Distributed energy output prediction method, system and device and computing equipment

Abstract

The invention discloses a distributed energy output prediction method, system and device and computing equipment, and belongs to the technical field of data processing, and the method comprises the steps: receiving the equipment state information of all distributed energy equipment at a plurality of time points; mapping the device state information of each distributed energy device at the plurality of time points to a virtual space to obtain a digital twinborn body of each distributed energy device; according to the equipment state information of the target digital twinborn at the multiple time points, the output condition of the target digital twinborn at the current time point is predicted through a pre-trained energy output prediction model, and the energy output prediction model is a pulse neural network comprising a gating unit; in the training process, the gating unit determines and updates model parameters according to the input information of the current time step and the membrane potentials of the first n time steps. According to the method, observability, measurability and controllability of different types of distributed energy states are realized, accurate distributed energy output prediction is realized, and system power consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a distributed energy output prediction method, system, device and computing equipment. Background Art

[0002] Distributed energy is not based on a single energy supply, but rather a combination of natural gas, wind, light, geothermal heat, water-source heat and other energy sources that are deployed in accordance with local conditions. Compared with traditional energy supply methods, distributed energy can not only realize the comprehensive utilization of surrounding waste heat resources, but also realize intelligent scheduling between national power grids, municipal gas, municipal heat and other backbone networks, which has a very high development significance.

[0003] However, since the output of distributed energy is random and intermittent, in order to ensure that distributed energy can supply energy normally, it is necessary to analyze the use effect of distributed energy before use, such as predicting the output of distributed energy. However, it is difficult to predict the output of distributed energy with existing technology. Summary of the invention

[0004] To this end, the present invention provides a distributed energy output prediction method, and also provides a distributed energy output prediction system, a distributed energy output prediction device, a computing device and a computer-readable storage medium, aiming to at least to some extent solve the technical problem in the related technology that the distributed energy output cannot be predicted.

[0005] To achieve the above object, a first embodiment of the present invention provides a distributed energy output prediction method, which is applied to a cloud platform. The method includes:

[0006] Receive device status information of each distributed energy device at multiple time points;

[0007] Mapping the device status information of each distributed energy device at multiple time points to the virtual space to obtain the digital twin of each distributed energy device;

[0008] According to the device status information of the target digital twin at multiple time points, the output of the target digital twin at the current time point is predicted by a pre-trained energy output prediction model, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

[0009] According to an embodiment of the present invention, the device status information includes actual output and output impact information. According to the device status information of the target digital twin at multiple time points, the output of the target digital twin at the current time point is predicted by a pre-trained energy output prediction model, including:

[0010] Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information;

[0011] Determine a first actual output corresponding to the first output impact information;

[0012] The first actual output, the first output impact information and the current output impact information are input into a pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

[0013] According to one embodiment of the present invention, after receiving the device status information of each distributed energy device at multiple time points, the method further includes:

[0014] According to the different distributed energy devices, the device status information of each distributed energy device at multiple time points is managed in pieces, and the device status information of each distributed energy device is updated in real time.

[0015] According to one embodiment of the present invention, after predicting the output of the target digital twin at the current time point through a pre-trained energy output prediction model according to the device status information of the target digital twin at multiple time points, it also includes:

[0016] Obtain output demand information of target distributed energy;

[0017] Generate maintenance strategies based on output demand information and output conditions;

[0018] A maintenance instruction carrying a maintenance strategy is sent to a terminal device to perform maintenance on multiple distributed energy devices.

[0019] According to one embodiment of the present invention, during the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, including:

[0020]

[0021] Among them, α represents the model parameter, σ represents the activation function, and I t represents the input of the current time step, represents the membrane potential at time step ti, W m and W inBoth represent the network weights of the energy output prediction model, FC represents fully connected processing, and conv represents convolution processing.

[0022] To achieve the above-mentioned object, a second aspect of the present invention provides a distributed energy output prediction system, the system comprising a terminal device and a cloud platform;

[0023] The terminal device is configured to send device status information of each distributed energy device at multiple time points to the cloud platform;

[0024] The cloud platform is configured to map the device status information of each distributed energy device at multiple time points to the virtual space, obtain the digital twin of each distributed energy device, and predict the output of the target digital twin at the current time point through a pre-trained energy output prediction model based on the device status information of the target digital twin at multiple time points, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters based on the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

[0025] According to an embodiment of the present invention, the device status information includes actual output and output impact information, and the cloud platform includes an energy output prediction module, which is configured as follows:

[0026] Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information;

[0027] Determine a first actual output corresponding to the first output impact information;

[0028] The first actual output, the first output impact information and the current output impact information are input into a pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

[0029] According to one embodiment of the present invention, the cloud platform further includes a data mapping library, a logical entity space, a data warehouse module, and a resource slicing module;

[0030] a data mapping library configured to configure data mapping rules for mapping physical entities to virtual space;

[0031] The logical entity space is configured to map the device status information of each distributed energy device at multiple time points to the virtual space according to the data mapping rule to obtain the digital twin of each distributed energy device;

[0032] A data warehouse module is configured to receive and store device status information of each distributed energy device at multiple time points;

[0033] The resource slicing module is configured to manage the device status information of each distributed energy device at multiple time points in slices according to the different distributed energy devices, and update the device status information of each distributed energy device in real time.

[0034] According to one embodiment of the present invention, the cloud platform further includes a decision module, which is configured to:

[0035] Obtain output demand information of target distributed energy;

[0036] Generate maintenance strategies based on output demand information and output conditions;

[0037] A maintenance instruction carrying a maintenance policy is sent to the end-side device.

