[go: up one dir, main page]

TWI832131B - Management system of energy demand - Google Patents

Management system of energy demand Download PDF

Info

Publication number
TWI832131B
TWI832131B TW110148506A TW110148506A TWI832131B TW I832131 B TWI832131 B TW I832131B TW 110148506 A TW110148506 A TW 110148506A TW 110148506 A TW110148506 A TW 110148506A TW I832131 B TWI832131 B TW I832131B
Authority
TW
Taiwan
Prior art keywords
processing device
power consumption
data
load
energy demand
Prior art date
Application number
TW110148506A
Other languages
Chinese (zh)
Other versions
TW202326578A (en
Inventor
曾智一
戴佑宗
李國正
鄧皓宜
Original Assignee
友達光電股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 友達光電股份有限公司 filed Critical 友達光電股份有限公司
Priority to TW110148506A priority Critical patent/TWI832131B/en
Priority to CN202210498304.XA priority patent/CN114676938A/en
Publication of TW202326578A publication Critical patent/TW202326578A/en
Application granted granted Critical
Publication of TWI832131B publication Critical patent/TWI832131B/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for AC mains or AC distribution networks
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Health & Medical Sciences (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Quality & Reliability (AREA)
  • Development Economics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Educational Administration (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Power Engineering (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Primary Health Care (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Remote Monitoring And Control Of Power-Distribution Networks (AREA)

Abstract

A management system of energy demand includes a power management device and a processing device. The power management device controls a load of an electrical grid. The processing device is connected to the power management device and predicts multiple first predict power consumption corresponding to multiple first periods through a neural network according to multiple feature data. The processing device further makes a comparison between the first predict power consumption and an optimal contract capacity. When a first one of the first predict power consumption corresponding to a first one of the first periods is greater than the optimal contract capacity, the processing device monitors the load in the first one of the first periods to generate a control signal to the power management device, and the power management device reduces the load in response to the control signal.

Description

能源需量管理系統 energy demand management system

本案有關一種管理系統，特別是一種能源需量管理系統與方法。 This case relates to a management system, especially an energy demand management system and method.

能源使用情形可能由於產能、氣候、設備運轉等等的變數而無法掌握，從而使得調控能源與訂定契約容量相當被動。以習知的訂定契約容量方式而言，例如：一年一次以人力調整契約容量的方式，無法即時反應用電情形，調控能源的效率降低，導致所訂定的契約容量過高，進而造成基本電費的提升。因此，要如何發展能夠克服上述問題之相關技術為本領域重要之課題。 Energy usage may be difficult to control due to variables such as production capacity, climate, equipment operation, etc., making regulating energy and establishing contract capacity quite passive. Taking the conventional method of setting contract capacity, for example, manually adjusting the contract capacity once a year, it cannot reflect the power consumption situation in real time, and the efficiency of energy control is reduced, resulting in the contract capacity being set to be too high, which in turn causes Increase in basic electricity bills. Therefore, how to develop related technologies that can overcome the above problems is an important issue in this field.

本案的一個實施例是一種能源需量管理系統，包含電力管理裝置以及處理裝置。電力管理裝置調控電網的負載。處理裝置耦接上述電力管理裝置，並根據多個特徵資料透過類神經網路預測對應多個第一時間段的多個第一預測用電量。處理裝置將第一預測用電量與最佳化契約容量相比較。當相應於第一時間段之第一者的第一預測用電量的第一者大於最佳契約容量時，處理裝置在第一時間段之 第一者時監控該負載以產生管制訊號至電力管理裝置，並且上述電力管理裝置響應於管制訊號調降負載。 One embodiment of this case is an energy demand management system, including a power management device and a processing device. Power management devices regulate the load on the grid. The processing device is coupled to the power management device, and predicts a plurality of first predicted power consumption corresponding to a plurality of first time periods through a neural network based on a plurality of characteristic data. The processing device compares the first predicted power consumption with the optimized contract capacity. When the first predicted power consumption corresponding to the first time period is greater than the optimal contract capacity, the processing device The first is to monitor the load to generate a control signal to the power management device, and the power management device reduces the load in response to the control signal.

100:能源需量管理系統 100:Energy Demand Management System

110:電力管理裝置 110:Power management device

112:電網 112:Power grid

120:處理裝置 120: Processing device

122:管制訊號 122:Control signal

124:輸入裝置 124:Input device

126:顯示裝置 126:Display device

128:使用者輸入 128:User input

300:能源需量的管理方法 300: Energy Demand Management Methods

310,320,330,340,350:步驟 310,320,330,340,350: steps

C1:實際用電量 C1: Actual electricity consumption

C2:預測用電量 C2: Predict electricity consumption

C3,C4:曲線 C3,C4: Curve

L1:契約容量 L1: Contract capacity

L2:最佳化契約容量 L2: Optimized contract capacity

T1:時間點 T1: time point

為讓本案之上述和其他目的、特徵、優點與實施例能更明顯易懂，結合附圖閱讀時可以最好地理解本案內容的各方面。 In order to make the above and other purposes, features, advantages and embodiments of this case more obvious and understandable, all aspects of the content of this case can best be understood when read in conjunction with the accompanying drawings.

