Bidding system for electric power

Abstract

To provide a bidding system for electric power that generates bidding data with a high profitability in an electric power trading market by optimizing the bidding data using a deep neural network even in a trading environment in which the amount of electric power transmitted from secondary batteries change over time.SOLUTION: A bidding system for electric power that communicates with: a server device which provides bidding data to a bidding platform for electric power; and a secondary battery system which transmits electric power stored by secondary batteries to an electric power system or charges the secondary batteries by the electric power supplied from the electric power system to store the electric power includes: an input layer 23 formed by a plurality of parameters necessary for identifying a bidding price 26; an arithmetic layer 24 that performs arithmetic processing using a linear model in accordance with a deep neural network, an algorithm or a model including them; and an output layer 25 that generates bidding information in accordance with the arithmetic processing performed by the arithmetic layer 24. The bidding system generates the bidding data compatible with a power demand that changes over time.SELECTED DRAWING: Figure 4

Description

Technical Field

[0001] The present invention relates to a power bidding system for a battery system for grid power supply using a storage battery.

Background Art

[0002] Conventionally, as power bidding systems, the systems disclosed in Patent Documents 1 to 4 below have been published.

[0003] In the power bidding trading market, unlike the securities trading market etc., since it is premised on the delivery of electricity which is a physical good, bidding in accordance with the power supply and demand plan must be carried out so as not to conduct sell bids exceeding the power generation capacity of the power plant or buy bids exceeding the demand etc.

[0004] Also, except in emergencies such as generator dropout, losses due to power trading must be prevented. For example, when there is a large deviation between the predicted demand value expected before the delivery of electricity and the actual demand, a transaction that was generating a profit at the time of agreement may result in a loss because the self-generation cost fluctuates (increases) due to the occurrence of demand fluctuations at the time of electricity delivery.

[0005] Traders who conduct power bidding transactions must identify risks such as these demand fluctuations and generator troubles, and constantly monitor the bidding operations and their own demand. Also, conventionally, the operation mode in which traders manually create bidding information was mainly used, but in the power trading of the Zaraba method, especially in the "one-hour-ahead market trading" which is open 24 hours a day, it has been pointed out that this operation mode places a heavy burden on traders.

[0006] For this reason, Patent Document 1 below discloses that "in power trading, since power grid participants need to formulate bidding plans and power generation plans in consideration of the uncertainty of grid trading, a means 0102 for receiving an operation plan formulated for a power demand with a confirmed supply and a means 0106 for formulating a plurality of bidding plans for the grid based on the received plan are provided. At this time, by having the conditions specified for the uncertainty of the electricity unit price in grid settlement (0104), it becomes possible to perform predictive calculations on the transaction volume and revenue fluctuations resulting from the bidding results. Also, by having the conditions specified for the prediction of revenue and transaction volume, a means for screening the bidding plan required by the operator from a plurality of bidding plans is provided."

[0007] Patent Document 2 below discloses that "to provide an electricity contract mediation device that enables electricity consumers to select the cheapest electricity with a low burden and enables electricity suppliers to obtain opportunities for transactions with many electricity consumers with a low burden, the contract type DB109 stores information on each contract type. The bidding item registration process 102 registers the bidding items input from the electricity consumer terminal in the bidding item DB108. The automatic bidding process 107 creates bidding information for the bidding items registered in the bidding item DB109 based on the information on each contract type stored in the contract type DB109, associates the information with the bidding items, and stores the information in the bidding item DB108. The winning bid notification process 105 selects the winning bid from the bidding information stored in the bidding item DB108 associated with the registered bidding items and notifies the electricity supplier terminal and the electricity consumer terminal that input the bidding items that the bid has won."

[0008] Patent Document 3 below discloses that "in order to provide an easy-to-use automatic bidding system for the power trading market, an automatic bidding system 1 for the power trading market that generates bidding information for the power trading market includes an automatic bidding setting creation unit 11 that creates automatic bidding setting information including a plurality of setting values, an automatic bidding information creation unit 13 that creates automatic bidding information used in the power exchange system 3 based on the automatic bidding setting information created by the automatic bidding setting creation unit, and a bidding information transmission unit 15 that transmits the automatic bidding information created by the automatic bidding information creation unit to the power exchange system. The automatic bidding information creation unit changes the content of the automatic bidding information based on one or more predetermined setting values among the plurality of setting values."

[0009] Patent Document 4 below discloses that "in order to make it easier for a user to purchase power under conditions preferred by the user through automatic bidding, the automatic bidding system includes a vehicle agent 500 (computer). The vehicle agent 500 includes a bidding agent 530 (bidding unit) that performs automatic bidding for power trading according to an automatic bidding algorithm regarding the user's power, and an information collection agent 510 (information collection unit) that acquires user information indicating the user's preferences. The bidding agent 530 is configured to set parameters of the automatic bidding algorithm using the user information."

