Search

KR-102963376-B1 - SCHEDULING METHOD FOR RENEWABLE ENERGY-BASED MULTI-PURPOSE CONVERSION SYSTEM AND ELECTRONIC APPARATUS THEREFOR

KR102963376B1KR 102963376 B1KR102963376 B1KR 102963376B1KR-102963376-B1

Abstract

An electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure comprises: a memory in which at least one process for performing a process scheduling operation according to a time-series weather change is stored; and at least one processor that performs the operation according to the process; wherein the at least one processor is configured to perform process scheduling by generating process-related control parameters for energy and product flows between a plurality of processes according to a time-series weather change through a deep learning model learned based on time-series renewable energy production data and system operation data according to the same, and outputting the process-related control parameters in the form of a time-series sequence. The deep learning model may include: an input layer comprising an encoder that processes the input time-series renewable energy production data and system operation data according to the same; a common hidden layer connected to the input layer and a branched hidden layer connected to each output layer that outputs each process-related control parameter; and an output layer that outputs the time-series process-related control parameters according to the distribution ratio of energy and material between the processes.

Inventors

  • 이남경
  • 박종민

Assignees

  • 한국전력기술 주식회사

Dates

Publication Date
20260508
Application Date
20250813

Claims (10)

  1. In an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system, A memory storing at least one process for performing process scheduling operations based on time-series weather changes; and It includes at least one processor that performs the above operation according to the above process, and The above at least one processor is, Through a deep learning model trained on time-series renewable energy production data and the resulting system operation data, process-related control parameters for energy and product flows between multiple processes according to time-series weather changes are generated, and It is configured to perform process scheduling by outputting the above process-related control parameters in the form of a time-series sequence, and The above deep learning model is, An input layer comprising an encoder that processes the input time-series renewable energy production data and the system operation data based thereon; A common hidden layer connected to the input layer and a branched hidden layer connected to each output layer that outputs control parameters related to each process; and An output layer that outputs time-series process-related control parameters according to the distribution ratio of energy and material between the above processes; comprising Electronic device.
  2. In Article 1, The above process-related control parameter is characterized as being a control parameter for at least one of a renewable energy production process, an energy conversion process, and an energy or product storage process. Electronic device.
  3. In Article 1, The above deep learning model is a reinforcement learning-based actor-critic structure, and An actor network that generates distribution ratio control actions based on a probability distribution for the energy or product distribution ratio between the above processes; and It includes a critique network that outputs a value function of scalar values for the policy, and The above actor network and the above critique network share the above input layer and the above common hidden layer, Electronic device.
  4. In Paragraph 3, The above-mentioned at least one processor is, The above deep learning model is a structure that maximizes the total production volume of the final product by setting the production deviation by time series of the final product resulting from the process execution of the above multi-purpose transformation system as a compensation. A structure that minimizes the cost of the energy or the product by setting the time-series cost deviation according to the above process as compensation, and Designed to be reinforced learning such that an optimization goal is reflected in the process scheduling, based on at least one structure that maximizes the sales of the energy or the product by setting the sales deviation by time series according to the above process as a compensation. Electronic device.
  5. In Article 1, The branching hidden layer of the above deep learning model is, A structure comprising a time series decoder that sequentially generates branching tuples, wherein a normalization or loss weight adjustment mechanism is applied to prevent negative transfer between parameters at branching, Electronic device.
  6. In Article 1, The above-mentioned at least one processor is, Configured to input the operating status of the unit facility, the load of the unit facility, energy flow between processes, product flow between processes, and information on the cost, price, and demand and supply of the energy or product, along with the time-series renewable energy production data, into the deep learning model. Electronic device.
  7. In Article 1, The above-mentioned at least one processor is, Generate a training renewable energy dataset using a generative AI model trained on measured renewable energy data, and Generate an augmented dataset by performing goodness-of-fit filtering on the generated training renewable energy dataset, and Configured to train the deep learning model based on the augmented dataset and the system operation data, Electronic device.
  8. In Article 7, The above-mentioned at least one processor is, Configured to perform the suitability filtering by considering renewable energy production patterns according to weather patterns in each region where the above system is located, Electronic device.
  9. In Article 1, The above-mentioned at least one processor is, We perform mathematical modeling that takes multiple weather scenario data for renewable energy as input and calculates the hourly load, SoC, power flow, and converted product production for multiple facilities as outputs, and Configured to use the output data produced based on the above mathematical modeling results as training data for the above deep learning model. Electronic device.
  10. A process scheduling method for a renewable energy-based multi-purpose conversion system performed by a processor of a device, wherein the method comprises: A step of performing mathematical modeling to calculate the hourly load, SoC, power flow, and converted product production volume for multiple facilities as outputs, using multiple weather scenario data for renewable energy as input; A step of training a deep learning model using process-related control parameters within a preset period, generated based on the above mathematical modeling results, as training data; A step of generating process-related control parameters for energy and product flows between multiple processes according to time-series weather changes through the above-mentioned deep learning model; and The method includes the step of performing process scheduling by outputting the above process-related control parameters in the form of a time-series sequence; The above deep learning model is, An input layer including an encoder that processes input time-series renewable energy production data and system operation data based thereon; A common hidden layer connected to the input layer and a branched hidden layer connected to each output layer that outputs control parameters related to each process; and The output layer outputting time-series process-related control parameters according to the distribution ratio of energy and material between the processes; Configured to output the above process-related control parameters in the form of a time-series sequence, Process scheduling method for a renewable energy-based multi-purpose conversion system.