第1圖是根據一實施例的能源需量管理系統的示意圖。 Figure 1 is a schematic diagram of an energy demand management system according to an embodiment.

第2A圖是根據一實施例的能源需量管理系統之用電量的示意圖。 Figure 2A is a schematic diagram of electricity consumption of an energy demand management system according to an embodiment.

第2B圖是根據一實施例的能源需量管理系統之預測用電量的示意圖。 Figure 2B is a schematic diagram of predicted power consumption of the energy demand management system according to an embodiment.

第3圖是根據一實施例所繪示的能源需量的管理方法的流程圖。 Figure 3 is a flow chart of an energy demand management method according to an embodiment.

於本文中，當一元件被稱為「連接」或「耦接」時，可指「電性連接」或「電性耦接」。「連接」或「耦接」亦可用以表示二或多個元件間相互搭配操作或互動。此外，雖然本文中使用「第一」、「第二」、…等用語描述不同元件，該用語僅是用以區別以相同技術用語描述的元件或操作。除非上下文清楚指明，否則該用語並非特別指稱或暗示次序或順位，亦非用以限定本案。 In this document, when an element is referred to as "connected" or "coupled," it may mean "electrically connected" or "electrically coupled." "Connection" or "coupling" can also be used to indicate the coordinated operation or interaction between two or more components. In addition, although terms such as "first", "second", ... are used to describe different elements herein, the terms are only used to distinguish elements or operations described with the same technical terms. Unless the context clearly indicates otherwise, this term does not specifically refer to or imply a sequence or sequence, nor is it used to qualify the case.

這裡使用的術語僅僅是為了描述特定實施例的目 的，而不是限制性的。如本文所使用的，除非內容清楚地指示，否則單數形式「一」、「一個」和「該」旨在包括複數形式，包括「至少一個」。「或」表示「及/或」。如本文所使用的，術語「及/或」包括一個或多個相關所列項目的任何和所有組合。還應當理解，當在本說明書中使用時，術語「包括」及/或「包含」指定所述特徵、區域、整體、步驟、操作、元件的存在及/或部件，但不排除一個或多個其它特徵、區域整體、步驟、操作、元件、部件及/或其組合的存在或添加。 The terminology used herein is for the purpose of describing particular embodiments only. , rather than restrictive. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms including "at least one" unless the content clearly dictates otherwise. "Or" means "and/or". As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items. It will also be understood that when used in this specification, the terms "comprises" and/or "includes" designate the presence of stated features, regions, integers, steps, operations, elements and/or components but do not exclude one or more The presence or addition of other features, regions, steps, operations, elements, parts and/or combinations thereof.

第1圖是根據一些實施例的能源需量管理系統100的示意圖。能源需量管理系統100包含電力管理裝置110以及處理裝置120。電力管理裝置110用以調控電網112的負載。處理裝置120耦接於電力管理裝置110，並用以根據多個特徵資料透過類神經網路預測對應多個時間段的多個預測用電量，並進一步用以將預測用電量與最佳化契約容量相比較。當相應於上述多個時間段中一者的多個預測用電量中一者大於最佳化契約容量時，處理裝置120更用以在上述多個時間段中一者時監控該負載以產生管制訊號122至電力管理裝置110，並且電力管理裝置110更用以響應於管制訊號122調降負載。 Figure 1 is a schematic diagram of an energy demand management system 100 in accordance with some embodiments. The energy demand management system 100 includes a power management device 110 and a processing device 120 . The power management device 110 is used to regulate the load of the power grid 112 . The processing device 120 is coupled to the power management device 110 and is used to predict a plurality of predicted power consumption corresponding to multiple time periods through a neural network based on a plurality of characteristic data, and is further used to combine the predicted power consumption with optimization. Contract capacity compared. When one of the plurality of predicted power consumption corresponding to one of the plurality of time periods is greater than the optimal contract capacity, the processing device 120 is further configured to monitor the load in one of the plurality of time periods to generate The control signal 122 is sent to the power management device 110 , and the power management device 110 is further configured to reduce the load in response to the control signal 122 .