[0010] Note that Non-Patent Documents 1 to 3 below are publicly available as non-patent documents related to this application.

Prior Art Documents

Patent Documents

[0011]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Non-Patent Documents

[0012] [Non-Patent Document 1] “Statistical arbitrage trading across electricity markets using advantage actor-critic methods” Sustainable Energy, Grids and Networks 34(2023)101023 [Non-Patent Document 2] “Energy Storage Arbitrage in Real-Time Markets via Reinforcement Learning” Department of Electrical Engineering, University of Washington, Seattle, WA98195 [Non-Patent Document 3] “Energy Storage Arbitrage Under Day-Ahead and Real-Time Price Uncertainty” IEEE Transactions on Power Systems (Volume: 33, Issue: 1, January 2018) [Summary of the Invention] [Problems to be Solved by the Invention]

[0013] However, in the initial stage of conventional power trading bidding prices, since price increases or decreases were predicted and bidding was carried out based on human discretion, the accuracy of human discretion depends on the individual, and it has been pointed out that it is difficult for a single person to conduct transactions at their discretion in the electricity market, which is a 24-hour market.

[0014] Therefore, in the published automatic bidding system, although the bidding is automatic, a method called a linear model has been used, which statistically processes and quantifies input data and outputs it as a predicted value of the bidding price. Although these automatic bidding systems based on linear models show certain prediction results, they cannot be said to have sufficiently high accuracy for the bidding price compared to the calculated value of the back data based on the obtained values, and further improvement has been desired. These factors are due to the fact that the target input data focuses on artificial elements and does not focus on multiple parameters that vary depending on the natural environment of the storage battery and the weather, resulting in a deviation between the bidding price and the expected value.

[0015] The present invention has been made to solve the above problems. Even in a trading environment where the amount of power transmitted from a storage battery varies with time, by optimizing bidding data by focusing on multiple parameters using a deep neural network and an algorithm, it is possible to generate highly profitable bidding data in the power trading market, and an object of the present invention is to provide a power bidding system that can expand profits and the life of the storage battery and improve investment efficiency.

Means for Solving the Problems

[0016] The power bidding system according to the present invention is a power bidding system that communicates with a server device that provides bidding data to a power bidding platform via a predetermined communication medium and a battery system that charges the battery using the power stored in the battery to transmit power to the power grid or the power supplied from the power grid. It is characterized by comprising a generation means for generating bidding data that adapts to the constantly fluctuating power demand, comprising an input layer with a plurality of parameters necessary for specifying the bidding price, an arithmetic layer that performs arithmetic processing using a linear model or an algorithm according to a deep neural network and a model including them, and an output layer that generates bidding information by the arithmetic processing executed by the arithmetic layer.

Effects of the Invention

[0017] According to the present invention, even in a trading environment where the amount of power transmitted from a storage battery varies with time, by optimizing bidding data using a deep neural network or an algorithm and a model including them, highly profitable bidding data can be generated in the power trading market.

Brief Description of the Drawings

[0018] The drawings show specific embodiments of the present invention according to the present disclosure, including not only essential configurations of the invention but also optional and preferred embodiments.

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Embodiments for Carrying Out the Invention

[0019] 〔Embodiment〕 Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0020] FIG. 1 is a configuration diagram for explaining the concept of a bidding system in a grid-connected battery system using a storage battery. The storage battery is assumed to be connected to a grid power source located in a remote area. This system includes a server device that provides bidding data to a power bidding platform via a predetermined communication medium, and a power bidding system that communicates with a battery system that transmits the power stored in the battery to the power grid or charges the battery using the power supplied from the power grid to accumulate power. However, the system may be configured to include additional power facilities and monitoring systems. In addition, in this specification, the configuration described as a storage battery is a general term for examples of storage batteries that function as DC power sources including capacitor systems such as supercapacitors and lithium-ion capacitors, and secondary batteries such as lithium-ion batteries (including NAS batteries, etc.).

[0021] In FIG. 1, reference numeral 1 denotes a data terminal operated by a trading operator, which is configured to be connectable to a cloud server 3 via a network (not shown) by communication based on a predetermined protocol. It is assumed that an account (ID and PASS) has been registered in the data terminal 1 in advance. Reference numeral 2 denotes a market terminal, which is configured to be connectable to the cloud server 3 via a network (not shown) by communication based on a predetermined protocol. Although FIG. 1 shows an example in which one trading operator is connected, it is of course assumed that a plurality of operators may be connected simultaneously.

[0022] The cloud server 3 is configured to be connectable to an LTE base station 4 via a network (not shown) by communication based on a predetermined protocol. Reference numeral 5 denotes an LTE receiver, which executes a process of transmitting management data indicating the operating status including the amount of power stored in the storage battery 6 and the amount of power transmitted to the cloud server 3 at regular or optional timings.

