Power analysis methods, devices, equipment and media for hybrid energy industrial parks

Abstract

The present invention belongs to the technical field of electric power big data, and specifically relates to an electric power analysis method, device, equipment and medium for a hybrid energy industrial park. The method of the present invention is based on electricity consumption data and environmental data, and adopts an industrial park power demand prediction model constructed by a generative adversarial network to predict power demand; based on equipment sensor data, an industrial park power equipment failure prediction model based on Lasso and LSTM is adopted to predict equipment status, and a reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain an industrial park power strategy. It solves the limitations of traditional power analysis and management methods under complex energy structures, realizes accurate prediction of park power demand, early warning of equipment failures, and intelligent optimization of power strategy, ensuring the stable power supply and continuity of production operations in the park.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric power big data, and specifically relates to an electric power analysis method, device, equipment and medium for a hybrid energy industrial park. Background Art

[0002] With the continuous upgrading of the global industrial structure and the accelerated advancement of energy transformation, the scale and number of smart industrial parks, as a model of the deep integration of new industrialization and informatization, are growing rapidly. These parks not only integrate diverse production equipment and energy systems, but also widely adopt renewable energy such as wind energy and solar energy, forming a complex hybrid energy supply system. However, although this diversified energy structure improves the flexibility and sustainability of energy use, it also brings unprecedented challenges to the power management and regulation of the park.

[0003] On the one hand, the complementarity and intermittent characteristics between different energy sources significantly increase the complexity and uncertainty of power system operation. For example, the output of wind and solar energy is easily affected by weather conditions, which further highlights the randomness of the "source-load" side, making it difficult to accurately predict and dispatch supply and demand. On the other hand, the diversification and intelligence level of equipment in the park continues to improve, and the demand pattern for electricity has become more complex and changeable, requiring more refined power management and regulation strategies. In addition, with the increasing number and aging of power equipment in the park, the risk of equipment failure also increases. Once a key equipment fails, it will not only affect the normal production and operation of the park, but may also threaten the safe and stable operation of the power grid. Therefore, real-time monitoring of equipment status and fault prediction, so as to take measures in advance for maintenance or replacement, has become an important means to ensure the stable supply of power in the park.

[0004] However, traditional power analysis and management methods are powerless in the face of these challenges. Existing technologies often rely on a single data source and simple models, which makes it difficult to fully and accurately reflect the power operation status and demand changes of the park. Summary of the invention

[0005] The purpose of the present invention is to provide a power analysis method, device, equipment and medium for a hybrid energy industrial park to solve the problem that the existing technology often relies on a single data source and a simple model, and it is difficult to fully and accurately reflect the power operation status and demand changes of the park.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for analyzing power in a hybrid energy industrial park, comprising:

[0008] Obtain electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park;

[0009] Inputting the electricity consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network;

[0010] Input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM;

[0011] Reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park.

[0012] Furthermore, the power consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are obtained, including:

[0013] Based on the data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

[0014] Furthermore, the industrial park power demand prediction model is trained in the following manner:

[0015] Construct a generative adversarial network, including a generator network G and a discriminator network D;

[0016] Get historical electricity consumption data X electricity and historical environmental data X environment ;

[0017] The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Yprediction With real electricity demand data Y real differences;

[0018] The input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction , the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction ;

[0019] The generator network G and the discriminator network D are trained adversarially. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are updated first to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ;

[0020] When the generative adversarial network meets the preset prediction accuracy, the training of the industrial park power demand prediction model is completed.

[0021] Furthermore, the industrial park power equipment fault prediction model is trained in the following manner:

[0022] Acquire historical device sensor data, convert the historical device sensor data into time series data, and perform feature engineering on the time series data;

[0023] Use Lasso regression to select features of time series data after feature engineering, obtain important features that meet the preset correlation, and construct training set data based on important features;

[0024] The training set data is input into the LSTM model for training, and the parameters of the LSTM model are adjusted based on the cross entropy loss function and the gradient descent optimization algorithm. After the preset conditions are met, a trained industrial park power equipment fault prediction model is obtained.