[0038] According to one embodiment of the present invention, the terminal side device is further configured as follows:

[0039] Receive maintenance instructions and perform maintenance on multiple distributed energy devices based on maintenance strategies;

[0040] The device status information of each distributed energy device after maintenance is obtained and sent to the cloud platform so that the cloud platform can update the device status information of each distributed energy device.

[0041] According to one embodiment of the present invention, the system further includes a side device, and the side device is configured to:

[0042] The device status information of each distributed energy device at multiple time points sent by the receiving-side device is stored and forwarded to the cloud platform.

[0043] To achieve the above-mentioned purpose, a third aspect of the present invention provides a distributed energy output prediction device, the device comprising:

[0044] A receiving unit, used to receive device status information of each distributed energy device at multiple time points;

[0045] A mapping unit, used to map the device status information of each distributed energy device at multiple time points to a virtual space to obtain a digital twin of each distributed energy device;

[0046] The prediction unit is used to predict the output of the target digital twin at the current time point according to the device status information of the target digital twin at multiple time points through a pre-trained energy output prediction model, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

[0047] To achieve the above-mentioned purpose, a computing device is provided in a fourth aspect of an embodiment of the present invention, and the computing device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the distributed energy output prediction method as described in the first aspect above is implemented.

[0048] To achieve the above-mentioned purpose, a fifth aspect of the present invention provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the distributed energy output prediction method as described in the first aspect above is implemented.

[0049] The distributed energy output prediction method provided in an embodiment of the present invention is applied to a cloud platform, receiving device status information of each distributed energy device at multiple time points; mapping the device status information of each distributed energy device at multiple time points to a virtual space to obtain a digital twin of each distributed energy device; based on the device status information of the target digital twin at multiple time points, predicting the output of the target digital twin at the current time point through a pre-trained energy output prediction model, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters based on the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0. The above method uses digital twin technology to realize the logical mapping of distributed energy from physical space to virtual space, realizes the observable, measurable and controllable status of different types of distributed energy, and uses a pulse neural network including a gating unit to predict the output of distributed energy. During the training process, the model parameters are determined and updated according to the input information of the current time step and the membrane potential of the previous n time steps, which can improve the overall memory capacity of the model, thereby alleviating the gradient disappearance and explosion problems that may occur during model training, and improves the prediction performance of the model. In addition, the power consumption of the pulse neural network is low, so the energy output prediction model is used for output prediction, which not only improves the prediction accuracy and reliability, but also reduces the system power consumption.

[0050] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a structural schematic diagram of a distributed energy output prediction system provided according to an embodiment of the present invention;

[0052] Figure 2 is a structural schematic diagram of another distributed energy output prediction system provided according to an embodiment of the present invention;

[0053] Figure 3 is a structural schematic diagram of an energy output prediction model provided according to an embodiment of the present invention;

[0054] Figure 4 is a flow chart of a distributed energy output prediction method provided according to an embodiment of the present invention;

[0055] Figure 5 is a structural schematic diagram of a distributed energy output prediction device provided according to an embodiment of the present invention;

[0056] Figure 6 It is a structural schematic diagram of a computing device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be configured to explain the present invention, and are not to be construed as limiting the present invention.

[0058] Foreign research on the analysis of distributed energy characteristics focuses on the uncertainty characteristics of distributed energy output. In terms of photovoltaic output characteristics, the impact of cloud movement and different configurations of photovoltaic arrays on photovoltaic output characteristics is studied; the accuracy of deterministic and stochastic photovoltaic prediction models in day-ahead prediction is analyzed, and the photovoltaic prediction data is made more accurate by defining the "clear sky performance index"; the probabilistic power flow algorithm is used to evaluate the impact of photovoltaic power generation uncertainty on the performance of the transmission system, considering the correlation between input random variables and the correlation between adjacent photovoltaic resources. In terms of energy storage device characteristics, by considering the impact of the characteristics of the energy storage system on the distribution system, its impact on the operation plan, reliability and safety of the distribution system is studied. In terms of electric vehicles, by analyzing the impact of electric vehicles connected to charging piles on the power grid, the charging strategy for electric vehicles is optimized to improve the economy and safety of electric vehicles and system operation. In terms of flexible adjustable loads, in order to accurately describe the working characteristics of air-conditioning loads, based on the single air-conditioning load model, the Monte Carlo simulation method is used to establish a load aggregation model based on the distribution characteristics of various parameters, and a multi-region load aggregation model is established based on the spatial differences of key parameters. A random energy storage charging parameter model of aggregated load is proposed. A priority stack hysteresis algorithm based on non-interrupt control is used to track the regulation signal. The charging / discharging time of the energy storage model is calculated according to the change of indoor temperature after the air conditioner is turned on / off.

[0059] Foreign research on the characteristics analysis, operation regulation capability and modeling technology of distributed energy in distribution network substations is more focused on the characteristics analysis and operation regulation capability evaluation of one type of distributed energy. There is less research on the operation regulation capability of distributed energy in heterogeneous, multi-period coupled and uncertain distribution network substations, and the establishment of a distributed energy evaluation system still needs to be further improved. Therefore, it is necessary to establish a set of methods that comprehensively consider various influencing factors and conduct characteristics analysis and operation regulation capability evaluation of multiple distributed energy sources.

[0060] To this end, an embodiment of the present invention provides a distributed energy output prediction method, which can solve the above technical problems. For its specific implementation, please refer to the relevant description of the following embodiments.

[0061] The following describes the distributed energy output prediction method, system, apparatus, computing device and computer-readable storage medium provided by the embodiments of the present invention with reference to the accompanying drawings.

[0062] Figure 1 1 is a structural diagram of a distributed energy output prediction system provided according to an embodiment of the present invention. The system includes a terminal device 101 and a cloud platform 102.

