Energy system regulation and control method and device, electronic equipment and storage medium

Abstract

The invention relates to a regulation and control method and device of an energy system, electronic equipment and a storage medium, and is applied to the technical field of computers, the method comprises the steps that multi-source heterogeneous data of the energy system is acquired, and the multi-source heterogeneous data comprises data influencing energy regulation and control in multiple data sources; performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector; predicting the power consumption behavior of the user based on the high-order feature vector to obtain a prediction result; based on the multi-source heterogeneous data and the prediction result, determining a regulation and control planning strategy of the energy system; and regulating and controlling the power equipment of the energy system based on the prediction result and the regulation and control planning strategy.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular, to a method, device, electronic device, and storage medium for regulating an energy system. Background Art

[0002] In the era of the energy Internet, the "dual carbon" goal has promoted the wide application of clean energy and the rapid development of smart grid technology. As important components of the power system, the household power consumption system and electric vehicles (V2H) play a key role in the transformation of the energy structure. However, in practical applications, the power consumption demands of household electrical appliances and electric vehicles exhibit significant dynamic change characteristics: on the one hand, the start and stop of household electrical appliances have strong randomness and uncertainty; on the other hand, electric vehicles are not only power consumers (such as charging), but can also provide power support to households or the power grid through V2H technology. This two-way energy flow characteristic makes the operation of the energy system more complex.

[0003] Traditional energy switching methods are mainly based on fixed rules. For example, fixed switching thresholds or rule bases are used for energy selection, which may lead to inaccurate control and energy waste in the system when facing dynamically changing power consumption demands. Summary of the Invention

[0004] The present application provides a method, device, electronic device, and storage medium for regulating an energy system, so as to solve the problem in the prior art that using fixed rules for energy selection leads to inaccurate control and energy waste in the system.

[0005] According to the first aspect of the embodiments of the present application, a method for regulating an energy system is provided, including:

[0006] Obtain multi-source heterogeneous data of the energy system, where the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources;

[0007] Perform feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector;

[0008] Predict the user's power consumption behavior based on the high-order feature vector to obtain a prediction result;

[0009] Determine a regulation planning strategy for the energy system based on the multi-source heterogeneous data and the prediction result;

[0010] Regulate the power equipment of the energy system based on the prediction result and the regulation planning strategy.

[0011] Optionally, the performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector includes:

[0012] Extract features from the multi-source heterogeneous data to obtain the temporal features and spatial features of the data from each data source;

[0013] Perform weighted fusion on the temporal features and spatial features of the data from each data source to obtain the high-order feature vector.

[0014] Optionally, predict the user's electricity consumption behavior based on the high-order feature vector to obtain a prediction result, including:

[0015] Input the high-order feature vector into a pre-trained prediction model, and map the high-order feature vector to a high-dimensional embedding space through the embedding layer of the prediction model and add position encoding information to obtain a high-order feature matrix;

[0016] Based on the attention layer in the prediction model, determine the attention weights of the time dimension and space dimension of the high-order feature matrix, as well as the spatio-temporal correlation between different features in the high-order feature matrix, and generate the attention output result of the high-order feature matrix based on the attention weights and the spatio-temporal correlation;

[0017] Perform a non-linear transformation on the attention output result to obtain the prediction result.

[0018] Optionally, based on the multi-source heterogeneous data and the prediction result, determine the regulation and planning strategy of the energy system, including:

[0019] Integrate the multi-source heterogeneous data and the prediction result to obtain integrated data;

[0020] Based on the mixed integer programming algorithm, with a preset optimization goal as the optimization direction, perform basic optimization on the integrated data to obtain an initial optimization result, and the initial optimization result satisfies the preset constraint conditions; the optimization goal includes minimizing the electricity consumption cost and minimizing the life loss cost of the vehicle, and the constraint conditions include power balance constraints and charge and discharge constraints of power equipment;

[0021] Use the initial optimization result as the initial population, and optimize the initial optimization result using the genetic algorithm with the optimization goal as the optimization direction to obtain the regulation and planning strategy.

[0022] Optionally, after determining the regulation and planning strategy based on the multi-source heterogeneous data and the prediction result, it further includes:

[0023] If it is monitored that the multi-source heterogeneous data has changed, update the regulation and planning strategy using the particle swarm optimization algorithm based on the changed data.

[0024] Optionally, based on the prediction result and the regulation and planning strategy, regulate the power equipment of the energy system, including:

[0025] Normalize the predicted result and the regulation planning strategy to obtain normalized data;

[0026] Based on the normalized data, perform power adjustment on the regulation planning strategy to obtain an energy regulation strategy, and the energy regulation strategy can balance the power when the power equipment operates in coordination;

[0027] Control the power equipment to execute according to the energy regulation strategy.

[0028] Optionally, performing power adjustment on the regulation planning strategy based on the normalized data to obtain an energy regulation strategy includes:

[0029] Perform state representation on the normalized data to obtain a high-dimensional state matrix;

[0030] Adopt a reinforcement learning policy network to perform action selection on the high-dimensional state matrix based on a reward function to obtain the target action information of the power equipment; the target action information indicates the power adjustment amount of the power equipment, and the reward function indicates the quality of the target action information;

[0031] Based on the target action information, adjust the power information in the regulation planning strategy to obtain the energy regulation strategy.

[0032] According to the second aspect of the embodiments of the present application, a regulation device for an energy system is provided, including:

[0033] An acquisition unit, configured to acquire multi-source heterogeneous data of the energy system, and the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources;

[0034] A feature fusion unit, configured to perform feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector;

[0035] A prediction unit, configured to predict the user's electricity consumption behavior based on the high-order feature vector to obtain a prediction result;

[0036] A determination unit, configured to determine an energy system regulation planning strategy based on the multi-source heterogeneous data and the prediction result;

[0037] A regulation unit, configured to regulate the power equipment of the energy system based on the prediction result and the regulation planning strategy.

[0038] According to the third aspect of the embodiments of the present application, an electronic device is provided, including a memory and a processor;

[0039] The memory is connected to the processor and is used to store programs;

[0040] The processor is configured to implement the regulation method of the energy system as described in the first aspect by running the program in the memory.

[0041] According to a fourth aspect of the embodiments of the present application, there is provided a storage medium, on which a computer program is stored. When the computer program is run by a processor, the regulation method of the energy system as described in the first aspect is implemented.

[0042] According to a fifth aspect of the embodiments of the present application, there is provided a computer program product, including computer program instructions. When the computer program instructions are run by a processor, the processor is caused to execute the regulation method of the energy system as described in the first aspect.