[0025] Furthermore, the reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park, including:

[0026] The power consumption system of the hybrid energy industrial park is modeled as a reinforcement learning environment, including energy type, power demand, equipment status, and power price; the equipment status is determined based on the equipment failure prediction results;

[0027] Convert the power demand and equipment failure prediction results into state representation as the state of the agent in the environment;

[0028] Define the action space of the agent, including adjusting the operating status of the equipment and the power consumption plan;

[0029] Design multi-objective reward functions;

[0030] A deep reinforcement learning model is constructed based on a multi-objective deep deterministic policy gradient algorithm, including a policy network, a Q-value network, and a target network; the parameters of the policy network and the Q-value network are randomly initialized, and the parameters of the target network are initialized to be the same as those of the Q-value network; the environment state, the actions taken by the agent, the rewards obtained, and the next state are collected to form experience and stored in the experience replay buffer; a batch of experience is randomly sampled from the experience replay buffer; for each experience, the target Q-value is calculated through the target network; the policy network and the Q-value network are updated by maximizing the cumulative rewards of multiple objectives; the parameters of the Q-value network are periodically copied to the target network to update the parameters of the target network; after training is completed, multiple power consumption strategies are generated using the policy network;

[0031] Based on multiple electricity usage strategies, an initial population containing multiple electricity usage strategies is randomly generated; each individual in the population corresponds to a electricity usage strategy; iterative crossover and mutation operations are performed on the individuals in the population to generate candidate solutions; a multi-objective optimization algorithm is used to evaluate the fitness of the candidate solutions, and when the preset conditions are met, a Pareto optimal solution set is obtained; and the electricity usage strategy for the industrial park is selected from the Pareto optimal solution set.

[0032] Furthermore, in the step of designing a multi-objective reward function, the multi-objective reward function is expressed as:

[0033]

[0034] in, Indicates that at time t, when the agent is in state and take action is the reward obtained from the environment when ; K is the number of targets, is the weight of the kth target, is the kth target in state Next action The reward function of .

[0035] Furthermore, the policy network and Q-value network are updated by maximizing the cumulative rewards of multiple objectives. The update rule is expressed as:

[0036]

[0037]

[0038] in, For the policy network parameters The gradient of To sample the state in the experience replay buffer and actions expectations; The action output by the policy network; Sampling status and actions The sampling distribution of For the policy network about the state The gradient of is the gradient of the action function with respect to the action; Indicates that the action is equal to In the case of

[0039] is the action value function parameter The gradient of To sample the state in the experience replay buffer and actions ; is the gradient of the action value function with respect to the parameter; is the gradient when the action is equal to the current action; is the gradient of the policy network with respect to the action; is the output of the action-value function;

[0040] The target network parameter update rule is:

[0041] in, is the parameter of the Q value network; are the parameters of the target network; It is a soft update parameter.

[0042] In a second aspect, the present invention provides a power analysis device for a hybrid energy industrial park, comprising:

[0043] A data acquisition module is used to acquire electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park;

[0044] A first prediction module is used to input the power consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, and the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network;

[0045] The second prediction module is used to input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM;

[0046] The strategy formulation module is used to analyze the power demand and equipment failure prediction results using reinforcement learning algorithms to obtain the power consumption strategy for the industrial park.

[0047] Furthermore, the data acquisition module is specifically used for:

[0048] Based on the data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

[0049] Furthermore, in the first prediction module, the industrial park power demand prediction model is trained as follows:

[0050] Construct a generative adversarial network, including a generator network G and a discriminator network D;

[0051] Get historical electricity consumption data X electricity and historical environmental data X environment ;

[0052] The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Y prediction With real electricity demand data Y real differences;

[0053] The input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction, the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction ;

[0054] The generator network G and the discriminator network D are trained adversarially. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are updated first to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ;

[0055] When the generative adversarial network meets the preset prediction accuracy, the training of the industrial park power demand prediction model is completed.