[0063] In a possible implementation, the terminal device 101 can communicate directly with the cloud platform 102, and the terminal device 101 can send the device status information of each distributed energy device at multiple time points to the cloud platform 102. That is, this solution adopts the cloud-terminal structure collaborative working mode, and uses digital twin technology to realize the logical mapping of distributed energy from physical space to virtual space and the output prediction of distributed energy.

[0064] In an embodiment of the present invention, the terminal device 101 is configured to send the device status information of each distributed energy device at multiple time points to the cloud platform 102; the cloud platform 102 is configured to map the device status information of each distributed energy device at multiple time points to the virtual space to obtain the digital twin of each distributed energy device, and predict the output of the target digital twin at the current time point through a pre-trained energy output prediction model according to the device status information of the target digital twin at multiple time points, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

[0065] In a specific implementation, the end-side device 101 is the sum of physical entities, including multiple distributed energy devices and hardware facilities such as sensors. Sensors can be installed at key locations of distributed energy devices to obtain comprehensive and accurate device status information. The device status information may include distributed energy data information reported by distributed energy devices themselves (such as device configuration data and operation data of distributed energy devices) and sensor data information collected by sensors (such as meteorological data), and this information may be in units of time points, that is, the sensor data information may be the data information collected by the sensor at each time point.

[0066] As an example, distributed energy equipment may include photovoltaic equipment, wind power equipment, and energy storage equipment. The sensor data information collected by the sensors installed on the photovoltaic equipment includes irradiance, temperature, wind speed, humidity, etc., the sensor data information collected by the sensors installed on the wind power equipment includes operating temperature, altitude, wind speed, survival temperature, wind direction, humidity, atmospheric pressure, etc., and the sensor data information collected by the sensors installed on the energy storage equipment includes battery capacity, operating temperature, etc. The distributed energy data information reported by the photovoltaic equipment includes photovoltaic arrays, operating data, azimuth, etc., and the distributed energy data information reported by the wind power equipment includes wind wheel diameter, rated power, etc. In addition, the distributed energy data information reported by each distributed energy device may also include the actual output of the distributed energy device, that is, the actual efficiency of the distributed energy device.

[0067] In a specific implementation, after the cloud platform 102 receives the device status information of each distributed energy device at multiple time points, it uses the digital twin technology to map the device status information of each distributed energy device at multiple time points to the virtual space, and obtains the digital twin of each distributed energy device. The digital twin of each distributed energy device includes the device status information of the distributed energy device at multiple time points. For any digital twin, such as the target digital twin, the output of the target digital twin at the current time point can be predicted based on the device status information of the target digital twin at multiple time points through a pre-trained energy output prediction model.

[0068] In an embodiment of the present invention, brain-like computing different from traditional machine learning algorithms is applied to achieve distributed energy output prediction, that is, distributed energy output prediction is performed using a pulse neural network. Since the brain-like algorithm based on the pulse neural network inspired by the spatiotemporal dynamic characteristics of the human brain has broken through the problems of poor stability, high power consumption and low efficiency of the deep neural network model, the distributed energy output prediction based on the pulse neural network not only improves the accuracy and reliability of the prediction, but also reduces the power consumption of the system. In addition, the energy output prediction model used in the present invention is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, and can determine the influence of the membrane potential of the previous n time steps on the membrane potential calculation of the current time step, which can improve the overall memory capacity of the energy output prediction model, thereby reducing the problems of gradient disappearance and explosion that may occur during model training, thereby improving the performance of the trained energy output prediction model and improving the prediction accuracy of the model.

[0069] According to one embodiment of the present invention, the cloud platform 102 further includes a data mapping library, a logical entity space, a data warehouse module, and a resource slicing module;

[0070] a data mapping library configured to configure data mapping rules for mapping physical entities to virtual space;

[0071] The logical entity space is configured to map the device status information of each distributed energy device at multiple time points to the virtual space according to the data mapping rule to obtain the digital twin of each distributed energy device;

[0072] A data warehouse module is configured to receive and store device status information of each distributed energy device at multiple time points;

[0073] The resource slicing module is configured to manage the device status information of each distributed energy device at multiple time points in slices according to the different distributed energy devices, and update the device status information of each distributed energy device in real time.

[0074] In the specific implementation, the data mapping library mainly contains data mapping rules for mapping physical entities to logical entities. The logical entity space mainly implements the logical entity modeling of physical entities on the cloud platform. Through the data mapping rules, the device status information of each distributed energy device at multiple time points is mapped to the virtual space, and the digital twin of each distributed energy device is established to achieve fine-grained mapping of all information in the physical entity to the virtual space. The data warehouse module mainly implements real-time updates of the received device status information of each distributed energy device, and stores historical device status information to provide data support for the resource slicing module and the energy output prediction module. The resource slicing module mainly implements the classification management and detection of different distributed resources. It has an embedded intelligent algorithm, which can use the virtual-real interaction technology to integrate the device status information of distributed energy devices on the basis of the collection model to achieve real-time monitoring of the device status information.

[0075] As an example, the resource slicing module can store the device status information of distributed energy devices of the same distributed energy together, and store the device status information of distributed energy devices of different distributed energy separately, thereby realizing the unified management of the same type of distributed energy information. Alternatively, the resource slicing module can store the device status information of the same distributed energy device together, and store the device status information of different distributed energy devices separately, thereby realizing the unified management of the device status information of the same distributed energy device. This improves management efficiency.