舉例而言，在一些實施例中，處理裝置120對相應二十四小時的預測用電量與其相應的最佳化契約容量進行比較，當預測用電量中一者大於相應的最佳化契約容量時，即表示當預測的二十四個小時用電量其中有一個小時 的用電量超過預期的最佳化契約容量時，處理裝置120會在相應超過預期的最佳化契約容量的上述時間點時監控電網112的負載，並產生管制訊號122至電力管理裝置110。接著，電力管理裝置110會根據管制訊號122調降電網112的負載。例如：今日預測明日中午12點用電量是77kW，而最佳化契約容量是75kW，處理裝置120會在隔日中午12點(即超過預期的最佳化契約容量75kW的時間點)時監控電網112的負載。在一些實施例中，處理裝置120更計算當時實際用電量與最佳化契約容量之間的差值，並根據該差值傳送管制訊號122至電力管理裝置110。 For example, in some embodiments, the processing device 120 compares the predicted power consumption of the corresponding twenty-four hours with its corresponding optimal contract capacity, and when one of the predicted power usage is greater than the corresponding optimal contract capacity, capacity, that is, when one hour of the predicted twenty-four-hour electricity consumption When the power consumption exceeds the expected optimal contract capacity, the processing device 120 will monitor the load of the power grid 112 at the corresponding time point when the expected optimal contract capacity is exceeded, and generate a control signal 122 to the power management device 110 . Then, the power management device 110 reduces the load of the power grid 112 according to the control signal 122 . For example: Today's prediction is that electricity consumption is 77kW at 12:00 noon tomorrow, and the optimal contracted capacity is 75kW. The processing device 120 will monitor the power grid at 12:00 noon the next day (that is, the time point when the expected optimal contracted capacity of 75kW is exceeded). 112 loads. In some embodiments, the processing device 120 further calculates the difference between the current actual power consumption and the optimized contract capacity, and transmits the control signal 122 to the power management device 110 based on the difference.

在一些實施例中，調降電網112的負載可包含，例如：減載或卸載連接至電網112的用電裝置。 In some embodiments, reducing the load of the grid 112 may include, for example, reducing loads or unloading electrical devices connected to the grid 112 .

第2A圖是根據一些實施例的能源需量管理系統100之用電量的示意圖，以及第2B圖是根據一些實施例的能源需量管理系統100之預測用電量的示意圖。接著將參照第1圖的能源需量管理系統100說明第2A圖與第2B圖。 FIG. 2A is a schematic diagram of the power consumption of the energy demand management system 100 according to some embodiments, and FIG. 2B is a schematic diagram of the predicted power consumption of the energy demand management system 100 according to some embodiments. Next, FIGS. 2A and 2B will be described with reference to the energy demand management system 100 of FIG. 1 .

如第2A圖所示，直線L1表示原本的契約容量，直線L2表示最佳化契約容量，曲線C1表示實際用電量，曲線C2表示預測用電量，其中預測用電量C2由處理裝置120根據特徵資料透過類神經網路所預測。在一些實施例中，特徵資料相應於多個時間段，並包含氣象溫溼度資料、產能資料、即時需量資料、契約容量資料或以上的結合， 上述產能資料可以是包括產出和投入的多維產能資料，上述即時需量資料可以是即時用電最大需量。 As shown in Figure 2A, the straight line L1 represents the original contracted capacity, the straight line L2 represents the optimized contracted capacity, the curve C1 represents the actual power consumption, and the curve C2 represents the predicted power consumption, where the predicted power consumption C2 is obtained by the processing device 120 Predicted through neural network based on feature data. In some embodiments, the characteristic data corresponds to multiple time periods and includes meteorological temperature and humidity data, production capacity data, real-time demand data, contract capacity data, or a combination of the above, The above-mentioned production capacity data may be multi-dimensional production capacity data including output and input, and the above-mentioned real-time demand data may be the maximum real-time electricity demand.

如第1圖與第2A圖所示，在一些實施例中，處理裝置120取得最佳化契約容量L2與實際用電量C1的差值，並根據差值產生管制訊號122以調整負載。舉例而言，當最佳化契約容量L2減去實際用電量C1小於0.5MW(例如，在第2A圖的時間點T1)時，處理裝置120產生用於減載電網112的用電裝置的管制訊號122。 As shown in Figures 1 and 2A, in some embodiments, the processing device 120 obtains the difference between the optimal contract capacity L2 and the actual power consumption C1, and generates a control signal 122 according to the difference to adjust the load. For example, when the optimal contract capacity L2 minus the actual power consumption C1 is less than 0.5MW (for example, at the time point T1 in FIG. 2A), the processing device 120 generates a load reduction function for the power consumption device of the power grid 112. Control signal 122.

而當最佳化契約容量L2低於實際用電量C1時，處理裝置120產生用於啟動電網112的發電機的管制訊號122，以在尖峰用電時補充供電量，避免供電量不足的情況。 When the optimal contract capacity L2 is lower than the actual power consumption C1, the processing device 120 generates a control signal 122 for starting the generator of the power grid 112 to supplement the power supply during peak power consumption and avoid insufficient power supply. .