[0056] According to a third aspect of the present invention, there is provided an electronic device, comprising a processor and a memory, wherein the processor is configured to execute a computer program stored in the memory to implement the power analysis method for the hybrid energy industrial park as described above.

[0057] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by a processor, the power analysis method for the hybrid energy industrial park as described above is implemented.

[0058] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0059] The present invention proposes a power analysis method for a hybrid energy industrial park, which uses a generative adversarial network to predict power demand, uses Lasso and LSTM to predict equipment failures, and uses a reinforcement learning algorithm to analyze the power demand and equipment failure prediction results to obtain the power use strategy of the industrial park. It solves the limitations of traditional power analysis and management methods under complex energy structures, realizes accurate prediction of the park's power demand, early warning of equipment failures, and intelligent optimization of power use strategies, and ensures the park's stable power supply and continuity of production operations. The power analysis device, electronic device, and computer-readable storage medium for a hybrid energy industrial park provided by the present invention also solve the problems raised in the background technology section.

[0060] This solution is based on the power demand forecasting model constructed by the generative adversarial network. Through adversarial training, the model's ability to predict complex power demand is improved. It can fully consider historical power consumption data and environmental data, capture the nonlinear relationship between data, and achieve accurate prediction of future power demand, providing a scientific basis for the power dispatch and resource allocation of the park.

[0061] This solution uses the power equipment fault prediction model built by Lasso regression and LSTM model, and realizes early warning of equipment failure through feature selection and time series analysis. It improves the accuracy of fault prediction, reduces maintenance costs, and ensures the stable operation of key equipment in the park.

[0062] This solution achieves real-time and efficient collection of power consumption data, environmental data, and equipment sensor data by deploying a multi-level data collection network (edge ​​network, core network, and cloud network) in the park. The introduction of edge computing nodes reduces data transmission delays, improves the real-time and accuracy of data processing, and provides a data basis for subsequent analysis and prediction.

[0063] This solution uses a deep reinforcement learning model constructed by a multi-objective deep deterministic policy gradient algorithm, taking into account multiple factors such as power demand, equipment status, and power prices, and generates the optimal power strategy by maximizing the cumulative rewards of multiple objectives. This reduces the power consumption cost of the park, improves energy efficiency, and achieves a win-win situation in terms of economic and environmental benefits.

[0064] This solution combines a multi-objective optimization algorithm with a genetic algorithm to generate a set of power usage strategies containing multiple Pareto optimal solutions. Park managers can select the most appropriate power usage strategy from the set based on actual needs and preferences, achieving flexibility and personalized customization of power usage strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0066] Figure 1 This is a flow chart of a method for analyzing power in a hybrid energy industrial park according to an embodiment of the present invention;

[0067] Figure 2 This is a structural block diagram of a power analysis device for a hybrid energy industrial park according to an embodiment of the present invention;

[0068] Figure 3 The present invention is a structural block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0070] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit exemplary embodiments according to the present invention.

[0071] Example 1

[0072] Please refer to Figure 1 , a power analysis method for a hybrid energy industrial park, comprising the following steps:

[0073] S1. Obtain electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park;

[0074] S2. Inputting the electricity consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, and the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network;

[0075] S3. Input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM;

[0076] S4. Use reinforcement learning algorithm to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park.

[0077] The above scheme provides a power analysis method for a hybrid energy industrial park, which uses a generative adversarial network to predict power demand, uses Lasso and LSTM to predict equipment failure, and uses a reinforcement learning algorithm to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park.

[0078] To further explain the present solution, in an optional embodiment, a method for analyzing power in a hybrid energy industrial park is also provided, comprising the following steps:

[0079] S100. Obtain electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park.