[0076] According to an embodiment of the present invention, the device status information may include actual output and output impact information. The cloud platform 102 includes an energy output prediction module, and the energy output prediction module is configured as follows:

[0077] Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information;

[0078] Determine a first actual output corresponding to the first output impact information;

[0079] The first actual output, the first output impact information and the current output impact information are input into a pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

[0080] In the specific implementation, the energy output prediction module has two main functions. One is to use the equipment status information (historical equipment status information) stored in the data warehouse module to train and optimize the energy output prediction model within a certain period of time to improve the accuracy of the energy output prediction model. The other is to perform distributed energy output prediction based on the trained energy output prediction model.

[0081] In some embodiments, the output impact information is information in the device status information that affects the output of the distributed energy device, and can also be understood as information in the device status information other than the actual output. Exemplarily, the output impact information may include at least one of meteorological data and device configuration data, and the meteorological data may include temperature, humidity, wind direction, wind speed, etc.

[0082] As an example, the historical device status information of the target digital twin at multiple time points can be searched from the data warehouse module, and multiple historical output impact information can be obtained from the multiple historical device status information. Since each historical output impact information may include multiple data, the historical output impact information can be compared with the current output impact information. If a preset number or more of the multiple data included in any historical output impact information is the same as the data included in the current output impact information, the historical output impact information is determined to be the first output impact information. Among them, the preset number can be set according to actual needs, or it can be set by default by the device, and this embodiment does not limit this.

[0083] Exemplarily, if each data included in a certain historical output impact information is the same as the data included in the current output impact information; or, if a certain type of data included in a certain historical output impact information is the same as the data of that type included in the current output impact information; or, if a preset number of data among the multiple data included in a certain historical output impact information are the same as the data included in the current output impact information, it is determined that the historical data impact information is the first output impact information.

[0084] For example, assuming that the historical output impact information includes meteorological data, and the meteorological data includes temperature, humidity, irradiance, and wind speed. If the humidity, irradiance, and wind speed included in the historical output impact information are the same as the humidity, irradiance, and wind speed included in the current output impact information, but the temperature differs by 1 degree Celsius, it can be determined that the historical output impact information is the first output impact information.

[0085] Furthermore, each historical output impact information corresponds to an actual output, and after determining the first output impact information, the corresponding first actual output can be determined. The first actual output, the first output impact information, and the current output impact information are input into the trained energy output prediction model, and the energy output prediction model can predict the output of the target twin at the current time point.

[0086] According to one embodiment of the present invention, the cloud platform 102 also includes a decision module, which is configured to: obtain output demand information of the target distributed energy; generate a maintenance strategy based on the output demand information and output conditions; and send a maintenance instruction carrying the maintenance strategy to the terminal device 101.

[0087] Accordingly, the terminal device 101 is further configured as follows:

[0088] Receive maintenance instructions and perform maintenance on multiple distributed energy devices based on maintenance strategies;

[0089] The device status information of each distributed energy device after maintenance is obtained and sent to the cloud platform 102, so that the cloud platform 102 updates the device status information of each distributed energy device.

[0090] In the specific implementation, in order to ensure the safety of distributed resource grid connection and the stable operation of distributed resource output, the decision module will generate a maintenance strategy for distributed energy equipment based on the output demand information of the target distributed energy and the predicted output, and carry the maintenance strategy in the maintenance instruction and send it to the end-side device 101. The end-side device 101 will maintain multiple distributed energy devices to ensure the stability of distributed resource output.

[0091] As an example, the output demand information may be the total amount of distributed energy generation required by a certain area. The maintenance strategy may be a grid connection plan, that is, to integrate the distributed energy equipment into other power grids in the same area, or into the power grids of other areas. For example, area 1 needs the distributed energy generation to reach A, but according to the predicted output of all distributed energy equipment in area 1, it is determined that the demand cannot be met, then it is necessary to call distributed energy equipment from other areas and integrate it into area 1.

[0092] In the specific implementation, after the end-side device receives the maintenance instruction, it can realize the safe grid connection of the distributed energy equipment according to the maintenance strategy, and after the grid connection, obtain the updated device status information of each distributed energy device and send it to the cloud platform 102, thereby completing the iterative update of the device status information of each distributed energy device to ensure the reliability of the predictive maintenance function.

[0093] It should be noted that the above is an explanation of the distributed energy output prediction using the cloud-end structure as an example. In another possible implementation, the system may also include a side device 103, which is configured to: receive the device status information of each distributed energy device at multiple time points sent by the end-side device 101, store and forward it to the cloud platform 102.

[0094] In a specific implementation, since the distributed energy output prediction structure system established by the present invention needs to realize information transmission and information interaction between physical entities and logical entities, and is applied in the entire process of distributed resource output prediction. Digital twin technology maps physical entities in physical space to virtual space to obtain logical entities, and a virtual-real interaction module can be configured, which can be configured on the side device 103.

[0095] In some embodiments, the edge device 103 includes an acquisition layer and a transmission layer. The function of the acquisition layer is to aggregate the device status information reported by the end device 101, and the transmission layer transmits the device status information to the cloud platform 102 as the basis for the subsequent establishment of a virtual space for data calls. In addition, data transmission can be carried out using serial communication, HPLC communication (low-voltage power line high-speed carrier communication), WIFI communication, micro-power communication, 6G / 5G / 4G and other communication protocols. The transmission layer realizes efficient transmission of data, and stores these data on both the cloud platform 102 and the edge device 103, so that dynamic response and efficient mutual calls can be achieved according to demand.

[0096] As an example, the edge device 103 includes converged terminals, smart terminals and other devices, and the edge device 103 is responsible for aggregating, collecting and storing the device status reported by the terminal device 101 connected to it and the data information collected by itself, and also has edge computing capabilities.