[0043] The above technical solutions provided by the embodiments of the present application have the following advantages compared with the prior art: In the method provided by the embodiments of the present application, by acquiring multi-source heterogeneous data of the energy system, the multi-source heterogeneous data including data affecting energy regulation in multiple data sources; performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector; predicting the electricity consumption behavior of users based on the high-order feature vector to obtain a prediction result; determining a regulation planning strategy for the energy system based on the multi-source heterogeneous data and the prediction result; and regulating the power equipment of the energy system based on the prediction result and the regulation planning strategy. In this way, considering the interaction between the power equipment in the energy system, the prediction result and the regulation planning strategy determined by using the acquired multi-source heterogeneous data can more accurately regulate the power equipment, and moreover, through the prediction of the users' electricity consumption behavior and the regulation planning by the multi-source heterogeneous data, the users' needs can be more accurately met, and the energy utilization rate can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present application, and those of ordinary skill in the art can also obtain other drawings according to the provided drawings without creative efforts.

[0045] Figure 1 It is a flowchart of a regulation method for an energy system provided by an embodiment of the present application;

[0046] Figure 2 It is a structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts belong to the scope of protection of the present application.

[0048] Exemplary implementation environment

[0049] The energy system regulation method according to the embodiments of the present application can be executed by an electronic device such as a terminal device or a server. The terminal device can be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The server can be an independent physical server, a server cluster composed of multiple physical servers, or a cloud server capable of performing cloud computing. This method can be implemented by the processor calling the computer-readable program instructions stored in the memory. In this application, the energy system regulation method is explained by taking the server as an example, but it is not limited thereto.

[0050] Exemplary method

[0051] Please refer to Figure 1 , in an exemplary embodiment, a method for regulating an energy system is provided, including:

[0052] Step 101, obtain multi-source heterogeneous data of the energy system, where the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources.

[0053] In some embodiments, the multi-source heterogeneous data can be raw data collected in real time from various devices and sensors. The multi-source heterogeneous data includes: user behavior data, device status data, and environmental parameter data.

[0054] Among them, the user behavior data can be obtained by collecting the activity trajectories and behavior patterns of users through a human infrared sensor, smart home appliances (such as air conditioners, TVs, water heaters), and a smart home gateway. The data forms include: user activity time series (such as the time of entering the room, the time of leaving the room), device start / stop records, etc.

[0055] The device status data can be used to collect the status information of electrical equipment through a smart meter, an energy storage battery management system, and a photovoltaic power generation inverter. The data forms include: device power change rate, operation time sequence, state of charge (SOC) of the energy storage, etc.

[0056] Environmental parameter data can obtain environmental parameters through temperature and humidity sensors, light intensity sensors, and external weather forecast APIs. The data forms include indoor and outdoor temperature, humidity, light intensity, etc.

[0057] The method of collecting multi-source heterogeneous data can be obtained through wired / wireless communication interfaces, edge computing nodes, and time series database storage. Among them, the wired / wireless communication interface uses communication protocols such as RS485, Wi-Fi, or ZigBee to realize data collection between devices and sensors. Deploy edge computing nodes (such as Raspberry Pi or smart gateways) within the home local area network to receive and preliminarily process sensor data in real time. Time series database storage: Store the collected raw data in a time series database (such as InfluxDB) for subsequent analysis and processing.

[0058] After collecting multi-source heterogeneous data, it can also be preprocessed to achieve denoising and anomaly processing. For example, remove sensor noise through moving average filtering or median filtering algorithms. Identify and remove outliers based on statistical methods (such as Z-score detection) or machine learning methods (such as Isolation Forest). Use interpolation methods (such as linear interpolation or KNN interpolation) to fill in missing values caused by sensor failures or communication interruptions.

[0059] Since multi-source heterogeneous data is obtained from multiple data sources, the data dimensions of different data sources may be different. Therefore, it is also possible to perform normalization processing (such as Min-Max normalization or Z-score standardization) on data with different dimensions to ensure that data of each modality is fused on the same scale.

[0060] Step 102: Perform feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector.

[0061] In some embodiments, for the obtained multi-source heterogeneous data, it can be respectively input into independent feature extraction networks to perform deep feature extraction on data of each modality (such as extracting temporal features from temporal channels and spatial correlation features from spatial channels), and the feature vectors of different modalities are weighted and fused through an attention mechanism to generate the final high-order feature vector.

[0062] In an alternative embodiment, the performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector includes:

[0063] Perform feature extraction on the multi-source heterogeneous data to obtain the temporal features and spatial features of data from each data source;

[0064] Perform weighted fusion on the temporal features and spatial features of data from each data source to obtain the high-order feature vector.

[0065] In some embodiments, a dual-channel feature extraction network can be adopted to extract features through a temporal channel and a spatial channel respectively. The temporal channel uses a gated recurrent unit (GRU) or a long short-term memory network (LSTM) to model the temporal features and extract dynamic characteristics such as the device power change rate and the amplitude of the electricity consumption load fluctuation. The spatial channel uses a graph neural network (GNN) or a convolutional neural network (CNN) to model the spatial correlation features and extract the correlation between devices in different regions and the spatial distribution characteristics of environmental parameters. A spatio-temporal attention mechanism is introduced to perform weighted fusion on data of different modalities. The attention in the time dimension captures the importance weights of user behaviors and device states in the time series; the attention in the spatial dimension captures the spatial correlation weights between devices and environmental parameters in different regions. Finally, a high-order feature vector that comprehensively considers temporal and spatial characteristics is generated.

[0066] The finally obtained high-order feature vector includes the following information: user behavior patterns (such as the morning wake-up time and the evening bedtime), device state characteristics (such as the air conditioner power change rate and the energy storage battery state), and environmental parameter characteristics (such as the indoor-outdoor temperature gradient and the light intensity change trend).

[0067] Step 103: Predict the user's electricity consumption behavior based on the high-order feature vector to obtain a prediction result.

[0068] In some embodiments, based on the high-order feature vector, the user's electricity consumption behavior is accurately predicted through a deep learning network, and the prediction result is passed to subsequent functional modules.

[0069] Among them, the user's electricity consumption behavior includes the user's electricity consumption demand, the start and stop times of devices, and the change trend of the load. Furthermore, the prediction result includes the total electricity consumption demand curve of the user in a future period of time; the start and stop time series of key household appliances (such as air conditioners and water heaters); the change trend curve of the grid load; and the correlation analysis report between devices in each region.

[0070] In an alternative embodiment, predicting the user's electricity consumption behavior based on the high-order feature vector to obtain a prediction result includes:

[0071] Inputting the high-order feature vector into a pre-trained prediction model, and mapping the high-order feature vector to a high-dimensional embedding space through the embedding layer of the prediction model and adding position encoding information to obtain a high-order feature matrix;

[0072] Based on the attention layer in the prediction model, determine the attention weights in the time dimension and the spatial dimension of the high-order feature matrix, as well as the spatio-temporal correlation between different features in the high-order feature matrix, and generate an attention output result of the high-order feature matrix based on the attention weights and the spatio-temporal correlation;

[0073] Perform a non-linear transformation on the attention output result to obtain the prediction result.