[0080] Specifically, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are obtained, including: based on a data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

[0081] More specifically, a data collection network is deployed in the hybrid energy industrial park, and the data collection network is divided into different levels. Based on smart sensors, the power data and environmental data of the hybrid energy industrial park are collected in real time. The smart sensors may include power sensors (voltage, current sensors), temperature and humidity sensors, light sensors, etc., to monitor the power usage and environmental parameters of the hybrid energy industrial park in real time.

[0082] Optionally, the data collection network includes an edge network, a core network and a cloud network. Edge computing nodes and edge gateway devices are deployed in the industrial park for data collection and preprocessing. The edge network is responsible for real-time data collection, preliminary processing and screening, and reduces the amount of data transmitted to the core network and the cloud network. A core network is established as a bridge connecting the edge network and the cloud network, which is responsible for data aggregation, processing and forwarding. The core network adopts a high-bandwidth, low-latency transmission channel to ensure high-speed data transmission and reliability. The cloud network is used for data storage, analysis and application. The cloud network is responsible for in-depth analysis, mining and application of the received data, and provides intelligent power analysis and decision-making support for the industrial park.

[0083] Optionally, the efficient data transmission protocol MQTT is used between the edge network and the core network to achieve fast and reliable data transmission. The high-speed and secure communication protocol HTTPS is used between the cloud network and the core network. Strict security policies and mechanisms are implemented at each network level, including access control, identity authentication, data encryption, etc., to ensure the security and privacy protection of data transmission.

[0084] Among the optional solutions, in order to improve the communication quality, a routing optimization algorithm can be used to dynamically adjust the data transmission path according to the multi-level network topology and real-time traffic conditions, select the optimal transmission path, and reduce the data transmission delay and packet loss rate. The details are as follows:

[0085] (1) Initialize the starting point and the end point, and add the starting point to the openList. The starting point and the end point represent the starting node and the destination node of the data transmission, which can be routers, switches, servers or other network devices in the network. They depend on the network topology and the data transmission path requirements, and can be adjusted according to the actual application scenario.

[0086] (2) Repeat the following steps until the optimal path is found or openList is empty:

[0087] It should be noted that initially, there is only one starting point in the openList. At this time, the starting point is selected for exploration, the neighboring nodes of the starting point are traversed, the neighboring node with the smallest cost function is selected as the target node for the next exploration and added to the openList, and the starting point is added to the closedList; for the target node in the openList, its neighboring nodes are traversed, and the node corresponding to the smallest cost function among the neighboring nodes of the target node is selected as the next target node, the next target node is added to the openList, and the current target node is added to the closedList; iterative exploration is performed until node n is the end point. During the search process, it is necessary to select the next node for exploration according to the preset rules to gradually find the optimal path to the end point. The process of selecting a node is to determine the next node to be explored by evaluating the cost function of each node (such as f(n) = g(n) + h(n)), where f(n) is the total cost of node n; g(n) represents the actual cost from the starting point to node n, that is, the path length from the starting point to node n; h(n) represents the estimated cost from node n to the end point, that is, the straight-line distance from node n to the end point. The node with the smallest cost function value is selected for exploration in the hope of finding the optimal path.

[0088] Select the node n with the smallest f(n) from openList for exploration;

[0089] If node n is the end point, the path search is completed and the path is returned;

[0090] Otherwise, move node n from openList to closedList and traverse the adjacent nodes of node n:

[0091] If the adjacent node is not in openList and closedList, calculate g(n), h(n) and f(n) and add the adjacent node to openList;

[0092] If the adjacent node is in the openList, compare the cost of the current path with the previous path, and update the path and cost;

[0093] (3) Obtain the optimal path based on the paths recorded in closedList.

[0094] S200. Input the electricity consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park electricity demand prediction model, and the industrial park electricity demand prediction model outputs the electricity demand of the industrial park; wherein the industrial park electricity demand prediction model is constructed based on a generative adversarial network.