[0097] In this case, the edge device 103 and the terminal device 101 have a binding relationship, and one edge device 103 can be bound to multiple terminal devices 101, that is, one edge device 103 is responsible for the terminal device 101 of one substation. Therefore, the cloud platform 102 receives the device status information of multiple distributed energy devices in the substation that it is responsible for, which is reported from different edge devices 103. In order to generate a suitable maintenance strategy after predicting the output of distributed energy and connect the distributed energy devices to the grid, the binding relationship between the distributed energy devices and the edge devices 103 can be stored in the data mapping library of the cloud platform 102. When performing data mapping, the distributed energy devices are first registered on the cloud platform 102, that is, each distributed energy device is authenticated on the cloud platform 102, and then the physical entity is mapped to the logical entity.

[0098] See also Figure 2 , Figure 2 It is a structural diagram of another distributed energy output prediction system provided according to an embodiment of the present invention. The system includes a terminal device 101, a cloud platform 102 and an edge device 103.

[0099] The end-side device 101 provides a physical space. The end-side device 101 includes photovoltaic equipment, wind power equipment and energy storage equipment. For photovoltaic equipment, the sensor collects data information including irradiance, temperature, wind speed, humidity, etc. For wind power equipment, the sensor collects data information including operating temperature, altitude, wind speed, survival temperature, wind direction, humidity, atmospheric pressure, etc. For energy storage equipment, the sensor collects data information including battery capacity, operating temperature, etc. The distributed energy data information reported by the photovoltaic equipment includes photovoltaic array, operation data, azimuth, etc., and the distributed energy data information reported by the wind power equipment includes wind rotor diameter, rated power, etc.

[0100] The edge device 103 includes an acquisition layer and a transmission layer, and the acquisition layer is used to receive sensor acquisition data information and distributed energy data information, and the transmission methods of the transmission layer include serial communication, HPLC communication, WIFI communication, micro-power communication and 6G / 5G / 4G.

[0101] The cloud platform 102 provides a virtual space, which includes a processing layer and a distributed energy output prediction layer, and can provide system modeling and function warehouses. The processing layer includes a resource slicing module, a data warehouse module, and an energy output prediction module, and the resource slicing module monitors and updates the device status information in real time, and sends the update results to the data warehouse module. The energy output prediction module is used for data processing and output prediction, and it obtains the device status information from the data warehouse module. In addition, after the energy output prediction module performs output prediction, the output obtained by the prediction will correct the device status information of the data warehouse module and the resource slicing module. The processing layer provides data support and algorithm support for the distributed energy output prediction layer. The distributed energy output prediction layer includes a logical entity space, a data mapping library, and a decision module.

[0102] In addition, there is data update and data mapping between the end-side device 101 and the cloud platform 102. The cloud platform 102 generates a maintenance strategy and sends it to the end-side device 101, which makes decisions and deploys it. The edge device 103 and the cloud platform 102 also interact through the sensing access technology.

[0103] It should be noted that the energy output prediction model used in the present invention is obtained by pre-training a pulse neural network including a gating unit, and the pulse neural network can use LIF (Leaky Integrate and Fire) neurons or LIAF (Leaky Integrate and Analog Fire) neurons, which is not limited in the embodiments of the present invention. The structure and training process of the energy output prediction model are introduced below.

[0104] See also Figure 3 , Figure 3 is a schematic diagram of a structure of an energy output prediction model provided according to an embodiment of the present invention. The energy output prediction model includes a gating unit (i.e. Figure 3 The dotted box marked part) and RAM (Random Access Memory, random access memory), and the energy output prediction model includes integration, triggering and resetting functions.

[0105] The membrane potential changes and pulse actions of a single neuron are expressed as:

[0106]

[0107] In formula 1, τ m represents the membrane potential time constant, V represents the membrane potential of the neuron, V th Represents the preset threshold potential, V reset represents the resting potential, V t-1 W represents the membrane potential of the neuron after all cell activities in the previous time step have ended. j represents the weight of the jth forward synapse, I j (t) represents the input of the jth forward synapse at the current time step.

[0108] First, the neuron at time t (current time step) receives the pulse stream transmitted from the forward synapse at the current time step and updates its own membrane potential state. The traditional model uses pulses to calculate the membrane potential state at the current time step, and the current time step does not directly output pulse information, but outputs the membrane potential after activation.

[0109] As described in Formula 2, depending on whether the threshold potential is used, the activation mode can be divided into two modes: TR (Threshold Related) and NTR (No Threshold Related). th Synonymous with Formula 1, Act() can be any nonlinear activation function, such as sigmoid, relu.

[0110]

[0111] (1) Integrate the input of the current time step

[0112]

[0113] In formula 3, u t represents the temporary membrane potential of the current time step after calculation, I t Represents the input of the forward synapse of the cell at the current time step, and W represents the weight matrix of the forward synapse. According to different requirements, it can be divided into two forms: fully connected (FC) and convolution (conv).

[0114] (2) Accumulate the membrane potential of one time step

[0115]

[0116] In formula 4, represents the membrane potential after the integration of the temporary membrane potential and the previous time step, Represents the membrane potential of the neuron at the previous time step.

[0117] (3) Compare with the threshold potential and decide whether to send a pulse

[0118]

[0119] In Formula 5, F t Is a Boolean value, true (i.e. F t The value is 1) indicates that the cell has a pulse at time t (the current time step), otherwise (i.e. F t The value is 0) means no pulse is emitted, V th It is a preset threshold potential. In engineering implementation, it can be manually set or obtained by autonomous learning of the model.

[0120] (4) Get the final membrane potential of the current time step

[0121]

[0122] In Formula 6, represents the final membrane potential of the current time step, ⊙ represents the matrix dot product, τ α With τ β are leakage constants, which are consistent with τ in Formula 1. m This step is processed separately according to whether a pulse is issued. If a pulse is issued, the membrane potential is reset to the resting potential, otherwise it remains unchanged. Then the membrane potential decay is calculated based on the leakage factor to obtain the final membrane potential of the current time step.

