Dynamic nonlinear optimization of battery energy storage systems

Abstract

The invention is entitled "Dynamic Nonlinear Optimization of Battery Energy Storage Systems." The invention provides a system and method for optimizing a battery energy storage system (BESS) that can involve: inputting data to an echo state network, the data relating to operation of a battery energy storage system and including one or more of: load data, renewable energy data, non-renewable energy data, carbon emissions, carbon credits, and weighted energy costs; and optimizing the operation of the battery energy storage system based on the data input to and processed by the echo state network and relating to the operation of the battery energy storage system.

Description

Technical Field

[0001] The implementation plan relates to the field of energy storage systems, including battery energy storage systems. It also relates to methods and systems for optimizing battery energy storage systems (BESS). Furthermore, it relates to methods and systems for dynamic nonlinear optimization of BESS operation, taking into account carbon emissions, credits, and weighted energy costs. Background Technology

[0002] Deploying renewable and distributed generation assets helps minimize carbon emissions associated with fossil fuel-based power plants. Due to the intermittent and variable nature of renewable energy generation, it is impossible to completely eliminate fossil fuel-based generators. However, if carbon reduction and electricity availability are considered as two distinct objectives to be optimized (as in a multi-objective optimization system), carbon reduction can sometimes directly conflict with the requirement to maximize electricity availability. This is a fundamental problem in optimization algorithms that attempt to maximize energy availability and minimize carbon emissions.

[0003] Attempts have been made to optimize carbon reduction as a separate objective from economic optimization. However, treating carbon reduction as a separate objective remains an unresolved issue because it is unclear who bears the costs of this objective. This objective could run in the opposite direction to economic optimization, potentially causing economic losses to the entities bearing the costs, and in some cases, producers might not have the opportunity to recover these costs from consumers.

[0004] Carbon credits are designed as a market-oriented mechanism for reducing greenhouse gas emissions, allowing the cost of carbon footprint reduction to be quantified and traded as a commodity with associated monetary value. A carbon footprint represents the estimated emissions of carbon dioxide (CO2) and other greenhouse gases (GHG) associated with electricity generation. Generating electricity using renewable energy sources allows power generators to obtain carbon credits after discounting the carbon footprint of products such as solar photovoltaic panels and lithium-ion batteries, which may generate a smaller carbon footprint during their manufacturing process.

[0005] For example, a carbon footprint can be expressed as tons of carbon dioxide or tons of carbon emitted annually. Currently, many versions of carbon calculators are available for calculating carbon footprints. Some examples of third-party vendors specializing in carbon footprint calculations include Environmental Resources Trust, Inc., Blue Source, LLC, and others. On the other hand, carbon credits (e.g., measured in tons of CO2) may represent actions taken to reduce or avoid GHG emissions in one location to offset GHG emissions occurring elsewhere.

[0006] Battery storage systems are becoming increasingly cheaper as lithium costs decrease and new alternatives to sodium and iron emerge. Microgrids or grid-connected energy forms are rapidly emerging, including those using battery storage systems and devices alongside renewable energy sources. Energy costs and availability at any point of operation can be highly dynamic due to the intermittency of renewable energy, load variations, and the dynamic pricing structure of energy with time of day (TOD), as well as demand charges that can be applied during dynamic (e.g., time-varying) peaks in energy demand.

[0007] Conventional battery systems operate on a "timing" principle. That is, for example, the battery can be charged when solar or wind power reaches its capacity, and discharged when solar or wind power is unavailable. This simple switching may not be optimal, because energy at any given time can be subject to strange supply and demand equations.

[0008] Solutions have been proposed that maximize energy obtained from a source based on its availability. Unfortunately, energy costs vary significantly at different points in time, and simply maximizing availability may lead to optimal energy costs. In other words, in many cases, maximum energy output may not represent the consumer's lowest energy bill. Similarly, maximum energy output may not represent the energy service provider's maximum revenue.

[0009] Carbon credits earned from the sale of green electricity can be considered as revenue, and carbon emissions from the operation of diesel generators can be considered as costs within the economic optimization objective, rather than treating carbon reduction as a separate optimization objective or a multivariate optimization objective. Energy management systems (EMS) can have counters that can accumulate carbon credits over a given period of time.

[0010] EMS can also have a counter for carbon emissions that can accumulate over the same period. During the counter's operating window, net carbon credits can be calculated by subtracting the carbon credits earned from the microgrid's emissions. EMS can connect to an external carbon trading platform and can convert accumulated net carbon credits into monetary value, which can be determined based on carbon pricing signals from the carbon trading platform.

[0011] When battery energy storage systems are used in multiple use cases, value or revenue can be stacked. Each use case has certain revenue and carbon credit potential. Furthermore, each use case has certain associated costs, which may include carbon emission costs. An optimal solution is needed that maximizes revenue, including carbon credits, and minimizes costs, including consumer emission costs. Summary of the Invention

[0012] The following description of the invention is provided to facilitate understanding of some features of the embodiments and is not intended to be a complete description.

[0013] Therefore, one aspect of the implementation plan is to provide an improved energy storage system.

[0014] Another aspect of the implementation plan is to provide an improved battery energy storage system (BESS).

[0015] Another aspect of the implementation plan is to provide methods and systems for optimizing BESS.

[0016] Another aspect of the implementation plan is to provide methods and systems for dynamic nonlinear optimization of BESS operations while taking into account carbon emissions, carbon credits and weighted energy costs.

[0017] The above aspects and other objectives can now be achieved as described herein. In one embodiment, a method for optimizing a battery energy storage system may involve: inputting data into an echo state network, the data being relevant to the operation of the battery energy storage system and including at least: load data, renewable energy data, non-renewable energy data, carbon emissions, carbon credits, and weighted energy costs; and optimizing the operation of the battery energy storage system based on the data input into and processed by the echo state network and relevant to the operation of the battery energy storage system.

[0018] In one embodiment of the method, the echo state network may include a neural network.

[0019] One implementation of the method may further involve controlling the battery energy storage system after optimizing the operation of the battery energy storage system based on data input to and processed by the echo state network.

[0020] In one implementation of this method, optimizing the operation of the battery energy storage system based on data input to and processed by the echo state network may further involve generating a day-ahead schedule for the operation of the battery energy storage system based on data output from an optimization result analyzer and a day-ahead schedule output from the echo state network. This optimization result analyzer enables a day-ahead schedule selector to choose the best or optimal day-ahead schedule from those generated by a mixed-integer nonlinear solver. The echo state network integrates prediction and optimization functions. That is, the mixed-integer nonlinear solver can also generate the day-ahead schedule.

[0021] In one implementation of the method, at least a portion of the data may undergo a fast Fourier transform before being input into the echo state network.

[0022] In one implementation of this method, the output of the Fast Fourier Transform can be input into an echo state network.

[0023] In one implementation of the method, optimizing the operation of the battery energy storage system based on data input to and processed by the echo state network may further involve dynamic nonlinear optimization of the operation of the battery energy storage system.