在另一實施例中，當最佳化契約容量L2減去實際用電量C1大於0.5MW時，處理裝置120保持監控電網112的負載，並且耦接處理裝置120的顯示裝置126根據上述用電狀態顯示不同燈號。 In another embodiment, when the optimized contract capacity L2 minus the actual power consumption C1 is greater than 0.5MW, the processing device 120 keeps monitoring the load of the power grid 112, and the display device 126 coupled to the processing device 120 adjusts the power consumption according to the above power consumption. The status shows different lights.

在一些實施例中，預測用電量C2由處理裝置120透過類神經網路預測所得。在訓練類神經網路的過程中，處理裝置120以多個訓練特徵資料與對應的多個訓練預測用電量進行訓練，上述類神經網路包含卷積神經網路(Convolutional Neural Network，CNN)。 In some embodiments, the predicted power consumption C2 is predicted by the processing device 120 through a neural network. In the process of training a neural network, the processing device 120 performs training with multiple training feature data and corresponding multiple training prediction power consumptions. The above-mentioned neural network includes a convolutional neural network (CNN). .

接著，處理裝置120根據多個驗證特徵資料預測多個預測用電量，並在預測用電量不同於相應驗證特徵資料的驗證預測用電量時，處理裝置120的至少一參數被調 整。在另一實施例中，如第2B圖所示，處理裝置120透過類神經網路根據多個驗證特徵資料預測多個預測用電量，並在多個預測用電量中的一者(例如，P點)超過信賴預測區間時，處理裝置120的至少一參數被調整。 Next, the processing device 120 predicts a plurality of predicted power consumption according to the plurality of verification feature data, and when the predicted power consumption is different from the verified predicted power consumption of the corresponding verification feature data, at least one parameter of the processing device 120 is adjusted. all. In another embodiment, as shown in FIG. 2B , the processing device 120 predicts a plurality of predicted power consumption according to a plurality of verification characteristic data through a neural network, and performs the calculation on one of the plurality of predicted power consumption (for example, , point P) exceeds the reliable prediction interval, at least one parameter of the processing device 120 is adjusted.

舉例而言，以預測二十四個小時的最大用電量為例，處理裝置120以七十二個小時的特徵資料與相應的用電量訓練卷積神經網路並驗證。處理裝置120藉由卷積神經網路中的濾波器篩選全部七十二個小時的特徵資料，以取得七十二個小時的特徵資料的權重，其中篩選條件可包含，例如：用電量的多寡、何時具有最大用電量等等。在驗證經訓練的卷積神經網路的過程中，當七十二個小時的驗證預測用電量與相應七十二小時的預測用電量不同時，處理裝置120會調整上述權重。而當預測用電量(例如，第2B圖的P點)超過信賴預測區間(即曲線C3與C4之間的區域)時，處理裝置120會調整上述權重。接著，處理裝置120利用經驗證的權重取得二十四個小時的最大用電量。藉此調整權重的方式，使預測用電量允許更大的誤差以更貼近實際操作情況。 For example, taking the prediction of the maximum power consumption for twenty-four hours as an example, the processing device 120 uses the characteristic data of seventy-two hours and the corresponding power consumption to train the convolutional neural network and verify it. The processing device 120 filters all 72 hours of feature data through the filter in the convolutional neural network to obtain the weight of the 72 hours of feature data. The filtering conditions may include, for example: electricity consumption. How much, when has the maximum power consumption, etc. In the process of validating the trained convolutional neural network, when the verified predicted power consumption for seventy-two hours is different from the predicted power consumption for the corresponding seventy-two hours, the processing device 120 will adjust the above-mentioned weights. When the predicted power consumption (for example, point P in Figure 2B) exceeds the reliable prediction interval (ie, the area between curves C3 and C4), the processing device 120 will adjust the above weight. Then, the processing device 120 uses the verified weight to obtain the maximum power consumption for twenty-four hours. By adjusting the weights, the predicted power consumption allows for a larger error to be closer to the actual operating conditions.