[0123] (5) Calculate the final output of the current time step.

[0124]

[0125] In Formula 7, O t It represents the final output at time t (current time step) and is calculated in two forms. If the energy output prediction model is LIF, it outputs whether a pulse is issued at time t; if the energy output prediction model is LIAF, it outputs the activation value calculated by formula 2.

[0126] In this scheme, an approximate gradient function is set to serve as Formula 5. First, Formula 5 is slightly transformed to obtain the Dirac delta function in the form of Formula 8:

[0127]

[0128] Afterwards, in the back propagation phase, the derivative function of Formula 8 is calculated according to Formula 9:

[0129]

[0130] In Formula 9, μ is usually a small positive number, and δ is usually an integer greater than 1. Generally, μ can be set to 0.5 and δ can be set to 1.

[0131] The present invention embeds a gating unit into the model described by Formula 2-Formula 9 to solve the problems of gradient explosion and gradient disappearance, so that the energy output prediction model has long-term memory capability while maintaining the low power consumption characteristic of the pulse neural network.

[0132] Figure 3 The gate control unit shown in the dashed box can be described by formula 10:

[0133]

[0134]

[0135] In Formula 10, α represents the model parameter, σ represents the activation function, and I t represents the input of the current time step, represents the membrane potential at time step ti, W m and W in Both represent the network weights of the energy output prediction model, FC represents fully connected processing, and conv represents convolution processing.

[0136] The gate control unit receives I t and As input, after calculation by formula 10, α is returned. α is a floating point number with a value range between [0,1]. α will replace τ in formula 6 α , and τ in Formula 6 β Replaced by Participate in subsequent calculations of the model.

[0137] As an example, the energy output prediction model can be pre-trained based on multiple sets of historical equipment status information, and multiple historical output impact information can be input into the energy output prediction model to obtain multiple predicted output conditions. The predicted output conditions are compared with the actual output conditions, and the model parameters are adjusted according to the difference to obtain a trained energy output prediction model.

[0138] It can be seen that the embodiment of the present invention proposes an improved pulse neural network, whose gating unit uses the input information of the current time step and the membrane potential of the previous n time steps to determine the leakage factor, and then determines the influence of the membrane potential of the previous n time steps on the calculation of the membrane potential of the current time step. This can improve the overall memory capacity of the model, thereby solving the problems of gradient vanishing and explosion that may occur during model training, so that the model can learn useful features, improve the performance of the model, and further improve the prediction accuracy of the model.

[0139] The distributed energy output prediction method provided by the embodiment of the present invention uses digital twin technology to realize the logical mapping of distributed energy from physical space to virtual space, realizes the observable, measurable and controllable status of different types of distributed energy, simplifies process efficiency, improves operation accuracy, and proposes the concept of resource slice management, realizing the unified management of the device status information of the same type of distributed energy. In addition, the present invention uses a pulse neural network including a gating unit to predict the output of distributed energy, which improves the accuracy and reliability of the prediction while reducing the power consumption of the system. In addition, the gating unit uses the input information of the current time step and the membrane potential of the previous n time steps to determine the leakage factor, and then determines the influence of the membrane potential of the previous n time steps on the membrane potential of the current time step, which can improve the overall memory capacity of the energy output prediction model, thereby alleviating the gradient disappearance and explosion problems that may occur during the training of the energy output prediction model, thereby improving the performance of the energy output prediction model, so that the accuracy of the distributed energy output prediction using the energy output prediction model is higher.

[0140] Figure 4 : is a flow chart of a distributed energy output prediction method provided according to an embodiment of the present invention, the method is applied to a cloud platform, and comprises the following steps:

[0141] Step 401: Receive device status information of each distributed energy device at multiple time points.

[0142] According to some embodiments of the present invention, after receiving the device status information of each distributed energy device at multiple time points, the method further includes:

[0143] According to the different distributed energy devices, the device status information of each distributed energy device at multiple time points is managed in pieces, and the device status information of each distributed energy device is updated in real time.

[0144] In a specific implementation, the data warehouse module of the cloud platform receives and stores the device status information of the distributed energy equipment at multiple time points. The device status information may be sent by the end-side device or by the edge-side device through the transport layer.

[0145] Step 402: Map the device status information of each distributed energy device at multiple time points to the virtual space to obtain a digital twin of each distributed energy device.

[0146] Step 403: Based on the device status information of the target digital twin at multiple time points, predict the output of the target digital twin at the current time point through a pre-trained energy output prediction model, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters based on the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

[0147] According to some embodiments of the present invention, the device status information includes actual output and output impact information. According to the device status information of the target digital twin at multiple time points, the output of the target digital twin at the current time point is predicted by a pre-trained energy output prediction model, including:

[0148] Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information;

[0149] Determine a first actual output corresponding to the first output impact information;

[0150] The first actual output, the first output impact information and the current output impact information are input into a pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

[0151] According to some embodiments of the present invention, during the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, including:

[0152]

[0153] Among them, α represents the model parameter, σ represents the activation function, and It represents the input of the current time step. represents the membrane potential at time step ti, W m and W in Both represent the network weights of the energy output prediction model, FC represents fully connected processing, and conv represents convolution processing.

[0154] According to some embodiments of the present invention, after predicting the output of the target digital twin at the current time point by using a pre-trained energy output prediction model according to the device status information of the target digital twin at multiple time points, the method further includes:

[0155] Obtain output demand information of target distributed energy;

[0156] Generate maintenance strategies based on output demand information and output conditions;

[0157] A maintenance instruction carrying a maintenance strategy is sent to a terminal device to perform maintenance on multiple distributed energy devices.