[0024] In one implementation, a system may be provided, which may include: an echo state network, wherein data is input to the echo state network relating to the operation of a battery energy storage system and including at least: load data, renewable energy data, non-renewable energy data, carbon emissions, carbon credits, and weighted energy costs; and a battery energy storage system wherein the operation of the battery energy storage system may be optimized based on data input to and processed by the echo state network and relating to the operation of the battery energy storage system.

[0025] In one implementation of the system, the echo state network may include a neural network.

[0026] One implementation of the system may further include a controller for controlling the battery energy storage system after optimizing the operation of the battery energy storage system based on data input to and processed by the echo state network.

[0027] One implementation of the system may further include generating a day-ahead schedule for the operation of the battery energy storage system based on data output from an optimization results analyzer and a day-ahead schedule output from an echo state network. The optimization results analyzer enables a day-ahead schedule selector to choose from the best or optimal day-ahead schedule generated by a mixed-integer nonlinear solver, which integrates prediction and optimization functions. As previously indicated, the mixed-integer nonlinear solver may also generate the day-ahead schedule.

[0028] In one implementation of the system, at least a portion of the data may undergo a Fast Fourier Transform (FFT) before being input into the echo state network, and the output of the FFT may be input into the echo state network.

[0029] In one implementation of the system, optimizing the operation of the battery energy storage system based on data input to and processed by the echo state network can further involve dynamic nonlinear optimization of the operation of the battery energy storage system.

[0030] In another embodiment, a system for dynamic nonlinear optimization of BESS operations may include: a mixed-integer nonlinear solver that generates day-ahead timetables; a day-ahead timetable generator based on an echo state network model that integrates prediction and optimization functions to minimize the impact of uncertainties or errors that may affect the optimization results of the mixed-integer nonlinear solver; an optimization result analyzer that compares the results of two different solvers, wherein at least one of the two different solvers may include the mixed-integer nonlinear solver; and a training dataset generator that generates data for dynamically retraining the echo state network model when the optimization results are not as optimal as those of the mixed-integer nonlinear solver.

[0031] One implementation of a system for dynamic nonlinear optimization of BESS operations may further include a day-ahead schedule selector that selects the optimal output from two different solvers.

[0032] One implementation of a system for dynamic nonlinear optimization of BESS operations may further include a time series forecast generator that can use nonlinear regression analysis relative to the input data.

[0033] In one implementation of a system for dynamic nonlinear optimization of BESS operations, at least a portion of the input data may undergo a Fast Fourier Transform (FFT), and the output of the FFT may be input into an echo state network model.

[0034] One implementation of a system for dynamic nonlinear optimization of BESS operation may further include a history module that stores past input data.

[0035] In one implementation of a system for dynamic nonlinear optimization of BESS operation, the echo state neural network model may further include a recurrent neural network. Attached Figure Description

[0036] The accompanying drawings also illustrate the invention and, together with the specific embodiments thereof, serve to explain the principles of the invention, wherein similar reference numerals throughout the separate views refer to the same or functionally similar elements and are incorporated in and form part of the specification.

[0037] Figure 1 A block diagram of an energy system that can be implemented according to one or more implementation schemes is shown;

[0038] Figure 2 The architecture of an energy management system that can interact with an energy and carbon trading platform is shown according to one implementation scheme;

[0039] Figure 3 A block diagram of a dynamic nonlinear optimization solver that can be implemented according to one implementation scheme is shown;

[0040] Figure 4 A flowchart depicting the logical operational steps of a method for solving hybrid nonlinear optimizations according to one implementation scheme is shown;

[0041] Figure 5 A schematic diagram of a computer system according to one embodiment is shown; and

[0042] Figure 6 A schematic diagram of a software system including modules, an operating system, and a user interface according to one implementation scheme is shown. Detailed Implementation

[0043] The specific values ​​and configurations discussed in these non-restrictive examples are variable and are cited only to illustrate one or more implementations, and are not intended to limit their scope.

[0044] The subject matter will now be described more fully below with reference to the accompanying drawings, which form part of the subject matter and illustrate specific exemplary embodiments by way of illustration. However, the subject matter can be embodied in many different forms, and therefore the subject matter covered or claimed is intended to be construed as not being limited to any of the exemplary embodiments listed herein; exemplary embodiments are provided merely for illustrative purposes. Likewise, the subject matter intended to be claimed or covered has a suitably broad scope. Among other things, the subject matter can be embodied as a method, apparatus, component, or system. Therefore, embodiments can take the form, for example, hardware, software, firmware, or combinations thereof. Thus, the following detailed description is not intended to be construed as limiting.

[0045] Throughout the specification and claims, terms may have nuanced meanings as the context dictates or implies, in addition to their expressly stated meanings. Similarly, phrases such as “in one embodiment” or “in an exemplary embodiment” and their variations, as used herein, may not necessarily refer to the same embodiment, and phrases such as “in another embodiment” or “in another exemplary embodiment” and their variations, as used herein, may or may not refer to different embodiments. For example, the claimed subject matter is intended to include, in whole or in part, combinations of exemplary embodiments.

[0046] Generally, terms can be understood at least in part from their usage in the context. For example, terms such as “and,” “or,” or “and / or” as used herein can have a variety of meanings that can depend at least in part on the context in which such terms are used. Generally, “or,” when used in an associative list, such as A, B, or C, is intended to indicate A, B, and C used herein in an inclusive sense, and A, B, or C used herein in an exclusive sense. Furthermore, the term “one or more,” as used herein, depends at least in part on the context and can be used to describe any feature, structure, or characteristic in a singular sense, or to describe a combination of features, structures, or characteristics in a plural sense. Similarly, terms such as “a,” “an,” or “the” also depend at least in part on the context and can be understood to convey a singular usage or to express a plural usage. Moreover, the term “based on” can be understood not necessarily to convey a set of exclusive factors, but can depend at least in part on the context, allowing for additional factors that are not necessarily explicitly described again.

[0047] Figure 1 A block diagram of an energy system 100 that can be implemented according to one or more embodiments is shown. Figure 1 The energy system 100 depicted may include a power plant 88 (i.e., centralized power generation) operatively connected to a high-voltage (HV) transmission network 114. The energy system 100 may also communicate with a wide-area communication network 112. The HV transmission network 114 and the wide-area communication network 112 may be implemented as two separate networks, to which the energy system 100 may be operatively connected. The HV transmission network 114 may also communicate with one or more microgrid sites, including, but not limited to, a first microgrid site 116, a second microgrid site 118, and a third microgrid site 120.

[0048] It should be noted that the term "microgrid site" is used interchangeably with the term "microgrid" in this document to refer to the same characteristic. The term "microgrid" can refer to a localized grouping of generation, energy storage, and loads that can be operatively connected to a conventional centralized power grid (e.g., a distribution network or macro grid) via a common point of coupling. A microgrid can be controlled by a controller, which can be centralized or distributed (e.g., controlling distributed energy resources according to voltage or current control schemes).