在另一實施例中，相對於等於或大於二十四小時的時間段，處理裝置120透過卷積神經網路預測相應的預測用電量。對於預測小於二十四小時的時間段的用電量預測而言，以預測兩個小時的用電量為例，處理裝置120利用八個小時的特徵資料及對應的用電量訓練長短期記憶(Long Short-term Memory，LSTM)網路並驗證，以 取得兩個小時的最大用電量。長短期記憶網路是一種具有時間循環的類神經網路，可以在任意時間間隔內儲存資訊，並在後續操作中使用前一個時間段所儲存的資訊。舉例而言，每十五分鐘的資料作為一筆特徵資料，處理裝置120透過長短期記憶網路將前一筆特徵資料(即前十五分鐘的資料)與後一筆特徵資料(即後十五分鐘的資料)作為輸入資料，並結合上述特徵資料的狀態條件(例如，是否處於高溼度)與相應的用電量，以取得兩個小時的最大用電量。 In another embodiment, relative to a time period equal to or greater than twenty-four hours, the processing device 120 predicts the corresponding predicted power consumption through a convolutional neural network. For prediction of electricity consumption for a time period less than twenty-four hours, taking the prediction of electricity consumption for two hours as an example, the processing device 120 uses eight hours of characteristic data and corresponding electricity consumption to train long and short-term memory. (Long Short-term Memory, LSTM) network and verification to Get the maximum power usage for two hours. The long short-term memory network is a neural network with a time loop that can store information at any time interval and use the information stored in the previous time period in subsequent operations. For example, every fifteen minutes of data is regarded as a piece of characteristic data, and the processing device 120 uses the long short-term memory network to combine the previous piece of characteristic data (i.e., the data of the previous fifteen minutes) and the next piece of characteristic data (i.e., the data of the next fifteen minutes). data) as input data, and combine the status conditions of the above characteristic data (for example, whether it is in high humidity) with the corresponding power consumption to obtain the maximum power consumption for two hours.

在另一實施例中，針對短暫性特殊用電的情況，處理裝置120以具有特殊用電資料的特徵資料及對應的用電量訓練卷積神經網路並驗證，並在相應特殊用電資料的預測用電量超過信賴預測區間(例如，第2B圖的P點)時，處理裝置120將權重調整至進入信賴預測區間，藉此使特殊用電的預測用電量更貼近實際短暫性的特殊用電情況。 In another embodiment, for the situation of temporary special power consumption, the processing device 120 uses the characteristic data with special power consumption data and the corresponding power consumption to train and verify the convolutional neural network, and in the corresponding special power consumption data When the predicted power consumption exceeds the reliable prediction interval (for example, point P in Figure 2B), the processing device 120 adjusts the weight to enter the reliable prediction interval, thereby making the predicted power consumption of the special power closer to the actual short-term power consumption. Special power usage situations.

在一些實施例中，處理裝置120根據對應多個較短時間段(例如，上述實施例中的二十四個小時)的多個預測用電量預測對應多個較長時間段的多個預測用電量。舉例而言，當取得的二十四個小時預測用電量C2皆小於相應的最佳化契約容量L2時，處理裝置120透過卷積神經網路根據二十四個小時的預測用電量C2預測十二個月的預測用電量，以取得十二個月的最佳化契約容量。 In some embodiments, the processing device 120 predicts multiple predictions corresponding to multiple longer time periods based on multiple predicted power consumption corresponding to multiple shorter time periods (eg, twenty-four hours in the above embodiment). Electricity usage. For example, when the obtained twenty-four-hour predicted power consumption C2 is less than the corresponding optimized contract capacity L2, the processing device 120 uses the convolutional neural network to calculate the predicted power consumption C2 according to the twenty-four-hour forecast. Forecast the forecast electricity consumption for twelve months to obtain the optimal contract capacity for twelve months.

在一些實施例中，最佳化契約容量由處理裝置120根據電費公式取得與十二個月中的最大用電量相應的電費，再結合過去契約容量L1所得。在一些實施例中，電費公式 所需的電費參數包含基本電費、超約附加費、復電費等等。在另一實施例中，如第1圖所示，耦接處理裝置120的輸入裝置124接收用於調整電費參數的使用者輸入128。 In some embodiments, the optimized contract capacity is obtained by the processing device 120 obtaining the electricity fee corresponding to the maximum electricity consumption in twelve months according to the electricity fee formula, and then combining it with the past contract capacity L1. In some embodiments, the electricity bill formula The required electricity fee parameters include basic electricity fee, over-contract surcharge, recharge fee, etc. In another embodiment, as shown in FIG. 1 , the input device 124 coupled to the processing device 120 receives user input 128 for adjusting electricity bill parameters.

與習知一年一次以人力調整契約容量的方式相比，藉由本案的配置，可節省至少2%的費用，達到有效且即時能源效率調控。 Compared with the traditional method of manually adjusting the contract capacity once a year, through the configuration of this case, at least 2% of the cost can be saved, achieving effective and real-time energy efficiency control.

第3圖是根據一些實施例所繪示的能源需量的管理方法300的流程圖。能源需量的管理包含步驟310、步驟320、步驟330、步驟340以及步驟350。接著將參照第1圖的能源需量管理系統100說明能源需量的管理方法300。 Figure 3 is a flow chart of an energy demand management method 300 according to some embodiments. The management of energy demand includes step 310 , step 320 , step 330 , step 340 and step 350 . Next, the energy demand management method 300 will be described with reference to the energy demand management system 100 in FIG. 1 .