[0158] It should be noted that the relevant descriptions of steps 401 to 403 in this embodiment can refer to the relevant descriptions of the above-mentioned system embodiment, and this embodiment will not be repeated here.

[0159] The distributed energy output prediction method provided by the embodiment of the present invention uses digital twin technology to realize the logical mapping of distributed energy from physical space to virtual space, realizes the observable, measurable and controllable status of different types of distributed energy, simplifies process efficiency, improves operation accuracy, and proposes the concept of resource slice management, realizing the unified management of the device status information of the same type of distributed energy. In addition, the present invention uses a pulse neural network including a gating unit to predict the output of distributed energy, which improves the accuracy and reliability of the prediction while reducing the power consumption of the system. In addition, the gating unit uses the input information of the current time step and the membrane potential of the previous n time steps to determine the leakage factor, and then determines the influence of the membrane potential of the previous n time steps on the membrane potential of the current time step, which can improve the overall memory capacity of the energy output prediction model, thereby alleviating the gradient disappearance and explosion problems that may occur during the training of the energy output prediction model, thereby improving the performance of the energy output prediction model, so that the accuracy of the distributed energy output prediction using the energy output prediction model is higher.

[0160] Figure 5 : is a schematic diagram of a distributed energy output prediction device provided according to an embodiment of the present invention, the device may include:

[0161] A receiving unit 501 is used to receive device status information of each distributed energy device at multiple time points;

[0162] A mapping unit 502 is used to map the device status information of each distributed energy device at multiple time points to a virtual space to obtain a digital twin of each distributed energy device;

[0163] The prediction unit 503 is used to predict the output of the target digital twin at the current time point according to the device status information of the target digital twin at the multiple time points through a pre-trained energy output prediction model, wherein the target digital twin is a digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, and n is a positive integer greater than 0.

[0164] According to an embodiment of the present invention, the device status information includes actual output and output impact information, and the prediction unit 503 is further used to:

[0165] Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information;

[0166] Determine a first actual output corresponding to the first output impact information;

[0167] The first actual output, the first output impact information and the current output impact information are input into a pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

[0168] According to one embodiment of the present invention, the device further includes a slice management unit, which is used to:

[0169] According to the different distributed energy devices, the device status information of each distributed energy device at multiple time points is managed in pieces, and the device status information of each distributed energy device is updated in real time.

[0170] According to one embodiment of the present invention, the device further comprises a maintenance unit, which is used to:

[0171] Obtain output demand information of target distributed energy;

[0172] Generate maintenance strategies based on output demand information and output conditions;

[0173] A maintenance instruction carrying a maintenance strategy is sent to a terminal device to perform maintenance on multiple distributed energy devices.

[0174] According to one embodiment of the present invention, the apparatus further comprises a training unit, wherein the training unit is configured to:

[0175]

[0176] Among them, α represents the model parameter, σ represents the activation function, and I t represents the input of the current time step, represents the membrane potential at time step ti, W m and W in Both represent the network weights of the energy output prediction model, FC represents fully connected processing, and conv represents convolution processing.

[0177] The distributed energy output prediction method provided by the embodiment of the present invention uses digital twin technology to realize the logical mapping of distributed energy from physical space to virtual space, realizes the observable, measurable and controllable status of different types of distributed energy, simplifies process efficiency, improves operation accuracy, and proposes the concept of resource slice management, realizing the unified management of the device status information of the same type of distributed energy. In addition, the present invention uses a pulse neural network including a gating unit to predict the output of distributed energy, which improves the accuracy and reliability of the prediction while reducing the power consumption of the system. In addition, the gating unit uses the input information of the current time step and the membrane potential of the previous n time steps to determine the leakage factor, and then determines the influence of the membrane potential of the previous n time steps on the membrane potential of the current time step, which can improve the overall memory capacity of the energy output prediction model, thereby alleviating the gradient disappearance and explosion problems that may occur during the training of the energy output prediction model, thereby improving the performance of the energy output prediction model, so that the accuracy of the distributed energy output prediction using the energy output prediction model is higher.

[0178] The above is a schematic scheme of a distributed energy output prediction device according to an embodiment of the present invention. It should be noted that the technical scheme of the distributed energy output prediction device and the technical scheme of the distributed energy output prediction method described above belong to the same concept, and the details not described in detail in the technical scheme of the distributed energy output prediction device can be found in the description of the technical scheme of the distributed energy output prediction method described above.

[0179] Figure 6 6 is a schematic diagram of a computing device according to an embodiment of the present invention. The computing device 600 includes: a memory 601, a processor 602, and a computer program stored in the memory 601 and executable on the processor 602. When the processor 602 executes the computer program, a distributed energy output prediction method provided in any of the above embodiments is implemented.

[0180] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a distributed energy output prediction method as provided in any of the above embodiments.

[0181] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions configured to implement logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wirings (electronic devices), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM, Compact Disk Read Only Memory). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0182] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit configured to implement a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA, Programmable Gate Array), a field programmable gate array (FPGA, Field Programmable Gate Array), etc.

[0183] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" 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 may be combined in any one or more embodiments or examples in a suitable manner.

[0184] In addition, the terms "first" and "second" are only configured for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0185] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0186] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A distributed energy output prediction method, characterized in that: Applied to a cloud platform, the method includes: Receive device status information of each distributed energy device at multiple time points; Mapping the device status information of each distributed energy device at multiple time points to the virtual space to obtain the digital twin of each distributed energy device; According to the device status information of the target digital twin at the multiple time points, the output of the target digital twin at the current time point is predicted by a pre-trained energy output prediction model, wherein the target digital twin is a digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

2. The distributed energy output prediction method according to claim 1, characterized in that: The device status information includes actual output and output impact information. The output of the target digital twin at the current time point is predicted by a pre-trained energy output prediction model based on the device status information of the target digital twin at the multiple time points, including: Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information; Determining a first actual output corresponding to the first output impact information; The first actual output, the first output impact information and the current output impact information are input into the pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

3. The distributed energy output prediction method according to claim 1 or 2, characterized in that: After receiving the device status information of each distributed energy device at multiple time points, the method further includes: According to the different distributed energy devices, the device status information of each distributed energy device at multiple time points is managed in pieces, and the device status information of each distributed energy device is updated in real time.