[0049] The first microgrid site 116 may include a photovoltaic (PV) system 117 capable of communicating with a local area network (LAN), and a distribution transformer 123 operable to the HV transmission network 114. The first microgrid site 116 may further include a battery energy storage system 119 capable of communicating with the distribution transformer 123, and a remote control thin client 125 operable to both the distribution transformer 123 and the LAN 119. Load “LOAD” data can also be accessed via the LAN, as illustrated in the schematic diagram of the first microgrid site 116.

[0050] Battery energy storage is a technology that enables power system operators and utilities to store energy for later use. A BESS is an electrochemical device that can charge (or collect) energy from the grid, microgrid, or power plant and then release that energy later to provide electricity or other grid services when needed.

[0051] The second microgrid site 118 may include a remote control thin client 131 that can communicate with a LAN 129, which in turn can communicate with a BESS 135 and a diesel generator 127. The second microgrid site 118 may also include a distribution transformer 133 that can communicate with the remote control thin client 131.

[0052] The third microgrid site 120 may include a remote control thin client 139, a LAN 141, a distribution transformer 143, and a BESS 137. Additionally, the third microgrid site 120 may include one or more wind turbines 145 (e.g., wind energy converters) that convert the kinetic energy of wind into electrical energy. The turbines 145 may communicate with the distribution transformer 143. Load data can be accessed via the LAN 141. Distributed generation may occur at the first microgrid site 116, the second microgrid site 118, and the third microgrid site 120, which can be considered "behind" the respective distribution transformers 123, 133, and 143.

[0053] Wide area communication network 112 can also communicate with remote control thin clients 125, 131, and 139, as well as with a VPN / firewall 110 associated with network 102, which may include a Supervisory Control and Data Acquisition (SCADA) module 104 (e.g., via the cloud (Experion Elevate SCADA)) and a process control (PCOC) 106 via the cloud. PCOC 106 may also provide a Control as a Service (CaaS) application 108. It should be noted that the term SCADA refers to “Supervisory Control and Data Acquisition” and relates to a control system architecture that includes computers, networked data communications, and a graphical user interface (GUI) for advanced process supervision and management, as well as other peripherals such as programmable logic controllers (PLCs) and discrete proportional-integral-derivative (PID) controllers that can interact with process plants or machinery. SCADA modules and systems can also be used for the management and operation of project-driven processes. It should be noted that, as used herein, the term “supervisory controller” can also refer to a SCADA module, such as SCADA module 104.

[0054] Wide area communication network 112 may further communicate with VPN / firewall 98, which in turn communicates with energy center remote operation center 96, which may include thin client desktop computer 94, widescreen monitor 92 and server-class computer 90.

[0055] Including BESS (for example, Figure 1 The term "value stacking" in energy systems (such as Energy System 100) described in BESS 119, BESS 135, and BESS 137 can refer to utilizing batteries to support multiple use cases. Examples of use cases may include peak shaving, off-peak operation, backup power and frequency, active and reactive power support, black start, etc. Each use case may include revenue opportunities and associated costs. The goal of value stacking optimization is to maximize the value of distributed generation assets coupled with energy storage systems by maximizing revenue and / or minimizing the costs associated with each use case.

[0056] Conventional approaches treat this problem as a multi-objective optimization problem with multiple objectives or use cases, attempting to maximize availability. The value-added optimization described in the implementation plan may consider revenue associated with each use case, which may include sales revenue from energy and carbon credits obtained from distributed generation assets, as well as costs associated with energy purchased from the grid, energy generated by burning diesel or other fuels, operating costs (which may include maintenance costs of generation assets), and costs associated with battery cycle counts and carbon emission costs (including the carbon footprint of renewable generation assets).

[0057] Some non-renewable energy sources (such as coal) are cheaper than other non-renewable fuels (such as diesel), but may have a higher associated carbon cost. The true cost of such non-renewable fuels will be factored in when considering the cost of carbon footprint and CO2 equivalent greenhouse gas emissions. Previous approaches treated carbon footprint reduction as a separate optimization objective. The problem with treating carbon footprint reduction as a separate optimization objective is that it may directly conflict with energy availability requirements. Treating energy availability as a constraint rather than an optimization objective allows for achieving the lowest cost and maximum revenue from every use case.

[0058] This implementation optimizes, rather than minimizes, expectations of weighted energy costs, which can include expectations of consumer energy bills (e.g., purchases from utilities), operating costs associated with distributed energy generation, the cost of battery charging cycles, carbon credits associated with renewable energy generation, and carbon costs associated with fossil fuel-based power plants. In this type of energy bill calculation, all operating costs, carbon costs, and credits are included in addition to the base electricity price for grid and fuel costs. This all-inclusive calculation allows for a comprehensive picture and the final optimization factors.

[0059] Due to the time-coupled effect of energy storage systems on the entire energy system, the optimization problem of the charging and discharging cycles of energy storage systems differs from combinatorial optimization problems, which address the issue of putting generators into operation at any given time. The current state of the system (including the battery's state of charge, current supply and demand, and consequently, electricity prices) is a function of past decisions, which influence future states and even future decisions. It is practically impossible to go back in time and optimize the past. Therefore, optimization problems may need to consider anticipated energy costs, demand, generation, and electricity prices. Alternative approaches to handling time-coupled effects in optimization (such as dynamic programming) are plagued by the "curse of dimensionality," a term used by mathematician Richard Bellman in his book on dynamic programming.

[0060] This concept can be illustrated in the following context. Consider a microgrid, which includes the following:

[0061] • Battery energy storage systems with capacities of 235 kWh and 125 kW

[0062] · 125KW solar power plant

[0063] ·680KW basic load

[0064] · 175KW fixed load

[0065] ·505KW interruptible load

[0066] Variable or peak loads from 0 kW to 217.5 kW

[0067] 175KW diesel generator set

[0068] A 120kW wind power plant.

[0069] Solar power plants can charge batteries, which in turn can alleviate peak demand during peak periods of 1800 to 2000 hours. Figure 1 In the exemplary energy system 100 shown, a solar power plant can be represented by photovoltaics 117. Consider a four-day span, where peak demand is 217.5 kWh, 195 kWh, 127.5 kWh, and 127.5 kWh. This profile can be known a priori from short-term load forecasting.

[0070] Figure 1 BESS operations, represented by BESS 119, BESS 135, and BESS 137, are performed by charging the batteries from PV 117 in the morning and discharging them during peak hours. The cost per charge / discharge cycle is Rs. 1470, derived from the battery cost of Rs. 2058600 and the battery life of Rs. 1400. This cycle cost may also be included in the energy bill. Other operating costs exist, but for simplicity, these may not be included in the calculation.