[0095] Specifically, the industrial park power demand prediction model is trained in the following manner:

[0096] Construct a generative adversarial network, including a generator network G and a discriminator network D;

[0097] Get historical electricity consumption data X electricity and historical environmental data X environment ;

[0098] The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Y prediction With real electricity demand data Y real The difference between L gen :

[0099]

[0100] in, Represents historical electricity consumption data X electricity and environmental dataX environment expectations;

[0101] The discriminator network D evaluates the authenticity of the prediction results generated by the generator network and helps the generator network to continuously optimize the generated results; the input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction , the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction , that is, maximize the discriminative ability:

[0102]

[0103] in, L dis For the ability to discern; Real power demand data Y real expectations;

[0104] The generator network G and the discriminator network D are trained against each other to continuously improve the prediction ability of the generator network and enable it to realistically generate future power demand data. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are first updated to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ;

[0105] When the generative adversarial network meets the preset prediction accuracy, the training of the industrial park power demand prediction model is completed.

[0106] S300. Input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein, the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM.

[0107] Specifically, the industrial park power equipment fault prediction model is trained in the following manner:

[0108] Acquire historical device sensor data, convert the historical device sensor data into time series data, and perform feature engineering on the time series data;

[0109] Use Lasso regression to select features of time series data after feature engineering, obtain important features that meet the preset correlation, and construct training set data based on important features;

[0110] The goal of Lasso regression is to minimize the following loss function:

[0111]

[0112] Where N is the number of samples, p is the number of features, is the target value of the ith sample, is the jth feature of the ith sample, is the coefficient of the jth feature, is the intercept term and λ is the regularization parameter.

[0113] The training set data is input into the LSTM model for training, and the parameters of the LSTM model are adjusted based on the cross entropy loss function and the gradient descent optimization algorithm. After the preset conditions are met, a trained industrial park power equipment fault prediction model is obtained;

[0114] Optionally, the hidden state of the last LSTM unit is used, and the cross entropy loss function is used to calculate the loss between the predicted result and the true label. When the training set data is input into the LSTM model for training, the model parameters are adjusted based on the cross entropy loss function and the gradient descent optimization algorithm; the model is evaluated using the validation set, and the model's hyperparameters are adjusted based on the evaluation results to improve the model's generalization ability and prediction accuracy.

[0115] S400. Use reinforcement learning algorithm to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park.

[0116] Specifically, the reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park, including:

[0117] The power consumption system of the hybrid energy industrial park is modeled as a reinforcement learning environment, including factors such as energy type, power demand, equipment status, and power price, so that the agent can learn and formulate the best power consumption strategy; among them, the equipment status is determined based on the equipment failure prediction results;

[0118] Convert power demand and equipment failure prediction results into state representation as the state of the agent in the environment for decision making;

[0119] Define the action space of the agent, including adjusting the operating status of the equipment and the power consumption plan;

[0120] Design a multi-objective reward function to evaluate the agent's behavior at each time step to guide the agent to learn and formulate the best power consumption strategy. The reward function can take into account factors such as power cost, equipment operation efficiency, and campus power balance.

[0121] A deep reinforcement learning model is constructed based on a multi-objective deep deterministic policy gradient algorithm, including a policy network, a Q-value network, and a target network; the parameters of the policy network and the Q-value network are randomly initialized, and the parameters of the target network are initialized to be the same as those of the Q-value network; the environment state, the actions taken by the agent, the rewards obtained, and the next state are collected to form experience and stored in the experience replay buffer; a batch of experience is randomly sampled from the experience replay buffer; for each experience, the target Q-value is calculated through the target network; the policy network and the Q-value network are updated by maximizing the cumulative rewards of multiple objectives; the parameters of the Q-value network are periodically copied to the target network to update the parameters of the target network; after training is completed, multiple power consumption strategies are generated using the policy network;

[0122] Based on multiple electricity usage strategies, an initial population containing multiple electricity usage strategies is randomly generated; each individual in the population corresponds to a electricity usage strategy; iterative crossover and mutation operations are performed on the individuals in the population to generate candidate solutions; a multi-objective optimization algorithm is used to evaluate the fitness of the candidate solutions, and when the preset conditions are met, a Pareto optimal solution set is obtained; and the electricity usage strategy for the industrial park is selected from the Pareto optimal solution set.