4. The distributed energy output prediction method according to claim 1 or 2, characterized in that: After predicting the output of the target digital twin at the current time point by using a pre-trained energy output prediction model according to the device status information of the target digital twin at the multiple time points, the method further includes: Obtain output demand information of target distributed energy; generating a maintenance strategy according to the output demand information and the output situation; A maintenance instruction carrying a maintenance strategy is sent to a terminal device to perform maintenance on multiple distributed energy devices.

5. The distributed energy output prediction method according to claim 1 or 2, characterized in that: During the training process, the gate control unit determines and updates the model parameters according to the input information of the current time step and the membrane potential of the previous n time steps, including: Among them, α represents the model parameter, σ represents the activation function, and I t represents the input of the current time step, represents the membrane potential at time step ti, W m and W in Both represent the network weights of the energy output prediction model, FC represents fully connected processing, and conv represents convolution processing.

6. A distributed energy output prediction system, characterized in that: The system includes a terminal device and a cloud platform; The terminal device is configured to send device status information of each distributed energy device at multiple time points to the cloud platform; The cloud platform is configured to map the device status information of each distributed energy device at multiple time points to a virtual space to obtain a digital twin of each distributed energy device, and predict the output of the target digital twin at the current time point through a pre-trained energy output prediction model based on the device status information of the target digital twin at the multiple time points, wherein the target digital twin is the digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters based on the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

7. The distributed energy output prediction system according to claim 6, characterized in that: The device status information includes actual output and output impact information, and the cloud platform includes an energy output prediction module, and the energy output prediction module is configured as follows: Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information; Determining a first actual output corresponding to the first output impact information; The first actual output, the first output impact information and the current output impact information are input into the pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

8. The distributed energy output prediction system according to claim 6 or 7, characterized in that: The cloud platform also includes a data mapping library, a logical entity space, a data warehouse module and a resource slicing module; The data mapping library is configured to configure data mapping rules for mapping physical entities to virtual spaces; The logical entity space is configured to map the device status information of each distributed energy device at multiple time points to the virtual space according to the data mapping rule to obtain a digital twin of each distributed energy device; The data warehouse module is configured to receive and store the device status information of each distributed energy device at multiple time points; The resource slicing module is configured to manage the device status information of each distributed energy device at multiple time points in slices according to the different distributed energy devices, and update the device status information of each distributed energy device in real time.

9. The distributed energy output prediction system according to claim 6 or 7, characterized in that: The cloud platform further includes a decision module, which is configured to: Obtain output demand information of target distributed energy; generating a maintenance strategy according to the output demand information and the output situation; A maintenance instruction carrying a maintenance policy is sent to the terminal side device.

10. The distributed energy output prediction system according to claim 9, characterized in that: The terminal side device is further configured as follows: receiving the maintenance instruction, and performing maintenance on a plurality of distributed energy devices based on the maintenance strategy; The device status information of each distributed energy device after maintenance is obtained and sent to the cloud platform, so that the cloud platform updates the device status information of each distributed energy device.

11. The distributed energy output prediction system according to claim 6 or 7, characterized in that: The system further includes a side device, wherein the side device is configured to: Receive the device status information of each distributed energy device at multiple time points sent by the terminal side device, store and forward it to the cloud platform.

12. A distributed energy output prediction device, characterized in that: The device comprises: A receiving unit, used to receive device status information of each distributed energy device at multiple time points; A mapping unit, used to map the device status information of each distributed energy device at multiple time points to a virtual space to obtain a digital twin of each distributed energy device; A prediction unit is used to predict the output of the target digital twin at the current time point through a pre-trained energy output prediction model based on the device status information of the target digital twin at the multiple time points, wherein the target digital twin is a digital twin of any distributed energy device, and the energy output prediction model is a pulse neural network including a gating unit. During the training process, the gating unit determines and updates the model parameters based on the input information of the current time step and the membrane potential of the previous n time steps, where n is a positive integer greater than 0.

13. The distributed energy output prediction device according to claim 12, characterized in that: The device status information includes actual output and output impact information, and the prediction unit is further used to: Determine, from the output impact information of the target digital twin at multiple time points, first output impact information that is the same as or similar to the current output impact information; Determining a first actual output corresponding to the first output impact information; The first actual output, the first output impact information and the current output impact information are input into the pre-trained energy output prediction model to predict the output of the target digital twin at the current time point.

14. The distributed energy output prediction device according to claim 12 or 13, characterized in that: The device also includes a slice management unit, which is used to: According to the different distributed energy devices, the device status information of each distributed energy device at multiple time points is managed in pieces, and the device status information of each distributed energy device is updated in real time.

15. The distributed energy output prediction device according to claim 12 or 13, characterized in that: The device also includes a maintenance unit, which is used to: Obtain output demand information of target distributed energy; generating a maintenance strategy according to the output demand information and the output situation; A maintenance instruction carrying a maintenance strategy is sent to a terminal device to perform maintenance on multiple distributed energy devices.

16. A computing device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the distributed energy output prediction method as described in any one of claims 1 to 5 is implemented.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the distributed energy output prediction method as described in any one of claims 1 to 5 is implemented.

Images