[0074] In some embodiments, the input high-order feature vectors can be further processed through feature engineering, such as feature selection, feature transformation, and feature combination.

[0075] Feature selection can adopt an automatic feature selection method based on the attention mechanism to screen out the features that have the greatest impact on the prediction of electricity consumption intention. For example, the user's activity trajectory, the change rate of the power of key devices, etc.

[0076] Feature transformation can perform differential processing or sliding window processing on time-series features to extract short-term fluctuation characteristics, and perform graph embedding transformation on spatial correlation features to enhance the ability to express spatial information.

[0077] Feature combination can combine the user behavior pattern with the environmental parameter characteristics to construct a user comfort index (such as comprehensively considering temperature, humidity, and light intensity), and combine the device state characteristics with the environmental parameter characteristics to construct a device operation efficiency index (such as the relationship between air conditioner energy consumption and outdoor temperature).

[0078] For electricity consumption behavior prediction based on the multi-task learning-based electricity consumption intention prediction framework, it mainly includes predicting the user's electricity consumption demand (predicting the total electricity consumption demand of the user in a future period of time (such as the total power demand per hour)), predicting the start and stop time of devices (predicting the start and stop time series of key household appliances (such as air conditioners, water heaters)), and predicting the load change trend (predicting the change trend of the power grid load (such as peak hours, valley hours)).

[0079] Among them, the prediction model can adopt the deep neural network architecture "Spatio-Temporal Attention Enhanced Transformer". Through the embedding layer of the prediction model, the high-order feature vectors are mapped into a high-dimensional embedding space, and position encoding information is added to retain the time-series characteristics. Among them, Positional Encoding is a key technology for injecting the sequential information of sequence data (such as time series, spatial position) into the model. In the "Spatio-Temporal Attention Enhanced Transformer" architecture, its core role is to enable the model to perceive the temporal order (time position) or spatial position relationship of the data, making up for the defect that the Transformer itself does not have the ability to model sequence order.

[0080] The attention layer includes a spatio-temporal attention mechanism and a multi-head self-attention layer. The spatio-temporal attention mechanism pays attention in the time dimension to capture the importance weights of user behavior and device state in the time series. In the space dimension, it pays attention to capture the spatial correlation weights between devices and environmental parameters in different regions.

[0081] The multi - head self - attention layer captures temporal dependencies and spatial correlations at different granularities through multiple parallel attention heads.

[0082] Among them, multi - head parallel computing includes linear projection and attention calculation.

[0083] The linear projection defines learnable weight matrices for each attention head \(i\) (\(i = 1,2,\cdots,h\), where \(h\) is the number of attention heads)) and d k (the attention dimension for each head). Through these weight matrices, linear transformation is performed on the input features to obtain the query, key, and value for each head.

[0084] The attention calculation independently calculates the attention weights for each head and performs weighted summation. Using the scaled dot - product attention mechanism, the attention weights are calculated, and then the weighted sum of the outputs of each head is obtained.

[0085] The attention layer captures temporal dependencies at different granularities: short - term dependencies and long - term dependencies. Some attention heads can focus on capturing short - term temporal dependencies. For example, some heads may focus on changes between adjacent time steps, such as small fluctuations in device power at adjacent moments. The weight matrices of these heads will make them more focused on local time information. Other attention heads can capture long - term temporal dependencies. For example, some heads may focus on changes in power consumption patterns over several days or even weeks. Through the learned weight matrices, connections can be established between long - distance time steps to discover periodic power consumption behaviors.

[0086] Multi - head output concatenation and projection: Concatenate the outputs of all attention heads, and then perform linear projection through a learnable weight matrix to obtain the final output.

[0087] The attention layer also captures spatial correlations at different granularities.

[0088] Input the input feature matrix in the spatial dimension (containing spatial position encoding information) into the multi - head self - attention layer for multi - head parallel computing. Similar to the time dimension, learnable weight matrices are defined for each attention head, and linear transformation is performed on the input features to obtain the query, key, and value matrices for each head, and then the attention weights and outputs are calculated.

[0089] Capturing spatial correlations at different granularities includes local and global correlations. Some attention heads can focus on local spatial correlations, such as the collaborative working relationships between devices in adjacent areas. The weight matrices of these heads will make them pay more attention to the feature interactions between areas that are closer in distance. Other attention heads can capture global spatial correlations, such as the overall electricity consumption pattern correlations between different floors or different functional areas. These heads can establish connections across the entire space and discover more macroscopic spatial patterns.

[0090] Furthermore, the outputs of all spatial dimension attention heads are concatenated and linearly projected to obtain the final spatial dimension output.

[0091] Through the feed-forward neural network layer in the prediction model, the attention output is non-linearly transformed to generate the final task-specific representation.

[0092] Among them, the attention output refers to the feature representation after being processed by the spatio-temporal attention mechanism and the multi-head self-attention layer. Specifically: in the spatio-temporal attention mechanism, through the calculation of attention weights in the time and space dimensions, the input features (such as user behavior, device status, regional environmental parameters) are weighted and aggregated to generate features that integrate temporal importance and spatial correlations. The multi-head self-attention layer further captures different granularities of temporal dependencies and spatial correlations through multiple parallel attention heads, and outputs features containing multi-scale spatio-temporal information.

[0093] The task feature representation can reflect task adaptability, semantic abstraction, and the basis for task execution.

[0094] Task adaptability: For specific tasks (such as electricity consumption intention prediction, load change trend analysis), the general spatio-temporal features of the attention output are transformed into exclusive features that fit the task requirements. For example, feature patterns that are strongly correlated with peak electricity consumption and device start-stop patterns are extracted.

[0095] Semantic abstraction: Through the non-linear transformation of the feed-forward network (such as the ReLU activation function), the attention output is feature-refined to form a more abstract and discriminative representation. For example, features such as the original device power and user activity trajectories are abstracted into high-level semantic representations such as "high-energy consumption behavior patterns" and "regional electricity consumption collaboration patterns".

[0096] Basis for task execution: The final representation directly serves downstream tasks, such as inputting into a classifier for electricity consumption intention classification, or inputting into a regression model to predict load values. It is the key intermediate result connecting feature processing and task execution.

[0097] It can be understood that a dynamic weight adjustment mechanism can also be configured to dynamically adjust the loss function weights of each task according to the importance of real-time data. For example: increasing the weight of the load change trend prediction task during peak electricity consumption periods.

[0098] Among them, the training and optimization strategies of the above prediction model can be as follows:

[0099] First, for the loss function design, a weighted average loss function can be used to comprehensively consider the learning objectives of the three tasks. Among them, the loss weight of each task can be dynamically adjusted according to the importance of real-time data.

[0100] Second, for the optimization algorithm selection, the Adam optimizer combined with a learning rate decay strategy can be adopted for model training, and an early stopping mechanism can be introduced to prevent overfitting.

[0101] Third, for the model evaluation metrics, metrics such as mean squared error (MSE), mean absolute error (MAE), and accuracy (Accuracy) can be used to evaluate the performance of each task.