[0123] Specifically, in the step of designing a multi-objective reward function, the importance and weight of different indicators are considered to achieve multi-objective optimization. The reward function comprehensively considers the goals of minimizing cost, maximizing efficiency, and optimizing environmental protection. The multi-objective reward function is expressed as:

[0124]

[0125] in, Indicates that at time t, when the agent is in state and take action is the reward obtained from the environment when ; K is the number of targets, is the weight of the kth target, is the kth target in state Next action The reward function of .

[0126] Specifically, the policy network and Q-value network are updated by maximizing the cumulative rewards of multiple objectives. The update rule is expressed as:

[0127]

[0128]

[0129] in, For the policy network parameters The gradient of To sample the state in the experience replay buffer and actions expectations; The action output by the policy network; Sampling status and actions The sampling distribution of For the policy network about the state The gradient of is the gradient of the action function with respect to the action; Indicates that the action is equal to In the case of

[0130] is the action value function parameter The gradient of To sample the state in the experience replay buffer and actions ; is the gradient of the action value function with respect to the parameter; is the gradient when the action is equal to the current action; is the gradient of the policy network with respect to the action; is the output of the action-value function;

[0131] The target network parameter update rule is:

[0132] in, is the parameter of the Q value network; are the parameters of the target network; It is a soft update parameter.

[0133] Example 2

[0134] like Figure 2 As shown, based on the same inventive concept as the above embodiment, the present invention also provides a power analysis device for a hybrid energy industrial park, comprising:

[0135] A data acquisition module is used to acquire electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park;

[0136] A first prediction module is used to input the power consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, and the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network;

[0137] The second prediction module is used to input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM;

[0138] The strategy formulation module is used to analyze the power demand and equipment failure prediction results using reinforcement learning algorithms to obtain the power consumption strategy for the industrial park.

[0139] Furthermore, the data acquisition module is specifically used for:

[0140] Based on the data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

[0141] Furthermore, in the first prediction module, the industrial park power demand prediction model is trained as follows:

[0142] Construct a generative adversarial network, including a generator network G and a discriminator network D;

[0143] Get historical electricity consumption data X electricity and historical environmental data X environment ;

[0144] The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Y prediction With real electricity demand data Y real differences;

[0145] The input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction, the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction ;

[0146] The generator network G and the discriminator network D are trained adversarially. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are updated first to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ;

[0147] When the generative adversarial network meets the preset prediction accuracy, the training of the industrial park power demand prediction model is completed.

[0148] Example 3

[0149] like Figure 3 As shown, the present invention also provides an electronic device 100 for implementing a power analysis method for a hybrid energy industrial park according to the above embodiment;

[0150] The electronic device 100 includes a memory 101 , at least one processor 102 , a computer program 103 stored in the memory 101 and executable on the at least one processor 102 , and at least one communication bus 104 .

[0151] The memory 101 can be used to store a computer program 103 . The processor 102 implements the steps of a power analysis method for a hybrid energy industrial park in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101 .

[0152] The memory 101 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 may include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0153] At least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0154] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a power analysis method for a hybrid energy industrial park, and the processor 102 can execute the plurality of instructions to implement:

[0155] Obtain electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park;

[0156] Inputting the electricity consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network;

[0157] Input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM;

[0158] Reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park.