如第3圖所示，在步驟310中，透過處理裝置120中的類神經網路根據多個特徵資料預測對應多個第一時間段的多個第一預測用電量以取得最大用電量時間段。在一些實施例中，第一時間段相應於一日，處理裝置120透過卷積神經網路預測七日的預測用電量，以取得七日內的最大用電量時間段。 As shown in Figure 3, in step 310, the neural network in the processing device 120 predicts a plurality of first predicted power consumption corresponding to a plurality of first time periods according to a plurality of characteristic data to obtain the maximum power consumption. time period. In some embodiments, the first time period corresponds to one day, and the processing device 120 predicts the predicted power consumption for seven days through a convolutional neural network to obtain the maximum power consumption time period within seven days.

在步驟320中，透過處理裝置120中的類神經網路預測對應多個第二時間段的多個第二預測用電量。在一些實施例中，第二時間段相應於一個小時，處理裝置120透過卷積神經網路預測二十四個小時的預測用電量，以取得二十四個小時內的最大用電量時間段。 In step 320 , a plurality of second predicted power consumption corresponding to a plurality of second time periods are predicted through the neural network in the processing device 120 . In some embodiments, the second time period corresponds to one hour, and the processing device 120 predicts the predicted power consumption for twenty-four hours through a convolutional neural network to obtain the maximum power consumption time within twenty-four hours. part.

在步驟330中，將多個第二預測用電量與多個最佳化契約容量之一者相比較。在一些實施例中，處理裝置 120將二十四個小時的預測用電量與相應的最佳化契約容量進行比較。 In step 330, a plurality of second predicted power usages are compared with one of a plurality of optimized contract capacities. In some embodiments, the processing device 120 Compares the twenty-four-hour forecasted electricity usage with the corresponding optimal contract capacity.

在一些實施例中，當第二預測用電量小於最佳化契約容量之一者時，能源需量的管理方法300執行步驟340。相對地，當第二預測用電量中的至少一者大於最佳化契約容量中的一者時，能源需量的管理方法300執行步驟350。 In some embodiments, the energy demand management method 300 performs step 340 when the second predicted power consumption is less than one of the optimized contract capacities. Correspondingly, when at least one of the second predicted power consumption is greater than one of the optimized contract capacities, the energy demand management method 300 executes step 350 .

在步驟340中，透過類神經網路根據第二預測用電量預測對應多個第三時間段的多個第三預測用電量，處理裝置120根據多個第三預測用電量、電費公式以及過去契約容量計算對應於多個第三時間段的最佳化契約容量。在一些實施例中，第三時間段相應於一個月。處理裝置120透過卷積神經網路根據二十四個小時的用電量預測相應十二個月的預測用電量，以根據十二個月的預測用電量由電費公式與過去契約容量計算最佳化契約容量。 In step 340, a plurality of third predicted power consumption corresponding to a plurality of third time periods are predicted through the neural network based on the second predicted power consumption. and past contract capacity calculations corresponding to optimized contract capacities for multiple third time periods. In some embodiments, the third time period corresponds to one month. The processing device 120 predicts the predicted electricity consumption for the corresponding twelve months based on the twenty-four-hour electricity consumption through the convolutional neural network, and calculates the electricity bill formula and past contract capacity based on the twelve-month predicted electricity consumption. Optimize contract capacity.

在步驟350中，透過電力管理裝置110調整電網112的負載，其中透過電力管理裝置110調整電網112的負載包含：卸載連接於電網112的用電裝置。舉例而言，在一些實施例中，於二十四個小時的預測用電量中有一個小時的預測用電量大於相應的最佳化契約容量時，處理裝置120根據最佳化契約容量與實際用電量間的差值產生管制訊號122至電力管理裝置110，以透過電力管理裝置110調整電網112的負載。 In step 350 , the load of the power grid 112 is adjusted through the power management device 110 , where adjusting the load of the power grid 112 through the power management device 110 includes: unloading electrical devices connected to the power grid 112 . For example, in some embodiments, when the predicted power consumption for one hour among the twenty-four hours of predicted power consumption is greater than the corresponding optimal contract capacity, the processing device 120 determines the optimal contract capacity and the corresponding optimal contract capacity. The difference between the actual power consumption generates a control signal 122 to the power management device 110 to adjust the load of the power grid 112 through the power management device 110 .

綜上所述，本案藉由所提及的能源需量管理系統與 方法，基於以類神經網路預測的最佳化契約容量調控電網的負載，提前掌握能源使用趨勢，以即時調度電力，節省基本電費，進而使能源的調控更有效率、更精準。 In summary, this case uses the mentioned energy demand management system and This method regulates the load of the power grid based on the optimized contract capacity predicted by a neural network, and grasps the energy usage trend in advance to dispatch power in real time, save basic electricity bills, and thereby make energy regulation more efficient and precise.