[0102] Step 104: Based on the multi-source heterogeneous data and the prediction results, determine the regulation and planning strategy of the energy system.

[0103] In some embodiments, the regulation and planning strategy includes: energy supply path planning, vehicle charging and discharging plan, and energy storage scheduling plan. Among them, the energy supply path planning is used to provide the energy supply ratio and power distribution of each energy source within each period of time. The vehicle charging and discharging plan is used to provide the charging time window and discharging strategy of the vehicle. The energy storage scheduling plan is used to provide the charging and discharging time series and power adjustment plan of the energy storage battery.

[0104] Among them, the data used to determine the regulation and planning strategy can include but are not limited to: household electricity demand D(t), photovoltaic power generation Ppv(t), energy storage state, and vehicle charging and discharging demand.

[0105] Among them, the household electricity demand can be to obtain the total electricity demand curve within a future period of time (minute-level resolution), the photovoltaic power generation can be to obtain the real-time photovoltaic power generation and the predicted value for the next 1 hour, the energy storage state can include obtaining the state of charge SOC(t) of the energy storage battery and the charge and discharge power limit Pbatmax, and the vehicle charging and discharging demand can include obtaining the charging demand Pvehcharge(t) and discharging capacity Pvehdischarge(t) of the vehicle.

[0106] In an alternative embodiment, determining the regulation and planning strategy of the energy system based on the multi-source heterogeneous data and the prediction results includes:

[0107] Integrate the multi-source heterogeneous data and the prediction results to obtain integrated data;

[0108] Based on the mixed-integer programming algorithm, with the preset optimization objective as the optimization direction, the integrated data is basically optimized to obtain an initial optimization result, and the initial optimization result satisfies the preset constraint conditions; the optimization objective includes minimizing the electricity cost and minimizing the life loss cost of the vehicle, and the constraint conditions include power balance constraints and charge-discharge constraints of power equipment;

[0109] Taking the initial optimization result as the initial population, the initial optimization result is optimized by using the genetic algorithm with the optimization objective as the optimization direction to obtain the regulation and planning strategy.

[0110] In some embodiments, after obtaining the multi-source heterogeneous data and the prediction results, the data is first integrated, and the data from different sources is integrated into a unified time series format (such as one time point every 5 minutes), and it can also be calibrated in real time. The real-time data is calibrated by the Kalman filter algorithm to reduce sensor noise and prediction deviation. And priority allocation is performed. For example, priorities are assigned to different energy sources according to user needs and system goals (such as economy and stability).

[0111] Among them, the optimization objective function adopted by the optimization objective is:

[0112] min(C cost +α·(1 - η));

[0113]

[0114] Among them, Ccost is the economic cost, including the cost of purchasing electricity from the power grid and the life loss cost of energy storage and vehicles; η is the energy utilization efficiency; α is the weight coefficient used to balance the economic cost and the energy utilization efficiency; C grid (t) represents the power grid electricity price; Pgrid(t) represents the power obtained from the power grid; C bat represents the life loss cost of the energy storage battery; ΔSOC(t) represents the change in the SOC of the energy storage battery.

[0115] Mixed-integer programming (MIP) is used to solve the discrete decision problem of the energy supply path. The constraint conditions include power balance constraints and charge-discharge constraints of power equipment. Among them, the charge-discharge constraints of power equipment can include the charge-discharge constraints of energy storage batteries and the charge-discharge constraints of vehicles.

[0116] Specifically, the power balance constraint can be:

[0117] The charge-discharge constraint of the energy storage battery is:

[0118]

[0119] The charge-discharge constraint of the vehicle can be:

[0120]

[0121] Among them, P grid (t) represents the power obtained from the power grid (a positive value indicates power purchase, and a negative value indicates power sale); P pv (t) represents the photovoltaic power generation (a positive value indicates power generation, and a negative value indicates reverse power feeding); D(t) represents the total household electricity demand (with minute-level resolution); P bat (t) represents the charge and discharge power of the energy storage battery; P veh (t) represents the charge and discharge power of the vehicle; C grid (t) represents the grid electricity price; C bat represents the life loss cost of the energy storage battery; ΔSOC(t) represents the change in the SOC of the energy storage battery; SOC min represents the minimum SOC of the energy storage battery; SOC max represents the maximum SOC of the energy storage battery; P veh，min represents the minimum value of the vehicle charge and discharge power; P veh，max represents the maximum value of the vehicle charge and discharge power. represents the energy storage charging efficiency; represents the energy storage discharge efficiency; Δt represents the time step (e.g., 5 minutes).

[0122] Among them, mixed-integer programming (MIP), as a basic optimization tool, is a mathematical programming method used to solve discrete decision-making problems and provide accurate optimization solutions. In the energy supply path planning, MIP is used to handle problems involving discrete decisions, such as selecting photovoltaic power generation, energy storage, or external grid power supply.

[0123] The inputs of MIP include the household electricity demand D(t), the photovoltaic power generation Ppv(t), the energy storage state SOC(t), and the vehicle charge and discharge demand Pveh(t). The outputs include the accurate solutions of the energy supply ratio and power distribution of each energy source within each period of time.

[0124] Among them, the optimization function adopted by MIP is:

[0125]

[0126] Among them, P grid (t)+P pv (t)+P bat (t)+P veh (t) = D(t); 0 ≤ P bat (t) ≤ P batmax ;; SOC min ≤ SOC(t) ≤ SOC max ; P vehmin ≤ P veh (t) ≤ P vehmax .

[0127] The genetic algorithm (GA) is used for global search of the optimal energy supply strategy and can find the global optimal solution in complex multi-objective optimization scenarios. When the optimization problem involves multiple objectives (such as economy and stability) and there are a large number of possible solutions, GA can effectively search for the global optimal solution. Based on the exact solution provided by MIP, GA further expands the search scope to ensure finding the global optimal solution in complex multi-objective optimization scenarios. The fitness function of GA is based on the optimization objective function of MIP.

[0128] The input of GA takes the optimization result of MIP as the initial population, combines the optimization objectives (such as economy and energy utilization efficiency) and constraint conditions, and outputs the global optimal energy supply path planning.

[0129] The coding method of GA can adopt real number coding, and each individual represents an energy supply strategy. The fitness function can be based on the optimization objective function: min(Ccost + α·(1 - η)). The roulette wheel selection method, single-point crossover, and Gaussian mutation operations are used to complete the optimization process.

[0130] Furthermore, after determining the regulation planning strategy based on the multi-source heterogeneous data and the prediction result, it further includes:

[0131] If it is monitored that the multi-source heterogeneous data has changed, based on the changed data, the particle swarm optimization algorithm is used to update the regulation planning strategy.