[0071] You can specify upper and lower limits for the battery's state of charge to avoid three conditions that could degrade battery performance and shorten its lifespan. These three conditions are as follows:

[0072] 1. Overcharging

[0073] 2. Deep discharge

[0074] 3. Idle battery under low charge state

[0075] The upper limit is 95% of the battery capacity, for example, 223.25 kWh. The lower limit is 20% of the battery capacity, for example, 47 kWh. During the peak energy consumption half-hour, utilities may implement load shedding measures, disconnecting the entire microgrid's load. During this half-hour period, interruptible loads may be shut down. The remaining fixed loads and variable peak loads can be supported using BESS generation or diesel generation within the microgrid.

[0076] Three different battery charging and discharging methods can be considered, each with different associated costs, which demonstrates that the total weighted energy cost can include carbon costs and carbon credits.

[0077] Method 1 :

[0078] Diesel generators can be deployed for half an hour during utility load shedding or demand response load reduction. BESS can be deployed during peak hours to support variable peak loads, except for half an hour during which utility load shedding may occur. Exemplary microgrid operations are shown in Table 1 below.

[0079]

[0080] Table 1

[0081] Method 2:

[0082] In this approach, the battery can commit to supporting critical loads and some peak loads during a half-hour period of load shedding at the utility. This eliminates the need for a diesel generator in the backup power use case promised in Method 1. This also results in overall cost savings, as shown in Table 2 below.

[0083]

[0084] Table 2

[0085] This cost saving can be achieved without any additional investment in batteries for any other power generation assets. Conversely, the use of diesel generators may be eliminated entirely, potentially leading to additional capping and maintenance cost savings.

[0086] Method 3:

[0087] A key difference between Days 3 and 4 and Days 1 and 2 is that the fixed load can be reduced from 175 kW to 75 kW, while the interruptible load can be increased from 505 kW to 605 kW. Since the interruptible load can be reduced during demand response events or utility outages, the load on the backup power supply under BESS can be decreased, allowing for further optimization. Such optimization opportunities were not considered in Method 2.

[0088]

[0089] Table 3

[0090] Since the peak demand on day 4 was 75 kW, and the remaining energy in the battery at the start of day 4 was 103.25 kWh, charging was skipped on day 4, and the remaining energy was used to counteract the peak. During the outage event, the grid might have consumed an additional (75 – 103.25 + 47) kWh of energy. However, this avoids the charging cycle and saves Rs. 1470 in this project. This results in a net cost saving of Rs. 2357, or 1.78% of the day's energy bill, as shown in Table 4 below.

[0091]

[0092] Table 4

[0093] As can be seen from the foregoing, cost-weighted energy optimization can provide financial benefits, while energy maximization cannot. Supervisory control of such optimization algorithms can collect operational parameters and cost profiles. Based on this, downloading and calculating switching details, such as sending commands to an EMS (Energy Management System) to implement the switch, can bring economic benefits. It should be noted that an example of such an EMS is... Figure 2 EMS 140 is shown in the image.

[0094] Some solutions may minimize the energy consumed by users from the grid and may not be optimized using cost factors. However, the method disclosed in this invention differs from these conventional solutions, which makes the solution disclosed in this invention highly attractive.

[0095] Cost-weighted optimization can be performed using a supervisory controller such as SCADA module 104 to issue EMS commands based on this method (e.g., an algorithm). Instead of net energy, energy costs (or utility unit prices) can be calculated and optimized based on electricity prices. More compassionate operating costs may be part of the calculation. For example, battery life can be defined as the number of charging cycles. In this calculation, the monetary equivalent of a charging cycle can be calculated. Carbon credits generated from renewable energy generation can also be considered as income, while carbon emissions from operating fossil fuel-based generators (such as diesel generators) and the carbon footprint of renewable energy power plants can be considered as costs in this calculation, as follows:

[0096] E{cost wt (Δt)}=E{P 电网 ×D 费用}+E{P 电网 ×Δt}×E{C 电价 _ 使用时间}+E{C 储能}+E{C 循

[0097] 环}+E{C 备用}-E{ARB 收入}+E{CO2 e_成本}-E{CO2 e信用}

[0098] ·E{}: Expectation function; represents the expected value of the weighted energy cost, and the dependent variable as defined below.

[0099] ·P 电网Electricity drawn from the power grid

[0100] ·D 费用 Cost of electricity drawn during peak periods

[0101] • Δt: The time interval for completing energy cost calculations. Typically 15 minutes.

[0102] ·C 电价 _ 使用时间 Energy price based on usage time

[0103] ·C 储能 Energy storage costs, including round-trip efficiency and battery charging costs.

[0104] ·C 循环 Costs associated with battery cycles (battery cost / battery cycle life)

[0105] ·C 备用 Backup power cost

[0106] ·ARB 收入 Energy arbitrage income

[0107] ·cost wt (Δt) is the weighted cost, which considers costs associated with electricity demand, time-of-use energy costs, battery storage costs, battery charging cycle costs, and backup or auxiliary generation costs to avoid grid demand charges and also considers arbitrage revenue. The optimization objective is to minimize costs. wt Minimize (Δt) and power availability, or minimize P 电网 +P 储能 +P 备用 -P 负荷 >=0 as a constraint variable

[0108] CO2 e信用 : Accumulated carbon credits obtained through generating electricity from renewable power plants.

[0109] CO2 e_成本 Carbon costs are the emissions costs associated with greenhouse gas emissions from microgrids resulting from the use of non-renewable power plants (e.g., diesel generators) as backup power. Carbon costs also include the carbon costs of renewable power generation assets such as solar panels.

[0110] Optimization algorithms can be applied to predicted power generation (from all energy sources) and demand (from all loads) for at least 24 hours. For example, the result of an optimization algorithm could be a schedule for battery charging, discharging, and idling. Decisions about when to charge or not to charge, and when to discharge or not to discharge, can be based on optimization calculations performed on predicted power generation and demand for at least 24 hours. If the forecasts change during the day, optimization calculations can be performed on the new short-term and long-term power generation and load forecasts, and new schedules can be derived from these calculations.

[0111] Figure 2 An EMS140 capable of interacting with an energy and carbon trading platform, according to one implementation scheme, is shown. Note that... Figure 2 The various blocks and components shown can be implemented as modules and / or data. Figure 2 The EMS 140 depicted herein may include a blockchain decentralized ledger architecture 140, which may include or communicate with a carbon trading platform 144 capable of generating carbon price signals 146. The blockchain decentralized ledger architecture 140 may also communicate with an energy trading platform 152 capable of providing or generating energy price signals 154. It should be noted that the term "blockchain" as used herein may refer to a distributed ledger, similar to a distributed database, which may include a continuously growing list of records (called "blocks") linked and secured using cryptography, wherein each block may contain a hash pointer (i.e., a data structure that provides a pointer to a location where certain information is stored) as a link to the previous block, timestamp, and data. Private blockchains (PB) can be used as blockchains operated by an organization within a closed (non-public) communication network.

