generating an energy-saving and cost-saving strategy which considers the building thermal inertia, the user thermal comfort and the multi-source equipment coupling control by setting a boundary punishment and a rewarding model, and optimizing.
8. The artificial intelligence based method for predictive and intelligent management and control of a residential integrated energy system of claim 7, wherein the two-layer optimization model comprises: Establishing an upper reinforcement learning model to control indoor thermal inertia and thermal comfort optimization, wherein the upper reinforcement learning model comprises the load prediction result, a reward function optimization target and boundary conditions of the reward function; Constructing a lower heuristic or mathematical programming algorithm scheduling model to control energy supply equipment of a residential comprehensive energy system to operate optimally, wherein the scheduling model comprises an hour-by-hour load demand, an objective function and constraint conditions of the objective function of the algorithm; And the upper reinforcement learning model and the lower heuristic or mathematical programming algorithm scheduling model are subjected to interactive learning to obtain a solution of the double-layer optimization model, obtain an indoor thermal comfort state and equipment energy consumption condition and obtain an optimization operation strategy.
9. The method for predicting and intelligently controlling the comprehensive energy system of a residence based on artificial intelligence according to claim 8, wherein in the method for optimizing control, the reward function is reinforcement learning: Where r boundary is the boundary penalty, r cost is the negative of the sum of the costs of electricity, gas, etc. for the present hour, and r consumption is the negative of the sum of the energy consumption of electricity, gas, etc. for the present hour. Reinforcement learning the bonus function optimization objective is represented by the following formula: Wherein Reward is the maximum cost or energy consumption of each day including boundary penalty, t is each time step, N is the maximum time step, reward t is the reward of the t time step, and the obtained optimal result is the negative number of the maximum energy consumption or cost and the minimum cost or energy consumption on the premise of avoiding boundary penalty; Heuristic algorithm the objective function is represented by the following formula: Wherein E ele,t is the power consumption of each device at time t, cost ele,t is the electricity price at time t, E gas,t is the natural gas consumption at time t, and cost gas is the natural gas price; The boundary conditions for reinforcement learning the bonus function are represented by the following formula: Or (b) Wherein PMV is a predicted average somatosensory index, PMV set is a user set value, which is generally less than or equal to 1, delta PMV is the variation of the predicted average somatosensory index in each time step, T min,set is the set minimum temperature of a room, T is the temperature of the room, T max,set is the set maximum temperature of the room, and delta T is the variation of the room temperature in each time step; the constraint condition of the heuristic algorithm is represented by the following formula: Wherein Q equip.min is the minimum value of the equipment energy supply load, Q equip is the equipment energy supply load, Q equip,max is the maximum value of the equipment energy supply load, and Q demand is the load demand.
10. The method for predicting and intelligently controlling the comprehensive energy system of the residence based on artificial intelligence according to claim 7, wherein the double-layer collaborative optimization algorithm comprises the following steps: Step 1, reading a user thermal comfort constraint limit and an optimization target set by a user, and constructing an upper reinforcement learning per-hour thermal comfort PMV boundary setting and an energy-saving and cost-saving target setting of a double-layer optimization model; aiming at the daily user thermal comfort optimization method, the intelligent body carries out load demand control every time step, and the lower control constraint condition of the double-layer optimization model reaches the minimum cost or minimum energy consumption rewards in a single hour and is transmitted to the intelligent body; Step 3, the agent continuously interacts with the environment to complete one round of operation, obtain the rewarding value under each state and action change, and accumulate knowledge based on the state, action and rewarding; And step 4, obtaining an optimal intelligent agent to perform optimal operation, obtaining an optimal operation control load and equipment regulation and control under an optimization target, and obtaining an optimization strategy under the optimization target.

Description

Energy prediction and intelligent control method for residence comprehensive energy system based on artificial intelligence Technical Field The invention belongs to the technical field of building energy conservation, relates to prediction and optimization control of heat energy of a residence comprehensive energy system, and particularly relates to an artificial intelligence-based residence comprehensive energy system energy prediction and intelligent management and control method. Background The energy consumption in the building field accounts for a large proportion of the energy demand, and the energy-saving and carbon-reducing task is urgent. The comprehensive energy system uses electric power as a hub to couple cold, heat, electricity and renewable energy sources, realizes multi-energy complementation and cooperative energy supply, and becomes an effective path for improving the energy efficiency and reliability of the building. However, the multi-source coupling and the multi-time space scale linkage also obviously increase the complexity of system operation, and higher requirements are put forward on the prediction of the heat load and the optimization of energy consumption of the building side. On the one hand, building thermal inertia and thermal comfort can be regarded as 'virtual energy storage', and the building thermal inertia and thermal comfort are used for peak shifting, valley filling and flexible regulation and control, but are influenced by renewable energy fluctuation, weather and behavior uncertainty, and the prediction of the daily and daily thermal load must be compatible with strong nonlinearity and mutation. The existing statistical and simple machine learning methods are insufficient in describing complex space-time characteristics, and the prediction stability and precision are difficult to meet engineering application. On the other hand, the existing mathematical programming or heuristic algorithm capable of optimizing multi-dependence preset scenes and static parameters needs to be frequently solved again in the face of prediction errors and external environment disturbance, so that the calculation cost is high, and the instantaneity and the robustness are insufficient. Disclosure of Invention In order to overcome the defects of the prior art, the invention aims to provide an artificial intelligence-based prediction and intelligent control method for a residence comprehensive energy system, which adopts a plurality of deep learning networks to comprehensively consider the influences of environmental factors, time sequences and the like, so as to realize multi-feature fusion load prediction and improve prediction precision. The reinforcement learning is combined with the mathematical programming/heuristic algorithm, so that the residential energy thermal inertia and thermal comfort and multisource coupling optimization regulation and control are comprehensively regulated and controlled, and the personalized thermal comfort and energy-saving and cost-saving requirements of users are met. In order to achieve the above purpose, the technical scheme adopted by the invention is as follows: An artificial intelligence-based residential comprehensive energy system energy prediction and intelligent control method comprises the following steps: the method comprises the steps of carrying out multidimensional association modeling and mode extraction on meteorological features and time sequence features, and obtaining prediction results of different load demands of users based on correlation screening and multi-scale feature fusion; Adopting a multi-order RC building thermodynamic model, and generating a dynamic feasible load domain representing personalized indoor state constraint by combining building thermal inertia, user thermal comfort requirements and the prediction result; The heat comfort of the user and the individual requirements of energy conservation and cost saving are considered, a double-layer collaborative optimization algorithm is constructed, an energy conservation and cost saving strategy which takes the heat inertia of the building, the heat comfort of the user and the coupling control of multi-source equipment into consideration is constructed, and the energy consumption optimization management is realized. In one embodiment, the multi-dimensional correlation modeling and pattern extraction of the meteorological element features and the time series features comprises: acquiring various historical meteorological data and historical load data at least comprising ambient temperature, relative humidity, wind direction, wind speed, solar radiation and atmospheric pressure; The method comprises the steps of obtaining pearson correlation coefficients among all weather features, historical load data and other weather features, and selecting the weather features and time sequence features as input parameters according to the correlation degree; And carrying out time sequence and