[0132] In some embodiments, when the energy supply and demand state changes in real time (such as the fluctuation of photovoltaic power generation and the change of household electricity demand), the particle swarm optimization (PSO) can quickly adjust the energy supply path. The particle swarm optimization is used to dynamically adjust the energy supply strategy in real time and quickly respond to the changes in energy supply and demand.

[0133] When applying PSO, particle initialization is first performed, and each particle represents an energy supply strategy. And the speed update and position update are performed.

[0134] Among them, the speed update: v ik+1 = w·v ik + c1·r1·(p best,i - x ik ) + c2·r2·(g best - x ik );

[0135] The position update: x ik+1 = x ik + v ik+1 ;

[0136] Among them, $v_{ik}$ represents the velocity of the $i$-th particle at the $k$-th iteration; $x_{ik}$ represents the position of the $i$-th particle at the $k$-th iteration; $pbest_i$ represents the historical best position of the $i$-th particle; $gbest$ represents the global best position; $w$ represents the inertia weight; $c_1$ and $c_2$ represent the learning factors; $r_1$ and $r_2$ represent random numbers.

[0137] Based on the global optimal solution provided by GA, PSO performs real-time dynamic adjustment. It utilizes the fast convergence characteristic of PSO to ensure that in a real-time changing environment, the energy supply path planning can quickly adapt and maintain optimality.

[0138] Specifically, in the optimization process of the regulation planning strategy, MIP is used to solve discrete decision problems and generate an initial energy supply path planning. GA is used to perform global search based on MIP to find the global optimal solution. PSO is used to perform real-time dynamic adjustment based on GA to ensure the real-time optimality of the energy supply path planning. Through this collaborative logical relationship, MIP, GA, and PSO work together to ensure that the intelligent energy supply path planning system can not only provide accurate optimization solutions in a complex and dynamic energy management environment but also quickly adapt to real-time changes, ultimately achieving efficient and stable energy supply path planning.

[0139] Furthermore, the obtained regulation planning strategy includes:

[0140] Energy supply path planning: the energy supply ratio and power distribution of each energy source within each period of time;

[0141] Vehicle charging and discharging plan: the charging time window and discharging strategy of the vehicle;

[0142] Energy storage scheduling plan: the charging and discharging time series and power adjustment plan of the energy storage battery;

[0143] Switching timing and path: the best timing and switching path for energy source switching.

[0144] It can be understood that to achieve dynamic adjustment and smooth switching, a dynamic adjustment mechanism can be configured. Through this dynamic adjustment mechanism, the energy supply and demand status is monitored in real time by sensors. When a deviation occurs between the actual value and the predicted value, the optimization algorithm is triggered to recalculate the energy supply path. The rolling optimization strategy is adopted to recalculate the energy supply path every 5 minutes. Priority adjustment: Dynamically adjust the priority of the energy source according to the real-time situation.

[0145] The pre-synchronization control technology is adopted. Before the energy source switching, the voltage phase and frequency are synchronized through the phase-locked loop (PLL) algorithm. The transition buffer mechanism is adopted to introduce a transition buffer stage during the switching process to gradually adjust the power distribution and avoid power mutation. The power change curve is smoothed through the sliding window algorithm. Also, redundant energy sources are set at key energy supply nodes to ensure backup support in case of switching failure.

[0146] Step 105: Regulate the power equipment of the energy system based on the prediction result and the regulation planning strategy.

[0147] In some embodiments, the prediction result may include the total power consumption demand curve of users in a future period of time; the start-stop time series of key household appliances (such as air conditioners, water heaters); the change trend curve of the power grid load. The regulation planning strategy may include real-time energy supply path planning; the specific usage ratios of various energy sources (such as the power grid, photovoltaic, energy storage, etc.); the emergency response plan and the equipment priority list

[0148] By integrating the prediction result and the regulation planning strategy, smooth switching and power regulation between devices are achieved through a pre-synchronization control algorithm, and the optimization result is transmitted to the actuator, and a real-time feedback mechanism is provided to continuously optimize the system performance.

[0149] In an alternative embodiment, regulating the power equipment of the energy system based on the prediction result and the regulation planning strategy includes:

[0150] Normalize the prediction result and the regulation planning strategy to obtain normalized data;

[0151] Based on the normalized data, perform power regulation on the regulation planning strategy to obtain an energy regulation strategy, and the energy regulation strategy can balance the power when the power equipment operates in coordination;

[0152] Control the power equipment to execute according to the energy regulation strategy.

[0153] In some embodiments, after obtaining the prediction result and the regulation planning strategy, they can be standardized first, and the data with different dimensions are normalized (such as Min-Max normalization or Z-score standardization) to ensure that the data of each modality are analyzed on the same scale.

[0154] The optimization process is achieved by configuring a multi-dimensional pre-synchronization control network with a deep reinforcement learning network architecture to obtain an energy regulation strategy.

[0155] Among them, pre-synchronization mainly refers to the pre-coordination and synchronization in the time dimension of multiple key dimensions such as the power consumption demand curve, the energy supply path planning result, and the energy storage state during the operation of the energy system. Its core goal is to precisely match the power supply and demand, the equipment operation state, etc. in time sequence through advance planning and dynamic adjustment, and avoid problems such as insufficient power supply, equipment operation conflicts, or degraded user experience. For example, before the peak power period, pre-coordinate the charge and discharge strategies of energy storage equipment, the switching of energy supply paths, and the prediction of user power consumption demand to achieve multi-dimensional coordinated operation.

[0156] In an alternative embodiment, power adjustment is performed on the regulation planning strategy based on the normalized data to obtain an energy regulation strategy, including:

[0157] Perform state representation on the normalized data to obtain a high-dimensional state matrix;

[0158] Adopt a reinforcement learning policy network to perform action selection on the high-dimensional state matrix based on a reward function to obtain the target action information of the power equipment; the target action information indicates the power adjustment amount of the power equipment, and the reward function indicates the quality of the target action information;

[0159] Adjust the power information in the regulation planning strategy based on the target action information to obtain the energy regulation strategy.

[0160] In some embodiments, the normalized data is mapped to a high-dimensional state space, and time encoding information is added to retain the temporal characteristics. Among them, the state includes: electricity demand curve, energy supply path planning result, energy storage power state, etc.; in the action selection layer, a possible action space is defined (such as adjusting the energy storage charge and discharge power, adjusting the photovoltaic power generation output power, etc.), and the optimal action combination is selected through the policy network.

[0161] When mapping the input data to a high-dimensional state space, time encoding information is specifically added to retain the temporal characteristics. Through time encoding, the model can more accurately capture the time-dependent relationship of the data (such as the change law of the electricity demand curve over time), improve the understanding of the dynamic temporal process, and provide richer temporal information for subsequent decision-making. By defining the action space (such as adjusting the energy storage charge and discharge power, etc.) and selecting the optimal action combination through the policy network, the policy network automatically learns the optimal action strategy from reinforcement learning, can adapt to complex and changeable scenarios (such as the optimal adjustment when the power supply path changes dynamically), and improve the intelligence and flexibility of decision-making.