[0112] Carbon price signal 146 and energy price signal 154 can be fed to optimizer 148, which includes value-revenue overlay module 150. Optimizer 148 can be configured to operate as an echo state network (ESN) dynamic nonlinear optimizer. Optimizer 148 can then generate net carbon credits 156 and weighted energy costs 158. Net carbon credits 156 can be calculated from carbon counters, including total carbon credit counter 162 and total carbon emissions counter 160. Note that the term "carbon counter" as used herein can refer to a counter that provides an indication of carbon generated or delayed by energy.

[0113] Weighted average energy cost 158 ​​can be calculated from battery cycle counter 172, diesel fuel consumption module 164, and electricity meter 166 (net / bidirectional). Total carbon emission counter 160 can communicate with battery carbon footprint counter 168, diesel carbon emission counter 176, and solar PV carbon footprint counter 180. Total carbon credit counter 162 can communicate with wind carbon credit counter 182 and solar PV carbon credit counter 184. Furthermore, total carbon credit counter 162 can communicate with battery carbon credit counter 170. Battery carbon footprint counter 168 and battery carbon credit counter 170 can communicate with battery cycle counter 172, which in turn can communicate with BESS 174. Note that examples of systems that can be used to implement BESS 174 include... Figure 1 The BESS 119, BESS 135 and BESS 137 are shown.

[0114] The diesel carbon emission counter 176 can further communicate with the diesel generator 178, which in turn can communicate with the diesel fuel consumption module 164. That is, the diesel generator 178 can provide data to the diesel fuel consumption module 164. The solar PV carbon footprint counter 180, the wind energy carbon credit counter 182, and the solar PV carbon credit counter 184 can communicate with the renewable power plant 186 (e.g., solar PV, wind, etc.).

[0115] Therefore, EMS 140 can be based on a distributed and decentralized ledger architecture 142, which can interact with carbon trading platform 144 and energy trading platform 152. Carbon trading platform 144 can send “real-time” carbon price signals 145 to optimizer 148. Energy trading platform 152 can also send energy price signals 154 to optimizer 148. Due to the variable nature of distributed generation and load (including battery energy systems such as BESS 174), prices can be expected to be dynamic when demand and supply change rapidly.

[0116] EMS 140 maintains carbon footprint counters for batteries and solar panels because these assets have one-time carbon costs associated with the manufacturing process of these assets. A battery's carbon footprint may have a fixed component associated with battery life (in years) and a variable component associated with battery cycle life. If a battery is cycled more frequently, its cycle life will decrease more rapidly. Therefore, the carbon costs associated with battery cycling increase as batteries are used more frequently. This is one way to account for frequent cycling, which can degrade battery performance more quickly, potentially requiring battery replacement or recycling, thus increasing the carbon footprint of battery manufacturing and production. Similarly, carbon costs associated with the manufacturing of solar cells may be factored into the use of solar panels.

[0117] The diesel generator 178 can also be maintained with a diesel carbon emission counter 176 to count emissions generated from its operation. For example, the carbon emissions from the diesel generator could be approximately 2.68 kg per liter of diesel fuel consumed. The carbon footprint associated with electricity purchased from the main grid can also be considered. If grid electricity is generated by a coal-fired power plant, the emissions cost associated with grid electricity can be considered, for example, 1 ton of CO2 per MWh of grid energy consumed. 2e .

[0118] EMS 140 also maintains carbon credit counters for renewable energy generation assets, as well as BESS 174 for storing renewable energy and dispatching stored energy when no renewable energy generation is possible. Net carbon credits can be based on the carbon credits at sites (e.g., such as...). Figure 1 Carbon credits are calculated by discounting carbon emissions at the first microgrid site 116, the second microgrid site 118, and the third microgrid site 120 shown in the figure.

[0119] Net carbon credits can be traded through carbon swaps and may have an associated monetary value. This monetary value may vary by market or over time. Carbon prices are expected to rise as "low-hanging" carbon reduction projects are completed and new initiatives require more investment than older ones.

[0120] Figure 3 A block diagram of a dynamic nonlinear optimization solver 190, which can be implemented according to one embodiment, is shown. The dynamic nonlinear optimization solver 190 may include multiple modules, including, for example, a module 192 that implements an echo state network (ESN) learning algorithm. Module 192 receives data generated from a day-ahead schedule selector module 200, which may include a training dataset generator 202. Module 192 may provide data to an echo state network 194, which in turn may receive data from an FFT (Fast Fourier Transform) module 195. The echo state network 194 may be implemented as a type of reservoir computer using a recurrent neural network with sparsely and operationally connected hidden layers (e.g., typically with 1% connectivity). The connectivity and weights of the hidden neurons may be fixed and randomly assigned. Furthermore, the echo state network 194 may implement an ESN dynamic nonlinear optimizer model as part of the overall dynamic nonlinear optimization solver 190.

[0121] The dynamic nonlinear optimization solver 190 may further include a history module 196 that provides data to an optimization result analyzer 198, which in turn provides data to a day-ahead schedule selector module 200. Furthermore, the dynamic nonlinear optimization solver 190 may include a module 204 capable of performing nonlinear regression analysis and time series forecasting. Data from module 204 may be provided to module 206, which can be used as a mixed-integer nonlinear programming optimization solver. Data from module 206 may be provided to the optimization result analyzer 198, or to the day-ahead schedule selector module 200, which may then generate data, as indicated by arrow 205, indicating, for example, day-ahead schedules (i.e., for generator combination, battery charging, discharging, idling, and load shedding).

[0122] The dynamic, intermittent, and unpredictable nature of renewable energy generation, dynamic loads, and dynamic pricing of energy and carbon may necessitate optimizers (such as dynamic nonlinear optimization solvers190) to find the global minimum of weighted costs, taking into account expected future generation, load, and prices. Combinatorial optimization that does not consider future generation, demand, and prices cannot compute the true global minimum of weighted costs. Conventional solutions might involve forecasting functions for generation, load, and prices, and then using methods such as mixed-integer linear programming or dynamic programming to compute a day-ahead schedule that minimizes the weighted costs of energy and carbon while satisfying a set of defined constraints. Energy availability is one such constraint. Given the nonlinear nature of energy systems, some conventional solutions have attempted to improve optimization by applying mixed-integer nonlinear programming techniques.

[0123] On the other hand, the implementation scheme relates to systems and methods in which a prediction function and a cost-minimizing objective function can be combined and nonlinear programming problems can be solved using an echo-state network 194, which is a type of recurrent neural network with the characteristic of faster training of neural networks. Recurrent neural networks are neural networks with memory, whose characteristics make them suitable for time series prediction of functions with nonlinearity or discontinuity. Generalized recurrent neural networks (RNNs) are difficult to train, but echo-state networks are a type of RNN in which only the outer layer neurons are retrained, while the inner layer can be initialized only once and remains fixed during network retraining.