[0023] Reference numeral 7 denotes charging power, which is AC power supplied from the grid-connected power source 9, and is converted into DC power via an AD converter (not shown) and received by the storage battery 6 at a predetermined charging current. The storage battery 6 is provided with a power storage control controller that switches between a charging cycle and a discharging cycle.

[0024] 8 is the discharge power. After converting the DC power stored by the charging cycle into AC power through DA conversion, it is transformed and supplies AC power of a predetermined frequency to the grid-connected power source 9.

[0025] This power bidding system is composed of a cloud server 3 connected to the power trading market. In addition, there are a monitor for humans to confirm the overall control and control status, the bidding price, charging and discharging capacity determined by the cloud server 3, the control data of the EMS, a data terminal 1 that transmits the control data and receives data to control the EMS, and a storage battery 6 controlled by the EMS. The storage battery 6 is connected to the grid-connected power source 9.

[0026] Note that the cloud server 3 is equipped with a CPU, ROM, and RAM as hardware resources, and is configured to freely perform the statistical processing described later by loading the control program stored in the external memory into the RAM and executing it.

[0027] The cloud server 3 configured in this way has a plurality of parameters necessary for specifying the bidding price as an input layer 23, an arithmetic layer 24 that performs arithmetic processing using a linear model or algorithm according to a deep neural network and a model including them, and an output layer 25 that generates bidding information by the arithmetic processing executed by the arithmetic layer 24. It stores, as a program, a generation unit that generates bidding data adaptable to the power demand that fluctuates moment by moment, in the external memory.

[0028] Also, the generation unit executes a process of generating a bidding price and a bidding capacity as bidding data.

[0029] Furthermore, the generation unit executes a process of generating the optimal bidding capacity so as to indicate the discharge amount that extends the life of the storage battery 6.

[0030] Also, the above-mentioned plurality of parameters are configured to include weather forecast, spot price, time-previous price, and battery SOC.

[0031] Furthermore, the plurality of parameters includes, as other factors, the capacity of the storage battery 6 and the environmental information of the storage battery 6.

[0032] In addition, the environmental information includes the annual temperature variation, wind speed, and wind direction around the storage battery 6.

[0033] Furthermore, the generation unit executes a process of generating bidding data that adapts to the power demand that fluctuates moment by moment by linear statistical processing.

[0034] Figure 2 shows a state in which profits are generated by intraday arbitration (arbitrage trading) in the power trading market. The vertical axis indicates the power price (yen / kWh) 10, the horizontal axis indicates the time 13, and for example, it shows a 24-hour display starting from midnight.

[0035] As shown in this figure, the bidding price generally shows price fluctuations 12 as the price difference 11 between the high value and the low value according to the fluctuations in power demand over time 13.

[0036] Figure 3 is a block diagram showing a prediction method of a statistical linear model. In Figure 3, 14 is the spot price, and the bidding price 17 of the output layer 16 obtained by statistically processing the input layer 15 consisting of t = 0 to t = 47 is specified.

[0037] Therefore, in order to reduce the prediction error, the applicant of this application adopts a DNN (Deep Neural Network) method incorporating a predetermined algorithm or AI method while obtaining sequential latest information, and a configuration example is shown in Figure 4.

[0038] In Figure 4, the input layer 23 is composed of a plurality of sources, for example, weather forecast 18, spot price 19, previous time price 20, battery SOC 21, and others 22. Here, SOC is the abbreviation of State Of Charge of the storage battery 6 and is an index representing the charge rate or charge state. Also, it is an index with a fully charged state of 100% and a fully discharged state of 0%.

[0039] 24 is the calculation layer, which shows the correlation by combining a plurality of sources constituting the input layer 23, such as weather forecast 18, spot price 19, previous time price 20, battery SOC 21, and others 22. From this complex combination, as the output layer 25, it is characterized by a configuration for predicting the bidding price 26 and the bidding capacity 27.

[0040] Specifically, the controller of the data terminal 1 adopts the model shown in FIG. 4 to generate bidding data for generating bidding information for the power trading market. The output layer 25, which is the bidding data, simultaneously infers the bidding capacity 27 in addition to the bidding price 26, and maximizes the investment efficiency of the battery facility with the bidding capacity 27 that maximizes its total charge and discharge capacity.

[0041] At this time, the required data of the bidding price 26 and the bidding capacity 27 are input with the weather forecast 18, spot price 19, previous time price 20, battery SOC 21, and others 22 as the input layer 23. As an example, a deep neural network (DNN) is used in the calculation layer 24, which is a method of artificial intelligence, to finally obtain the bidding price 26 and the bidding capacity 27 that become the output layer 25.