[0162] Among them, a multi-dimensional reward function can be set, which comprehensively considers power adjustment accuracy, equipment operation stability, and user comfort. Among them, the reward function includes: a negative power deviation term, a positive equipment operation stability term, a positive user comfort term, etc. The reward function weights can be dynamically adjusted according to the real-time operation situation; for example: increasing the weight of power adjustment accuracy during peak power periods;

[0163] A value function evaluation layer can also be configured to use a deep neural network to evaluate the long-term value of taking a certain action in the current state, and introduce temporal difference learning (TD Learning) to update the value function;

[0164] Specifically, the reward function is an immediate feedback signal used to measure the immediate evaluation given by the environment after the agent takes a specific action at a certain moment. It is a short-term, direct reward for a specific action. For example, in a home energy system: when the energy supply mode is switched, if the power deviation is small (meeting the power regulation accuracy), the reward function will give a positive immediate reward; if the operating parameters of the device fluctuate greatly (poor stability), a negative reward will be given. The design of the reward function is directly related to the specific goals of the task (such as dimensions like power regulation accuracy, device stability, user comfort, etc.), and it is the "immediate guidance" for the agent to learn.

[0165] The value function is used to evaluate the expected future long-term cumulative reward after the agent takes a specific action in a certain state, that is, to measure the long-term value of the action. It solves the problem of "long-term planning" in reinforcement learning, enabling the agent to not only focus on immediate rewards but also consider the impact of the current action on subsequent states and rewards. For example: in a home energy system, a parameter adjustment action of a certain energy supply device may not have a high reward at the current moment (such as temporarily reducing the power regulation accuracy for the long-term stability of the device), but through the evaluation of the value function, it is found that this action can make the device operate more stably in the future and reduce greater losses caused by future failures (higher long-term cumulative reward). At this time, the value function value of this action will be relatively high. The value function guides the agent to choose strategies with long-term advantages by predicting long-term benefits, avoiding short-sighted behaviors. During the training process, it is continuously updated in combination with methods such as temporal difference learning (TD Learning) to enable the agent to adapt to environmental changes and optimize decision-making strategies.

[0166] Through the designed multi-dimensional reward function, a multi-dimensional reward function that comprehensively considers power regulation accuracy, device operating stability, and user comfort, including a negative power deviation term, a positive device operating stability term, and a user comfort term, etc. Through multi-dimensional design, different goals are balanced (for example, avoiding sacrificing device stability or user experience in pursuit of power accuracy), enabling the model to learn a comprehensive strategy that better meets the actual needs and improving the overall performance of the system.

[0167] Use a deep neural network to evaluate the long-term value of taking an action in the current state and introduce temporal difference learning (TD Learning) to update the value function. With the powerful fitting ability of the deep neural network, the long-term value can be estimated more accurately. Combined with real-time updating of TD Learning, the model can adapt to environmental changes faster, improving learning efficiency and decision-making quality.

[0168] The real-time adaptability of the dynamic adjustment mechanism dynamically adjusts the weights of the reward function according to the real-time operation conditions (such as increasing the weight of power regulation accuracy during peak power periods). Through dynamic adjustment, the model can prioritize key tasks (such as ensuring stable power supply during peak periods) at different times or in different scenarios, flexibly allocate optimization objectives, and enhance the adaptability of the system to complex real-time environments.

[0169] Furthermore, the training and optimization of the reinforcement learning policy network can be carried out through the following process:

[0170] Use the Deep Q-Network (DQN) combined with the experience replay strategy for training, and introduce a dual-network structure to prevent overfitting. Tune hyperparameters such as the learning rate and discount factor through grid search or Bayesian optimization methods. Use indicators such as power deviation rate, equipment operation stability index, and user comfort score to evaluate the performance of the model;

[0171] The energy regulation strategy obtained through the above embodiments can achieve a smooth transition of equipment start-stop, gradually adjust the power output during the equipment start-stop process, and avoid voltage fluctuations or equipment damage caused by mutations; accurately control power regulation, and adjust the power output of each energy device in real time according to the results of the energy supply path planning; and ensure the coordinated operation of multiple devices to ensure power balance when multiple energy devices (such as energy storage systems, photovoltaic power generation systems, power grids, etc.) operate in coordination.

[0172] Furthermore, the global dynamic optimization results can be transmitted to the actuators (such as energy storage system controllers, photovoltaic power generation inverters, etc.) to achieve real-time control of the home energy system. Real-time monitor the operation status of the home energy system (such as equipment operation status, energy storage power change, etc.) to ensure the accurate execution of the energy supply path planning scheme.

[0173] Configure a feedback and adaptive mechanism to transmit the actual operation data back to the upstream modules (such as the electricity demand prediction model and the energy supply path planning system) to continuously optimize the performance of the entire system. Start the adaptive mechanism in case of deviations or abnormal situations, and quickly adjust the energy supply path and power output to maintain system stability.

[0174] The control method of the energy system of the present application, compared with the traditional home energy management system that usually adopts fixed rules or experience-based control strategies, the energy distribution system based on reinforcement learning can perceive the environmental state in real time and find a dynamic balance among multiple objectives by continuously iterating and optimizing the strategy. At the same time, the combination of the pre-synchronization control technology and the intelligent energy supply path planning system makes the switching of the energy supply mode smoother and more efficient. Through the global dynamic optimization control system and the reinforcement learning algorithm, global optimization is achieved among multiple objectives such as reducing the electricity cost, increasing the energy self-sufficiency rate, and extending the battery life. It can integrate information from various data sources such as vehicle status, household electrical appliances, and environmental sensors, and clean, fuse, and analyze the data through advanced algorithms to provide high-quality data support for subsequent energy distribution and optimization. By real-time monitoring the status of the energy supply equipment and pre-adjusting the parameters of the target energy supply equipment (such as voltage matching, frequency synchronization, etc.) before switching, the continuity and stability of the energy supply are ensured. It can better adapt to the dynamic behavior changes of users and provide more intelligent and personalized energy management services.

[0175] Exemplary device

[0176] Correspondingly, the embodiment of the present application further provides a control device for an energy system, including:

[0177] An acquisition unit, configured to acquire multi-source heterogeneous data of the energy system, where the multi-source heterogeneous data includes data affecting energy control in multiple data sources;

[0178] A feature fusion unit, configured to perform feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector;

[0179] A prediction unit, configured to predict the electricity consumption behavior of a user based on the high-order feature vector to obtain a prediction result;

[0180] A determination unit, configured to determine a control planning strategy for the energy system based on the multi-source heterogeneous data and the prediction result;

[0181] A control unit, configured to control the power equipment of the energy system based on the prediction result and the control planning strategy.