雖然本案已以實施例揭露如上，然其並非用以限定本案，任何所屬技術領域中具有通常知識者，在不脫離本案的精神和範圍內，當可作些許的更動與潤飾，故本案的保護範圍當視後附的申請專利範圍所界定者為準。 Although this case has been disclosed as above using embodiments, they are not intended to limit this case. Anyone with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of this case. Therefore, the protection of this case The scope shall be determined by the appended patent application scope.

100:能源需量管理系統 100:Energy Demand Management System

110:電力管理裝置 110:Power management device

112:電網 112:Power grid

120:處理裝置 120: Processing device

122:管制訊號 122:Control signal

124:輸入裝置 124:Input device

126:顯示裝置 126:Display device

128:使用者輸入 128:User input

Claims (5)

一種能源需量管理系統，包含：一電力管理裝置，用以調控一電網的一負載；以及一處理裝置，耦接該電力管理裝置，其中該處理裝置用以根據複數個特徵資料透過一類神經網路預測對應複數個第一時間段的複數個第一預測用電量，並進一步用以將該些第一預測用電量與一最佳化契約容量相比較，且更用以透過該類神經網路根據複數個驗證特徵資料預測複數個預測用電量，其中當對應於該些第一時間段之一第一者的該些第一預測用電量的一第一者大於該最佳化契約容量時，該處理裝置更用以在該些第一時間段之該第一者時監控該負載以產生一管制訊號至該電力管理裝置，其中該處理裝置更用以根據一過去契約容量，以及透過一電費公式取得的一電費資料，計算該最佳化契約容量，其中當該最佳化契約容量低於一實際用電量時，該處理裝置更用以啟動該電網的一發電機，以提高該電網的供電量，其中當該些預測用電量中的一者超過一信賴預測區間時，該處理裝置的至少一參數被調整，以及該電力管理裝置更用以響應於該管制訊號調降該負載。 An energy demand management system includes: a power management device used to regulate a load of a power grid; and a processing device coupled to the power management device, wherein the processing device is used to use a type of neural network according to a plurality of characteristic data The road prediction corresponds to a plurality of first predicted power consumption in a plurality of first time periods, and is further used to compare the first predicted power consumption with an optimized contract capacity, and is further used to use the neural network The network predicts a plurality of predicted power usages based on a plurality of verification feature data, wherein a first one of the first predicted power usages corresponding to a first one of the first time periods is greater than the optimization When contracting capacity, the processing device is further used to monitor the load at the first time period to generate a control signal to the power management device, wherein the processing device is further used to monitor the load according to a past contracted capacity, and calculating the optimal contract capacity through an electricity fee data obtained through an electricity fee formula, wherein when the optimal contract capacity is lower than an actual electricity consumption, the processing device is further used to start a generator of the power grid, To increase the power supply of the power grid, when one of the predicted power consumption exceeds a reliable prediction interval, at least one parameter of the processing device is adjusted, and the power management device is further used to respond to the control signal Reduce the load. 如請求項1所述之能源需量管理系統，其中該些特徵資料對應於該些第一時間段，並包含一氣象溫溼 度資料、一產能資料、一即時需量資料、一契約容量資料或以上的結合。 The energy demand management system as described in claim 1, wherein the characteristic data corresponds to the first time periods and includes a weather temperature and humidity degree data, a production capacity data, a real-time demand data, a contract capacity data or a combination of the above. 如請求項1所述之能源需量管理系統，其中該處理裝置更用以透過該類神經網路對複數個訓練特徵資料與對應的複數個訓練預測用電量進行訓練，其中在該些預測用電量不同於該些驗證特徵資料對應的複數個驗證預測用電量時，該處理裝置的該至少一參數被調整，其中該類神經網路包含一卷積神經網路。 The energy demand management system of claim 1, wherein the processing device is further used to train a plurality of training feature data and a plurality of corresponding training predictions of electricity consumption through the neural network, wherein in the predictions When the power consumption is different from the plurality of verification prediction power consumption corresponding to the verification characteristic data, the at least one parameter of the processing device is adjusted, wherein the type of neural network includes a convolutional neural network. 如請求項1所述之能源需量管理系統，其中該處理裝置更用以根據該些第一預測用電量預測對應複數個第二時間段的複數個第二預測用電量，其中該些第二時間段的每一者的時間大於該些第一時間段的每一者的時間。 The energy demand management system of claim 1, wherein the processing device is further configured to predict a plurality of second predicted power consumption corresponding to a plurality of second time periods based on the first predicted power consumption, wherein the The time of each of the second time periods is greater than the time of each of the first time periods. 如請求項1所述之能源需量管理系統，該處理裝置更用以計算該最佳化契約容量與該實際用電量的一差值，並根據該差值產生該管制訊號以調整該負載。 For the energy demand management system described in claim 1, the processing device is further used to calculate a difference between the optimized contract capacity and the actual power consumption, and generate the control signal according to the difference to adjust the load. .
TW110148506A 2021-12-23 2021-12-23 Management system of energy demand TWI832131B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
TW110148506A TWI832131B (en) 2021-12-23 2021-12-23 Management system of energy demand
CN202210498304.XA CN114676938A (en) 2021-12-23 2022-05-09 Energy demand management system and method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
TW110148506A TWI832131B (en) 2021-12-23 2021-12-23 Management system of energy demand