[0159] Example 4

[0160] If the module / unit integrated in the electronic device 100 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media may include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0161] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0162] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0163] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0164] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0165] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A power analysis method for a hybrid energy industrial park, characterized in that: include: Obtain electricity consumption data, environmental data, and equipment sensor data of the hybrid energy industrial park; Inputting the electricity consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network; Input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM; Reinforcement learning algorithms are used to analyze power demand and equipment failure prediction results to obtain power consumption strategies for industrial parks. The industrial park power demand prediction model is trained in the following way: Construct a generative adversarial network, including a generator network G and a discriminator network D; Get historical electricity consumption data X electricity and historical environmental data X environment ; The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Y prediction With real power demand data Y real differences; The input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction , the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction ; The generator network G and the discriminator network D are trained adversarially. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are updated first to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ; When the generative adversarial network meets the preset prediction accuracy, the industrial park power demand prediction model training is completed; The industrial park power equipment fault prediction model is trained in the following manner: Acquire historical device sensor data, convert the historical device sensor data into time series data, and perform feature engineering on the time series data; Use Lasso regression to select features of time series data after feature engineering, obtain important features that meet the preset correlation, and construct training set data based on important features; The training set data is input into the LSTM model for training, and the parameters of the LSTM model are adjusted based on the cross entropy loss function and the gradient descent optimization algorithm. After the preset conditions are met, a trained industrial park power equipment fault prediction model is obtained; The reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park, including: The power consumption system of the hybrid energy industrial park is modeled as a reinforcement learning environment, including energy type, power demand, equipment status, and power price; the equipment status is determined based on the equipment failure prediction results; Convert the power demand and equipment failure prediction results into state representation as the state of the agent in the environment; Define the action space of the agent, including adjusting the operating status of the equipment and the power consumption plan; Design multi-objective reward functions; A deep reinforcement learning model is constructed based on a multi-objective deep deterministic policy gradient algorithm, including a policy network, a Q-value network, and a target network; the parameters of the policy network and the Q-value network are randomly initialized, and the parameters of the target network are initialized to be the same as those of the Q-value network; the environment state, the actions taken by the agent, the rewards obtained, and the next state are collected to form experience and stored in the experience replay buffer; a batch of experience is randomly sampled from the experience replay buffer; for each experience, the target Q-value is calculated through the target network; the policy network and the Q-value network are updated by maximizing the cumulative rewards of multiple objectives; the parameters of the Q-value network are periodically copied to the target network to update the parameters of the target network; after training is completed, multiple power consumption strategies are generated using the policy network; Based on multiple electricity usage strategies, an initial population containing multiple electricity usage strategies is randomly generated; each individual in the population corresponds to a electricity usage strategy; iterative crossover and mutation operations are performed on the individuals in the population to generate candidate solutions; a multi-objective optimization algorithm is used to evaluate the fitness of the candidate solutions, and when the preset conditions are met, a Pareto optimal solution set is obtained; and the electricity usage strategy for the industrial park is selected from the Pareto optimal solution set.

2. The power analysis method for a hybrid energy industrial park according to claim 1, characterized in that: Obtain electricity consumption data, environmental data, and equipment sensor data for the hybrid energy industrial park, including: Based on the data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