[0182] The control device of the energy system of the present application can integrate information from various data sources such as vehicle status (electricity quantity, location), household electrical equipment, and environmental sensors (such as temperature and humidity, light intensity), and clean, fuse, and analyze these data through advanced algorithms. It can effectively remove noise data and extract key feature information, providing high-quality data support for subsequent energy allocation and optimization. Its advantages lie in wide data sources, precise processing, and strong real-time performance. Based on factors such as historical electricity consumption data, household equipment usage habits, and weather changes, machine learning algorithms are used to accurately predict the electricity demand for a period of time in the future. This model can adapt to the dynamic behavior changes of users and conduct comprehensive analysis by combining information such as photovoltaic power generation and energy storage status. Its advantages are high prediction accuracy, strong adaptability, and good scalability. On the basis of real-time obtaining energy supply and demand information, the optimal energy supply path and strategy are designed through optimization algorithms. This system can flexibly switch between multiple energy supply methods according to factors such as household electricity demand, photovoltaic power generation, energy storage status, and vehicle charging and discharging demand, and ensure the efficiency and stability of energy flow. Its advantages are strong global optimization ability and fast dynamic adjustment speed. Seamless switching between different energy supply methods is achieved in the household energy system. This technology ensures the continuity and stability of energy supply by real-time monitoring the status of energy supply equipment (such as voltage, frequency, phase, etc.) and pre-adjusting the parameters of the target energy supply equipment (such as voltage matching, frequency synchronization, etc.) before switching. Specifically, this technology first real-time monitors the status of the current energy supply equipment and the target energy supply equipment and analyzes their operating characteristics; secondly, based on the characteristics of the target energy supply equipment (such as photovoltaic inverters, energy storage batteries, etc.), its operating parameters are predicted and adjusted in advance to match the current energy supply status; finally, the load is gradually transferred during the switching process to avoid power supply interruption or equipment damage caused by sudden changes. Through pre-synchronization control technology, problems such as power mutations and phase differences caused by energy supply method switching can be effectively reduced, significantly improving the operating stability of the household energy system and the user experience. Its advantages are strong real-time performance, high adaptability, and the ability to greatly reduce energy loss and equipment loss. It can coordinate and manage each sub-module of the household energy system (such as energy allocation, demand prediction, energy supply path planning, etc.) and continuously optimize the overall performance indicators (such as electricity cost, energy self-sufficiency rate, etc.) during real-time operation. This system realizes global optimal control in a complex dynamic environment through multi-objective optimization algorithms and has the ability to quickly respond to external changes. Its advantages are strong global coordination, significant optimization effect, and high operating stability.

[0183] The control device for the energy system provided in this embodiment belongs to the same inventive concept as the control method for the energy system provided in the above embodiments of the present application. It can execute the methods provided in any of the above embodiments of the present application and has the corresponding functional modules and beneficial effects for executing the methods. For the technical details not described in detail in this embodiment, reference can be made to the specific processing content of the control method for the energy system provided in the above embodiments of the present application, which will not be elaborated here.

[0184] In the above control device for the energy system, the functions realized by each unit can be respectively realized by the same or different processors, which is not limited in the embodiments of the present application.

[0185] It should be understood that each functional unit in the above device can be realized in the form of a processor calling software. For example, the device includes a processor, the processor is connected to a memory, and instructions are stored in the memory. The processor calls the instructions stored in the memory to implement any of the above methods or the functions of each unit of the device. The processor can be a general-purpose processor, such as a CPU or a microprocessor, etc., and the memory can be a memory inside the device or a memory outside the device. Alternatively, the units in the device can be realized in the form of a hardware circuit, and the functions of some or all of the units can be realized through the design of the hardware circuit. The hardware circuit can be understood as one or more processors. For example, in one implementation, the hardware circuit is an ASIC, and the functions of some or all of the above units are realized through the design of the logical relationship of the components in the circuit. Another example, in another implementation, the hardware circuit can be realized by a PLD. Taking an FPGA as an example, it can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured through a configuration file to realize the functions of some or all of the above units. All units of the above device can be all realized in the form of a processor calling software, or all realized in the form of a hardware circuit, or some realized in the form of a processor calling software and the remaining part realized in the form of a hardware circuit.

[0186] In the embodiments of the present application, the processor is a circuit with the ability to process signals. In one implementation, the processor can be a circuit with the ability to read and execute instructions, such as a CPU, a microprocessor, a GPU, or a DSP, etc. In another implementation, the processor can realize certain functions through the logical relationship of the hardware circuit, and the logical relationship of the hardware circuit is fixed or can be reconstructed. For example, the processor is a hardware circuit realized by an ASIC or a PLD, such as an FPGA, etc. In a reconfigurable hardware circuit, the process of the processor loading a configuration document to realize the configuration of the hardware circuit can be understood as the process of the processor loading instructions to realize the functions of some or all of the above units. In addition, it can also be a hardware circuit designed for artificial intelligence, which can be understood as a kind of ASIC, such as an NPU, a TPU, a DPU, etc.

[0187] It can be seen that each unit in the above device can be one or more processors (or processing circuits) configured to implement the above method. For example: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0188] In addition, each unit in the above device can be integrated in whole or in part, or can be independently implemented. In one implementation, these units are integrated together and implemented in the form of an SOC. The SOC can include at least one processor for implementing any of the above methods or implementing the functions of each unit of the device. The types of the at least one processor can be different. For example, it includes CPU and FPGA, CPU and artificial intelligence processor, CPU and GPU, etc.

[0189] Exemplary electronic device

[0190] Another embodiment of the present application also proposes an electronic device. As shown in Figure 2 the figure, the device includes:

[0191] a memory 200 and a processor 210;

[0192] Wherein, the memory 200 is connected to the processor 210 and is used for storing programs;

[0193] The processor 210 is used for implementing the regulation method of the energy system disclosed in any of the above embodiments by running the program stored in the memory 200.

[0194] Specifically, the above energy system regulation device may further include: a bus, a communication interface 220, an input device 230, and an output device 240.

[0195] The processor 210, the memory 200, the communication interface 220, the input device 230, and the output device 240 are interconnected through the bus. Among them:

[0196] The bus may include a path for transmitting information between various components of the computer system.

[0197] The processor 210 can be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or can be an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present invention. It can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0198] The processor 210 may include a main processor, and may also include a baseband chip, a modem, etc.

[0199] The memory 200 stores a program for implementing the technical solution of the present invention, and may also store an operating system and other critical services. Specifically, the program may include program code, and the program code includes computer operation instructions. More specifically, the memory 200 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk memory, a flash memory, etc.

[0200] The input device 230 may include devices for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor, etc.

[0201] The output device 240 may include devices for allowing information to be output to a user, such as a display screen, a printer, a speaker, etc.

[0202] The communication interface 220 may include devices of any transceiver type for communicating with other devices or communication networks, such as Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc.

[0203] The processor 210 executes the program stored in the memory 200 and calls other devices, and can be used to implement each step of any one of the energy system regulation methods provided in the above embodiments of the present application.