Publications (2)

Publication Number Publication Date
TW202326578A TW202326578A (en) 2023-07-01
TWI832131B true TWI832131B (en) 2024-02-11

Family

ID=82079880

Family Applications (1)

Application Number Title Priority Date Filing Date
TW110148506A TWI832131B (en) 2021-12-23 2021-12-23 Management system of energy demand

Country Status (2)

Country Link
CN (1) CN114676938A (en)
TW (1) TWI832131B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9002761B2 (en) * 2008-10-08 2015-04-07 Rey Montalvo Method and system for automatically adapting end user power usage
TWI504094B (en) * 2013-08-02 2015-10-11 元智大學 Power load monitoring and predicting system and method thereof
CN105512455A (en) * 2014-10-07 2016-04-20 英科德技术股份有限公司 Method and apparatus for estimating power consumption based on temperature

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102034143A (en) * 2010-10-26 2011-04-27 中华电信股份有限公司 Cost-saving energy-saving management system and method thereof
KR101865924B1 (en) * 2011-10-24 2018-06-11 한국전력공사 Appratus and method for estimation of weekly power load to improve processing time using neural network and revision factor
JP6221337B2 (en) * 2013-05-13 2017-11-01 富士電機株式会社 Demand control device and demand control system
CN104751233A (en) * 2013-12-27 2015-07-01 台达电子工业股份有限公司 Contract capacity optimization system and optimization method
CN106934497B (en) * 2017-03-08 2021-04-16 青岛卓迅电子科技有限公司 Intelligent community power consumption real-time prediction method and device based on deep learning
KR102327413B1 (en) * 2021-02-25 2021-11-16 국민대학교산학협력단 Power managing apparatus and method for the same

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9002761B2 (en) * 2008-10-08 2015-04-07 Rey Montalvo Method and system for automatically adapting end user power usage
TWI504094B (en) * 2013-08-02 2015-10-11 元智大學 Power load monitoring and predicting system and method thereof
CN105512455A (en) * 2014-10-07 2016-04-20 英科德技术股份有限公司 Method and apparatus for estimating power consumption based on temperature

Also Published As

Publication number Publication date
TW202326578A (en) 2023-07-01
CN114676938A (en) 2022-06-28

Similar Documents

Publication Publication Date Title
US11714441B2 (en) Method and apparatus for delivering power using external data
US11201491B2 (en) Method for balancing frequency instability on an electric grid using networked distributed energy storage systems
CN102075014B (en) Large grid real-time scheduling method for accepting access of wind power
CN107230984B (en) A method and control system for power demand response to participate in distribution network voltage regulation
CN107887933A (en) A kind of Multiple Time Scales rolling optimization microgrid energy optimum management method
CN108565874A (en) A kind of source lotus collaboration frequency modulation method based on LOAD FREQUENCY Controlling model
CN103001219A (en) Optimal coordination control method of multiple flexible alternative current transmission systems (FACTSs) based on trend entropy
CN104699051A (en) Demand response control method of temperature control device
JP2017050919A (en) Demand power prediction device, demand power prediction method and computer program
TWI832131B (en) Management system of energy demand
CN104156785A (en) Generation schedule optimization method taking into consideration thermal power unit-driven coal mill start and stop
US12015269B2 (en) Methods, systems, and apparatuses for the reset of a setpoint for committed demand
JP7557464B2 (en) Self-delivery system and self-delivery method
JP7492957B2 (en) Self-delivery system and self-delivery method
CN110994639B (en) Simulation constant volume method, device and equipment for power plant energy storage auxiliary frequency modulation
CN106816878A (en) A kind of demand response regulation device and method for being applied to drainage and irrigation hyelectric power facility
CN109167350B (en) Construction method of industrial load response model
CN111832898A (en) A Demand Response Scheduling Method for Power System Based on Air Conditioning Multifunction
CN108631368B (en) Energy storage configuration method for wind storage system joint dispatch considering energy storage operation loss
CN115954952A (en) Flexible resource planning method based on time sequence operation simulation
CN106655175A (en) Intelligent method for scheduling and optimizing electricity of resident users
CN104299055A (en) Power generation plan optimizing method for restraining unit reverse regulation in power plant
Li et al. Preventive control considering aggregated controllable loads for voltage stability in electrical power systems
Liao et al. Low-carbon demand response strategy of power system based on event-driven mechanism
CN119382139A (en) A virtual power plant dispatching system and method for power transmission