[0124] Nonlinear programming problems can first be transformed into primal-dual problems. Echo state networks 194 can be constructed based on the primal-dual problem, and their stable states are feasible solutions to the primal-dual problem, making the solutions to the primal-dual problem also solutions to the primal nonlinear programming problem.

[0125] The input to the echo-state network 194 can include not only the time-series data used as input, but also the signal obtained after transforming the signal into the frequency domain by applying an FFT through the FFT module 195. That is, the FFT module 195 at the input of the echo-state network 194 overcomes two fundamental problems existing in neural networks. If the training dataset does not have sufficient variance, the neural network tends to ignore important inputs.

[0126] The presence or absence of variance in the time domain can affect the training of neural networks and the effectiveness of inference after training them on datasets with insufficient variance. Transforming the time-domain input into the frequency domain and processing the input in the frequency domain may be an effective way to address this fundamental problem. A variance-free time-domain signal will have a peak at zero frequency in the frequency domain and zero values ​​at all other frequencies. Similarly, sine waves of different frequencies will have peaks at their respective frequencies, which allows neural networks to distinguish signals with similar amplitudes but different frequencies in the time domain.

[0127] Neural networks tend to have high convergence times when the input has broadband noise. If the noise amplitude is high relative to the sensor or input signal (e.g., low signal-to-noise ratio), the presence of broadband noise can cause the neural network to train indefinitely without convergence. Applying the FFT module 195 to transform the time-domain input signal into the frequency domain effectively filters broadband noise. Unlike conventional filters, the FFT module 195 does not remove any frequency-dependent signals. Instead, the FFT module operates based on the fundamental properties of the Fourier transform, where the spectrum of broadband noise in the time domain is a flat signal in the frequency domain, which is often ignored by neural networks because neural network training algorithms (such as error backpropagation) do not train on zero-sigma input signals. Therefore, hybrid nonlinear optimization solvers such as the dynamic nonlinear optimization solver 190 (which can utilize machine learning techniques and is applicable to optimization functions related to time series prediction as well as mixed-integer nonlinear optimization techniques) have not been previously designed and implemented.

[0128] Figure 4 A flowchart illustrating the logical operation steps of method 220 for hybrid nonlinear optimization solving, according to one embodiment, is shown. Method 220 can be followed to implement this method. Figure 3 The dynamic nonlinear optimization solver 190 is depicted in the diagram. As shown in box 224, the process of method 220 can be initiated. As shown next in box 226, the input data steps or operations can be implemented, including but not limited to data on load, non-renewable power generation, renewable power generation, energy prices, carbon prices, carbon credits, energy sales revenue, battery data, stage of charge (SOC), state of health (SOH), weather, and time of day.

[0129] It should be noted that, as used herein, the term State of Charge (SOC) can be related to the battery's SOC, which serves as a metric to quantify the remaining energy in the battery compared to its energy level when fully charged, and to indicate to the user how long the battery will continue to operate before needing to be recharged. On the other hand, the term State of Health (SOH) can be related to the battery's SOH, and can indicate the degree to which the battery has degraded due to aging.

[0130] As depicted below in box 228, the step or operation of transforming time-domain data into the frequency domain can be performed. Subsequently, as shown in box 230, the time-domain and frequency-domain data can be input to... Figure 3 The echo state network 194 shown is a step or operation for modeling a dynamic nonlinear optimization function. Then, as depicted in box 232, the input / output data can be stored in... Figure 3 The steps or operations shown in the history module 196.

[0131] Following the steps / operations shown in box 232, the step or operation of generating the day-ahead timetable (i.e., the dynamic nonlinear optimizer model) calculated by the echo state network 194 can be implemented as depicted in box 234. Next, as depicted in box 236, the step or operation of calculating the expected value of the total weighted cost of energy and carbon based on the day-ahead timetable generated by the echo state network 194 and the dynamic nonlinear optimization solver 190, using t0 = t_curr, can be implemented. The dynamic nonlinear optimization solver 190 can be used at least partially as a hybrid nonlinear integer programming (MINLP) solver.

[0132] It is important to note that a standalone MINLP solver cannot constitute a dynamic nonlinear optimization solver. That is, the MINLP solver can operate based on time series predictions generated by nonlinear regression analysis, which is prone to uncertainty or errors in the predictions. A dynamic optimization solver can be constructed by combining the echo state network 194, modules 204 and 206, as well as modules 192, FFT module 195, history module 196, optimization result analyzer 198, and day-ahead timetable selector module 200. If the day-ahead timetable generated by the echo state network 194 is less optimal than that generated by 206, the optimization result analyzer 198 allows the training dataset generator 202 to retrain the echo state network 194.

[0133] Then the decision-making operation can be performed as shown in box 238, as follows: E{Cost wt}_ESN>E{Cost wt}_MINLP. As shown in box 239, the step or operation of outputting the day-ahead schedule calculated by the ESN to the energy controller can be implemented. Alternatively, as shown in box 240, the step or operation of outputting the day-ahead schedule calculated by the MINLP solver to the energy controller can be implemented. After the processing of the operation depicted in box 240, the step or operation of generating the training dataset of the echo state network 194 (e.g., the ESN dynamic nonlinear optimizer model) can be implemented, as shown in box 242. Then, as shown in box 244, the ESN dynamic nonlinear optimizer (e.g., Figure 2 The steps or operations of the optimizer 148 shown.

[0134] Next, as depicted in box 246, the step or operation of performing a rule check (t1 = t_curr) can be implemented. This operation may also include using the data generated as a result of the step / operation shown in box 239. After the operation processing shown in box 246, the step or operation of checking for rule violations can be implemented as depicted in box 248. Based on the result of the operation performed as shown in box 248, the step or operation starting with the step / operation depicted in box 226 can be repeated. If not, the operation shown in box 250 can be implemented, where the day-ahead schedule for generator combination, load dispatch / reduction, and battery charging / discharging / idling can be executed in the energy controller. Next, as shown in decision box 252, the step or operation can be implemented as: t_curr = t1 + 15 minutes. Thereafter, as shown in decision box 254, the step or operation can be implemented as: t_curr = t0 + 24 hours.

[0135] Based on the foregoing, it can be understood that when using BESS to support multiple use cases, the optimization parameters can be dynamic, non-linear, and discontinuous. Cost-weighted optimization can be applied because each use case has associated costs and a certain revenue potential. Costs can include energy costs and carbon costs, with carbon costs having a monetary value associated with emissions. Revenue can include associated carbon credits.

[0136] This approach differs from conventional solutions that treat minimizing energy costs and carbon emissions as two separate objectives that sometimes conflict and can lead to suboptimal outcomes for either energy costs or carbon emissions, or both. An optimization objective could be to minimize the total weighted cost of energy and carbon, taking into account energy sales revenue and carbon credits, which may have an associated monetary value.