[0204] Exemplary computer program product and storage medium

[0205] In addition to the above methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions. When the computer program instructions are run by a processor, the processor is caused to execute the steps in the energy system regulation method according to various embodiments of the present application described in any of the above embodiments of this specification.

[0206] The computer program product can be written in any combination of one or more programming languages for executing the program code of the operations of the embodiments of the present application. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, executed as an independent software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0207] In addition, an embodiment of the present application can also be a storage medium on which a computer program is stored. The computer program is executed by a processor to perform the steps in the regulation method of the energy system according to various embodiments of the present application described in any of the above embodiments of this specification. Specifically, the following steps can be implemented:

[0208] Obtain multi-source heterogeneous data of the energy system, where the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources;

[0209] Perform feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector;

[0210] Predict the user's electricity consumption behavior based on the high-order feature vector to obtain a prediction result;

[0211] Determine the regulation planning strategy of the energy system based on the multi-source heterogeneous data and the prediction result;

[0212] Regulate the power equipment of the energy system based on the prediction result and the regulation planning strategy.

[0213] For the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be in other sequences or performed simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0214] It should be noted that each embodiment in this specification is described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0215] The steps in the methods of the embodiments of the present application can be adjusted in sequence, combined, and deleted according to actual needs, and the technical features recorded in the embodiments can be replaced or combined.

[0216] In the devices and terminals of the embodiments of the present application, the modules and sub-modules can be combined, divided, and deleted according to actual needs.

[0217] In several embodiments provided by the present application, it should be understood that the disclosed terminals, devices, and methods can be implemented in other ways. For example, the terminal embodiments described above are only illustrative. For example, the division of modules or sub-modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple sub-modules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling, direct coupling, or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of devices or modules can be in electrical, mechanical, or other forms.

[0218] The modules or sub-modules described as separate components may or may not be physically separated. The components as modules or sub-modules may or may not be physical modules or sub-modules, that is, they can be located in one place, or they can be distributed to multiple network modules or sub-modules. Some or all of the modules or sub-modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0219] In addition, in each embodiment of the present application, the functional modules or sub-modules can be integrated in a processing module, or each module or sub-module can exist physically alone, or two or more modules or sub-modules can be integrated in one module. The above-mentioned integrated modules or sub-modules can be implemented in the form of hardware or in the form of software functional modules or sub-modules.

[0220] Those skilled in the art can further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0221] The steps of the methods or algorithms described in connection with the embodiments disclosed herein may be implemented directly in hardware, in a software unit executed by a processor, or in a combination thereof. The software unit may be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0222] Finally, it should also be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0223] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A control method for an energy system, characterized in that, Including: Obtaining multi-source heterogeneous data of an energy system, where the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources; Performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector; Predicting the electricity consumption behavior of a user based on the high-order feature vector to obtain a prediction result; Determining a regulation planning strategy for the energy system based on the multi-source heterogeneous data and the prediction result; Regulating the power equipment of the energy system based on the prediction result and the regulation planning strategy.

2. The method according to claim 1, wherein The performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector includes: Performing feature extraction on the multi-source heterogeneous data to obtain the time series features and spatial features of the data of each data source; Performing weighted fusion on the time series features and spatial features of the data of each data source to obtain the high-order feature vector.

3. The method according to claim 1, wherein Predicting the electricity consumption behavior of a user based on the high-order feature vector to obtain a prediction result includes: Inputting the high-order feature vector into a pre-trained prediction model, and mapping the high-order feature vector to a high-dimensional embedding space through the embedding layer of the prediction model and adding position encoding information to obtain a high-order feature matrix; Based on the attention layer in the prediction model, determining the attention weights of the time dimension and space dimension of the high-order feature matrix, and the spatio-temporal correlation between different features in the high-order feature matrix, and generating an attention output result of the high-order feature matrix based on the attention weights and the spatio-temporal correlation; Performing a non-linear transformation on the attention output result to obtain the prediction result.

4. The method according to claim 1, wherein Determining a regulation planning strategy for the energy system based on the multi-source heterogeneous data and the prediction result includes: Integrating the multi-source heterogeneous data and the prediction result to obtain integrated data; Based on a mixed integer programming algorithm, taking a preset optimization goal as the optimization direction, performing basic optimization on the integrated data to obtain an initial optimization result, and the initial optimization result satisfies preset constraint conditions; the optimization goal includes minimizing the electricity consumption cost and minimizing the life loss cost of the vehicle, and the constraint conditions include power balance constraints and charge and discharge constraints of power equipment; Taking the initial optimization result as an initial population, and optimizing the initial optimization result using a genetic algorithm with the optimization goal as the optimization direction to obtain the regulation planning strategy.

5. The method according to claim 1 or 4, characterized in that, After determining the regulation planning strategy based on the multi-source heterogeneous data and the prediction result, it further includes: If it is monitored that the multi-source heterogeneous data has changed, updating the regulation planning strategy using a particle swarm optimization algorithm based on the changed data.

6. The method according to claim 1, characterized in that, Regulating the power equipment of the energy system based on the prediction result and the regulation planning strategy includes: Performing normalization processing on the prediction result and the regulation planning strategy to obtain normalized data; Performing power adjustment on the regulation planning strategy based on the normalized data to obtain an energy regulation strategy, and the energy regulation strategy can balance the power when the power equipment operates in coordination; Controlling the power equipment to execute according to the energy regulation strategy.

7. The method according to claim 6, wherein Performing power adjustment on the regulation planning strategy based on the normalized data to obtain an energy regulation strategy, including: Performing state representation on the normalized data to obtain a high-dimensional state matrix; Adopting a reinforcement learning policy network to perform action selection on the high-dimensional state matrix based on a reward function to obtain the target action information of the power equipment; the target action information indicates the power adjustment amount of the power equipment, and the reward function indicates the quality of the target action information; Adjusting the power information in the regulation planning strategy based on the target action information to obtain the energy regulation strategy.

8. A control device for an energy system, characterized in that, Including: An acquisition unit for acquiring multi-source heterogeneous data of an energy system, where the multi-source heterogeneous data includes data affecting energy regulation in multiple data sources; A feature fusion unit for performing feature fusion on the multi-source heterogeneous data to obtain a high-order feature vector; A prediction unit for predicting the electricity consumption behavior of a user based on the high-order feature vector to obtain a prediction result; A determination unit for determining a regulation planning strategy of the energy system based on the multi-source heterogeneous data and the prediction result; A regulation unit for regulating the power equipment of the energy system based on the prediction result and the regulation planning strategy.

9. An electronic device, characterized in that, Including a memory and a processor; The memory is connected to the processor and is used for storing programs; The processor is used for implementing the regulation method of the energy system according to any one of claims 1 to 7 by running the program in the memory.

10. A storage medium, characterized in that, A computer program is stored on the storage medium, and when the computer program is run by the processor, the regulation method of the energy system according to any one of claims 1 to 7 is implemented.

Images