3. A power analysis device for a hybrid energy industrial park, characterized in that: include: The data acquisition module is used to obtain the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park; A first prediction module is used to input the power consumption data and environmental data of the hybrid energy industrial park into a pre-trained industrial park power demand prediction model, and the industrial park power demand prediction model outputs the power demand of the industrial park; wherein the industrial park power demand prediction model is constructed based on a generative adversarial network; The second prediction module is used to input the equipment sensor data of the hybrid energy industrial park into a pre-trained industrial park power equipment failure prediction model, and the industrial park power equipment failure prediction model outputs the equipment failure prediction results of the industrial park; wherein the industrial park power equipment failure prediction model is constructed based on Lasso and LSTM; The strategy formulation module is used to analyze the power demand and equipment failure prediction results using reinforcement learning algorithms to obtain the power consumption strategy of the industrial park; In the first prediction module, the industrial park power demand prediction model is trained as follows: Construct a generative adversarial network, including a generator network G and a discriminator network D; Get historical electricity consumption data X electricity and historical environmental data X environment ; The generator network G takes the historical electricity consumption data X electricity and historical environmental data X environment As input, generate power demand forecasts for a period of time in the future Y prediction ; The goal of the generator network G is to minimize the power demand forecast result Y prediction With real power demand data Y real differences; The input of the discriminator network D is the real power demand data Y real The power demand forecast results generated by the generator network G Y prediction , the goal of the discriminator network D is to maximize the correct distinction of real power demand data Y real and power demand forecast results Y prediction ; The generator network G and the discriminator network D are trained adversarially. For each round of training, the generator network G and the discriminator network D are trained alternately. During training, the parameters of the generator network G are updated first to reduce the generation error. L gen , and then update the parameters of the discriminator network D to increase the discriminative ability L dis ; When the generative adversarial network meets the preset prediction accuracy, the industrial park power demand prediction model training is completed; In the first prediction module, the industrial park power equipment failure prediction model is trained in the following manner: Acquire historical device sensor data, convert the historical device sensor data into time series data, and perform feature engineering on the time series data; Use Lasso regression to select features of time series data after feature engineering, obtain important features that meet the preset correlation, and construct training set data based on important features; The training set data is input into the LSTM model for training, and the parameters of the LSTM model are adjusted based on the cross entropy loss function and the gradient descent optimization algorithm. After the preset conditions are met, a trained industrial park power equipment fault prediction model is obtained; In the strategy formulation module, a reinforcement learning algorithm is used to analyze the power demand and equipment failure prediction results to obtain the power consumption strategy of the industrial park, including: The power consumption system of the hybrid energy industrial park is modeled as a reinforcement learning environment, including energy type, power demand, equipment status, and power price; the equipment status is determined based on the equipment failure prediction results; Convert the power demand and equipment failure prediction results into state representation as the state of the agent in the environment; Define the action space of the agent, including adjusting the operating status of the equipment and the power consumption plan; Design multi-objective reward functions; A deep reinforcement learning model is constructed based on a multi-objective deep deterministic policy gradient algorithm, including a policy network, a Q-value network, and a target network; the parameters of the policy network and the Q-value network are randomly initialized, and the parameters of the target network are initialized to be the same as those of the Q-value network; the environment state, the actions taken by the agent, the rewards obtained, and the next state are collected to form experience and stored in the experience replay buffer; a batch of experience is randomly sampled from the experience replay buffer; for each experience, the target Q-value is calculated through the target network; the policy network and the Q-value network are updated by maximizing the cumulative rewards of multiple objectives; the parameters of the Q-value network are periodically copied to the target network to update the parameters of the target network; after training is completed, multiple power consumption strategies are generated using the policy network; Based on multiple electricity usage strategies, an initial population containing multiple electricity usage strategies is randomly generated; each individual in the population corresponds to a electricity usage strategy; iterative crossover and mutation operations are performed on the individuals in the population to generate candidate solutions; a multi-objective optimization algorithm is used to evaluate the fitness of the candidate solutions, and when the preset conditions are met, a Pareto optimal solution set is obtained; and the electricity usage strategy for the industrial park is selected from the Pareto optimal solution set.

4. The power analysis device for a hybrid energy industrial park according to claim 3, characterized in that: The data acquisition module is specifically used for: Based on the data collection network arranged in the hybrid energy industrial park, the electricity consumption data, environmental data and equipment sensor data of the hybrid energy industrial park are collected; wherein, the data collection network includes an edge network, a core network and a cloud network; the edge network includes smart sensors, edge computing nodes and edge gateway devices; the electricity consumption data and environmental data collected by the smart sensors are pre-processed in the edge computing nodes and then transmitted to the core network through the edge gateway device, and the core network integrates the received data and sends it to the cloud network.

5. An electronic device, characterized in that: It comprises a processor and a memory, wherein the processor is used to execute a computer program stored in the memory to implement the power analysis method for a hybrid energy industrial park as described in claim 1 or 2.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores at least one instruction, and when the at least one instruction is executed by the processor, the power analysis method for the hybrid energy industrial park as described in claim 1 or 2 is implemented.

Images