[0137] The method disclosed in this invention may involve an Energy Management System (EMS) capable of interacting with external energy trading platforms and carbon trading platforms, and capable of receiving energy and carbon pricing signals, which may be dynamic and time-dependent, such as... Figure 2As shown. This EMS maintains a counter for carbon emissions and the carbon credits obtained. The optimization problem can be viewed as a dynamic nonlinear optimization problem, and differential methods can be implemented to model the dynamic and nonlinear properties of the system, such as... Figure 4 As shown. Hybrid models can use recurrent neural networks, which can more quickly train a class of recurrent neural networks called echo-state networks, thereby modeling combined prediction and optimization solvers, as well as first-principles-based mixed-integer nonlinear programming solvers and time series forecasts based on nonlinear regression analysis.

[0138] The systems and methods disclosed in this invention overcome known limitations in conventional systems that use mixed-integer nonlinear solvers, which suffer from performance issues due to uncertainties or errors in the prediction function. Furthermore, the systems and methods disclosed in this invention overcome the limitations of dynamic nonlinear solvers by enabling dynamic retraining of the dynamic nonlinear solver, even when its performance is suboptimal compared to conventional mixed-integer nonlinear solvers. This is based on echo-state networks (a type of recurrent neural network).

[0139] Those skilled in the art will readily recognize that the subject matter disclosed herein can be practiced without one or more specific details or other methods. In other instances, well-known structures or operations have not been shown in detail to avoid obscuring certain aspects. This disclosure is not limited to the order of the actions or events shown, as some actions may occur in a different order and / or simultaneously with other actions or events. Furthermore, not all actions or events shown are necessary to implement the methods according to the embodiments disclosed herein.

[0140] The embodiments are described at least in part by referring to flowchart illustrations, steps, and / or block diagrams of methods, systems, computer program products, data structures, and scripts. It should be understood that each block and combination of blocks in the illustrations can be implemented by computer program instructions. These computer program instructions can be provided to a processor, such as a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus, to produce a machine that executes via the processor of that computer or other programmable data processing apparatus and creates means for implementing the functions / actions specified in one or more blocks.

[0141] For clarity, embodiments may be implemented in the context of, for example, a dedicated computer or a general-purpose computer or other programmable data processing apparatus or system. For instance, in some exemplary embodiments, the data processing apparatus or system may be implemented as a combination of a dedicated computer and a general-purpose computer. In this regard, a system comprising different hardware and software modules and different types of features can be considered as a dedicated computer intended for the purpose of synchronizing controllers in an automated control system, as described herein. However, in general, embodiments may be implemented as methods and / or computer program products at any possible level of technical detail integration. Such computer program products may include one or more computer-readable storage media having computer-readable program instructions thereon for causing a processor to perform aspects of the embodiments, such as the steps, operations, or instructions described herein.

[0142] The aforementioned computer program instructions may also be stored in a computer-readable storage medium that can instruct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions (e.g., steps / operations) stored in the computer-readable storage medium produce an article of writing including instruction means that implement the functions / actions specified in one or more blocks, flowcharts and other architectures shown and described herein.

[0143] The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus, thereby producing a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions / actions specified in one or more boxes herein.

[0144] The flowcharts and block diagrams in the figures can illustrate the architecture, functionality, and operation of possible specific implementations of systems, methods, and computer program products according to various embodiments (e.g., preferred or alternative embodiments). In this regard, each box in a flowchart or block diagram can represent a module, segment, or part of an instruction, which includes one or more executable instructions for implementing the specified logical function.

[0145] In some alternative embodiments, the functions indicated in the boxes may occur in the order shown in the accompanying drawings. For example, two boxes shown consecutively may actually be executed simultaneously, or these boxes may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each box and combination of boxes in the block diagrams and / or flowcharts may be implemented by a system based on the specified purpose hardware that performs the specified function or action or performs a combination of specified purpose hardware and computer instructions.

[0146] The functions described herein can be fully and non-abstractly implemented as physical hardware, fully implemented as physical non-abstract software (including firmware, resident software, microcode, etc.), or combined with non-abstract software and hardware implementations that can be collectively referred to herein as "circuit," "module," "engine," "component," "block," "database," "agent," or "system." Furthermore, aspects of this disclosure can take the form of a computer program product embodied in one or more non-transitory computer-readable media having computer-readable and / or executable program code embodied thereon.

[0147] Figure 5 and Figure 6 The illustration is merely an exemplary diagram of a data processing environment in which exemplary embodiments can be implemented. It should be understood that... Figure 5 and Figure 6 This is merely illustrative and is not intended to assert or imply any limitation on aspects of the implementation schemes or the environment in which the implementation schemes may be implemented. Many modifications may be made to the environment described and illustrated herein without departing from the spirit and scope of the implementation schemes.

[0148] like Figure 5 As shown, some embodiments and / or aspects of these embodiments can be implemented in the context of a data processing apparatus 400, which may include, for example, one or more processors such as processor 341 (e.g., CPU (Central Processing Unit) and / or other microprocessors), memory 342, controller 343, additional memory such as ROM / RAM 332 (i.e., ROM and / or RAM), peripheral USB (Universal Serial Bus) connection 347, keyboard 349 and / or another input device 345 (e.g., pointing device, such as mouse, trackball, pen device, etc.), display 346 (e.g., monitor, touch screen display, etc.) and / or other peripheral connections and components. In some embodiments, the database 114 previously shown and discussed herein may be located together with, for example, memory 342 or another memory.

[0149] System bus 110 serves as the main electronic information highway interconnecting the other illustrated components of the hardware of data processing device 400. In some embodiments, processor 341 may be a CPU, which serves as the central processing unit of data processing device 400, performing calculations and logical operations required to execute programs. Read-only memory (ROM) and random access memory (RAM) of ROM / RAM 344 ​​are examples of non-transitory computer-readable storage media.

[0150] Controller 343 can be coupled to system bus 110 with one or more optional non-transitory computer-readable storage media. These storage media may include, for example, external or internal DVD drives, CD-ROM drives, hard disk drives, flash memory, USB drives, etc. These various drives and controllers may be optional devices. Program instructions, software, or interactive modules for providing an interface and performing any queries or analyses associated with one or more datasets may be stored, for example, in ROM / RAM 344. Optionally, the program instructions may be stored on tangible non-transitory computer-readable media, such as optical discs, digital disks, flash memory, memory cards, USB drives, optical disc storage media, and / or other recording media.

[0151] As shown in the figure, various components of the data processing apparatus 400 can communicate electronically via a system bus 351 or a similar architecture. The system bus 351 may be, for example, a subsystem that transmits data between, for example, computer components within the data processing apparatus 400, to, and from other data processing devices, components, computers, etc. In some embodiments, the data processing apparatus 400 may be implemented as a server, for example, in a client-server based network (e.g., the Internet), or in a client-server context (i.e., where aspects are practiced on both the client and server).