[0042] Note that the controller of the data terminal 1 updates every 30 minutes if it is the previous time price 20, and recalculates the latest bidding price 26 and the bidding capacity 27.

[0043] FIG. 2 is a diagram showing the mechanism by which profits are generated by intraday arbitration (arbitration trading) in the power trading market. FIG. 3 shows an example of calculating the predicted value of the conventional power bidding by a linear model, and calculates the predicted value mathematically and statistically based on the conditions.

[0044] FIG. 4 shows that the predicted values used conventionally, which are mathematical and statistical, do not have a small difference from the actual transaction values. As a method for reducing these prediction errors, it shows a method of predicting the bidding value by an algorithm or an AI-based DNN method while obtaining sequential latest information, and predicting the amount of electricity corresponding to the bidding value that maximizes the battery life.

[0045] Figure 5 shows the profit values when using the algorithm and the AI-based DNN method with the profit of the agreed result predicted by the linear model in Figure 3 set to the standard value of 1.00 for the actually agreed value. Here, a significant improvement effect of 1.58 is obtained.

[0046] Also, Figure 6 shows the depth of discharge and battery life that occur in the lithium-ion battery of the energy storage battery 6 for the system. In particular, as an example, in the case of LiPO4; lithium iron phosphate ion battery, by reducing the depth of discharge, the battery life is significantly improved. In this case, it can be seen that the power amount in the total life of the energy storage battery 6 becomes three times with a 20% depth of discharge, and the investment efficiency of the profit from the capital investment can be significantly improved.

[0047] Also, it is expected that the investment efficiency can be significantly improved by the bidding value and power capacity derived by the algorithm and the AI-based DNN method shown in Figure 4. That is, the profit and battery life in the power system using the energy storage battery can be expanded and the investment efficiency can be improved.

[0048] 〔Effects of this Embodiment〕 According to this embodiment, even in a trading environment where the power amount transmitted from the energy storage battery 6 fluctuates over time, by optimizing the bidding data using a deep neural network, an algorithm, and a model including them, highly profitable bidding data can be generated in the power trading market.

[0049] The disclosure regarding the present invention described above can be summarized at least as the following matters.

[0050] (1) A power bidding system that communicates with a server device that provides bidding data to a power bidding platform via a specified communication medium and a battery system that stores power accumulated in the battery by transmitting the power stored in the battery to the power grid or charging the battery using the power supplied from the power grid. The power bidding system includes an input layer for a plurality of parameters required to specify a bidding price, an arithmetic layer that performs arithmetic processing using a linear model or algorithm according to a deep neural network and a model including them, and an output layer that generates bidding information by the arithmetic processing executed by the arithmetic layer, and is characterized by including generation means for generating bidding data that adapts to the constantly changing power demand.

[0051] (2) The generation means is characterized by generating a bidding price and a bidding capacity as bidding data.

[0052] (3) The generation means is characterized by generating the optimal bidding capacity so as to indicate the discharge amount that extends the life of the battery.

[0053] (4) The plurality of parameters include weather forecast, spot price, time-forward price, and battery SOC.

[0054] (5) The plurality of parameters include the capacity of the battery and the environmental information of the battery as other factors.

[0055] (6) The environmental information includes annual temperature fluctuations, wind speed, and wind direction around the battery.

[0056] (7) The generation means is characterized by generating bidding data that adapts to the constantly changing power demand by linear statistical processing.

Explanation of Signs

[0057] 1 Data terminal 3 Cloud server 6 Battery 9 Grid-connected power source

Claims

1. A power bidding system that communicates with a server device that provides bidding data to a power bidding platform via a predetermined communication medium and a battery system that transmits the power stored in the battery to the power grid or charges the battery using the power supplied from the power grid to store power, comprising generating means for generating bidding data that adapts to the constantly changing power demand, the generating means comprising an input layer for a plurality of parameters necessary to specify a bidding price, an arithmetic layer that performs arithmetic processing by a linear model or algorithm according to a deep neural network and a model including them, and an output layer that generates bidding information by the arithmetic processing executed by the arithmetic layer.

2. The power bidding system according to claim 1, wherein the generating means generates a bidding price and a bidding capacity as bidding data.

3. The power bidding system according to claim 2, wherein the generating means generates the optimal bidding capacity so as to indicate a discharge amount that extends the life of the battery.

4. The power bidding system according to claim 1, wherein the plurality of parameters include weather forecast, spot price, forward price, and battery SOC.

5. The power bidding system according to claim 1, wherein the plurality of parameters include the capacity of the battery and the environmental information of the battery as other factors.

6. The power bidding system according to claim 5, wherein the environmental information includes annual temperature fluctuations, wind speed, and wind direction around the battery.

7. The power bidding system according to claim 1, wherein the generating means generates bidding data that adapts to the constantly changing power demand by linear statistical processing.

Images