Description

Scheduling method for renewable energy-based multi-purpose conversion system and electronic apparatus therefor The present disclosure relates to process scheduling of a Power-to-X (P2X) system, and more specifically, to a process scheduling method for a renewable energy-based multi-purpose conversion system configured to optimize the flow of energy and products between processes despite the variability of renewable energy production due to weather changes, and to an electronic device for the same. Recently, as renewable energy has garnered attention as a means to achieve carbon neutrality, the utilization of renewable energy-based multi-purpose conversion systems that produce hydrogen and ammonia using wind and solar power is rapidly increasing. However, because these renewable energy sources exhibit significant variability in production volume due to climate change and possess intermittent characteristics, it is difficult to optimize real-time operation and achieve stable renewable energy production. In particular, renewable energy production is significantly affected by changes in solar energy over time or climate variations depending on regional location. Existing renewable energy production systems set process-related control parameters using fixed-scenario-based mathematical modeling based on statistical average weather conditions, or through rule-control logic based on engineers' experience in response to real-time climate fluctuations. Consequently, it was difficult to predict equipment load and production volume as they could not respond immediately to changes in external factors affecting renewable energy production. Furthermore, inefficient control of energy or material flow between facilities resulted in overloads within the facilities and bottlenecks between facilities. Furthermore, the inability to optimize process control has caused facility operating costs to increase significantly and securing efficient production volumes to become difficult when weather conditions differ from forecasts, leading to a decline in the economic viability of renewable energy. Therefore, there are limitations in securing production stability and economic viability of renewable energy production systems, and there is an urgent need to develop process control technologies that account for the real-time variability of renewable energy. FIG. 1 is a block diagram briefly illustrating the configuration of an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 2 is a simplified layout diagram illustrating the facility configuration of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 3 is a process diagram illustrating the process of collecting and augmenting a training dataset for a deep learning model that performs process scheduling inference of an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 4 is a process diagram illustrating a goodness-of-fit filtering process for an augmented dataset of an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 5 is a configuration diagram illustrating the structure of a deep learning model of an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 6 is a configuration diagram illustrating the actor-critic structure of a deep learning model of an electronic device for process scheduling of a renewable energy-based multi-purpose conversion system according to the present disclosure. FIG. 7 is a flowchart illustrating a process scheduling method for a renewable energy-based multi-purpose conversion system according to the present disclosure. Throughout this disclosure, the same reference numerals denote the same components. This disclosure does not describe all elements of the embodiments, and general content in the art to which this disclosure pertains or content that overlaps between embodiments is omitted. The terms ‘part, module, component, block’ as used in the specification may be implemented in software or hardware, and depending on the embodiments, a plurality of ‘parts, modules, components, blocks’ may be implemented as a single component, or a single ‘part, module, component, block’ may include a plurality of components. Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are directly connected but also cases where they are indirectly connected, and indirect connections include connections made via a wireless communication network. Furthermore, when it is stated that a part "includes" a certain component, this means that, unless specifically stated otherwise, it does not exclude other c