[0152] In some implementations, the data processing device 400 may be, for example, a standalone desktop computer, laptop computer, smartphone, tablet computing device, etc., wherein each such device is operatively connected to and / or communicates with a client-server based network or other types of networks (e.g., cellular networks, Wi-Fi, etc.). Examples of such networks may also include Figure 1 The wide area communication network 112 shown is an example. Other examples of the data processing device 400 may include, for example... Figure 1 The thin client desktop computer 94 and server-class computer 90 are shown. An example of monitor 346 is... Figure 1 The widescreen monitor 92 depicted in the image.

[0153] Figure 6 Showing the boot Figure 5The data processing apparatus 400 shown is operated by a computer software system 450. Software application 454 may be stored in, for example, memory 342 and / or another memory, and may include one or more modules, such as module 452. Computer software system 450 may also include a kernel or operating system 451 and a shell or interface 453. One or more applications, such as software application 454, may be "loaded" (i.e., transferred from, for example, mass storage or another memory location, to memory 342) for execution by data processing apparatus 400. Data processing apparatus 400 may receive user commands and data via interface 453. These inputs may then be executed by data processing apparatus 400 according to instructions (e.g., steps or operations) from operating system 451 and / or software application 454. In some embodiments, interface 453 may be used to display results, at which point user 459 may provide additional input or terminate the session. Software application 454 may include module 452, which may, for example, implement steps, instructions, operations, and scripts, such as those discussed and illustrated with respect to various diagrams, boxes, and components.

[0154] The following discussion aims to provide a brief general description of a suitable computing environment in which the systems and methods described herein may be implemented. Although not required, the implementation will be described in the general context of computer-executable instructions (such as program modules) executed by a single computer. In most cases, a “module” (also called an “engine”) can constitute a software application, but it can also be implemented as both software and hardware (i.e., a combination of software and hardware).

[0155] Typically, program modules include, but are not limited to, routines, subroutines, software applications, programs, objects, components, data structures, etc., that perform specific tasks or implement specific data types and instructions. Furthermore, those skilled in the art will understand that the disclosed methods and systems can be practiced using other computer system configurations, such as, for example, handheld devices, multiprocessor systems, data networks, microprocessor-based or programmable consumer electronics, networked PCs, microcomputers, mainframe computers, servers, etc.

[0156] It should be noted that, as used herein, the term "module" can further refer to a collection of routines and data structures that perform a specific task or implement a specific data type. A module can consist of two parts: an interface, which lists the constants, data types, variables, and routines that can be accessed by other modules or routines; and a concrete implementation, which is typically private (accessible only by that module) and includes the source code that actually implements the routines in that module. The term "module" can also simply refer to an application program, such as a computer program designed to help perform a specific task.

[0157] In some exemplary embodiments, the term "module" may also refer to a modular hardware component or a component as a combination of hardware and software. It should be understood that specific implementations and processing of the modules disclosed herein (whether primarily software-based modules and / or hardware-based modules or combinations thereof, as described herein) can lead to improved processing speed and ultimately energy savings and efficiency in devices such as the BESS described herein.

[0158] Implementations may also constitute improvements to technical systems (e.g., BESS such as those disclosed herein), rather than simply using computer systems as tools. The modules, instructions, steps, and functions disclosed herein can lead to specific improvements on existing systems, resulting in improved data processing systems.

[0159] Figure 5 and Figure 6 This is intended as an example and not as an architectural limitation of the implementation. Furthermore, such implementations are not limited to any particular application or computing or data processing environment. Rather, those skilled in the art will understand that the disclosed methods can be advantageously applied to a wide variety of systems and application software. Moreover, the implementations can be embodied on a variety of different computing platforms, including Macintosh, UNIX, LINUX, etc.

[0160] It should be understood that the specific order or hierarchical structure of steps, operations, or instructions in the processes or methods disclosed herein is an example of an exemplary method. For example, the various steps, operations, or instructions discussed herein may be performed in different orders. Similarly, the various steps, operations, and / or instructions discussed and shown herein may be changed and processed in different orders. Based on design preferences, it should be understood that the specific order or hierarchical structure of such steps, operations, or instructions in the processes or methods discussed and shown herein may be rearranged. For example, the appended claims present the elements of various steps, operations, or instructions in an exemplary order and are not intended to limit one to the specific order or hierarchical structure presented.

[0161] The inventors have recognized non-abstract technical solutions to technical problems to improve computer technology by increasing its efficiency, including improving automated control systems for, for example, industrial facilities. Embodiments can provide technical improvements to computer technology, such as automated control systems including data processing systems, and can also provide non-abstract improvements to computer technology through technical solutions to the technical problems identified in the background section of this disclosure.

[0162] The implementation scheme requires less processing time in terms of memory and processing power in the underlying system, and also requires fewer resources, including the use of computer technology to implement dynamic nonlinear optimization of BESS. Such improvements may arise from various specific implementations of the scheme. The solution protected by the claims may be rooted in computer technology to overcome specific problems arising in the fields of computers and computer networks, including battery energy storage systems.

[0163] It should be understood that the variations and other features and functions disclosed above, or alternative forms thereof, can be advantageously combined into many other different systems or applications. It should also be understood that various alternatives, modifications, variations, or improvements that are not currently foreseen or anticipated can subsequently be made by those skilled in the art, and these alternatives, modifications, variations, or improvements are also intended to be covered by the following claims.

Claims

1. A method for optimizing a battery energy storage system, the method comprising: Data is received by an echo state network of a reservoir computer having a sparse and operatively connected hidden layer and whose hidden neurons have fixed and randomly assigned connectivity and weights, wherein the data is related to the operation of the battery energy storage system, and wherein the data includes at least: load data, state of charge (SOC), state of health (SOH), renewable energy data, non-renewable energy data, carbon emission data, carbon credit data, and weighted energy cost. The operation of the battery energy storage system is optimized by the echo state network based on the data input to and processed by the echo state network and related to the operation of the battery energy storage system. The results of two different solvers are compared using an optimization result analyzer, wherein at least one of the two different solvers includes a mixed-integer nonlinear solver. Based on the data output from the optimization results analyzer and the data output from the echo state network, a day-ahead schedule for the operation of the battery energy storage system is generated.

2. The method of claim 1, wherein the echo state network comprises a neural network.

3. The method of claim 1, further comprising controlling the battery energy storage system after optimizing the operation of the battery energy storage system based on the data input to and processed by the echo state network.

4. The method of claim 1, wherein at least a portion of the data undergoes a fast Fourier transform before being input into the echo state network.

5. A system comprising: An echo state network, wherein data is input into the echo state network of a reservoir computer having sparsely and operationally connected hidden layers whose hidden neurons have fixed and randomly assigned connectivity and weights, wherein the data is related to the operation of a battery energy storage system, and wherein the data includes at least: load data, state of charge (SOC), state of health (SOH), renewable energy data, non-renewable energy data, carbon emission data, carbon credit data, weighted energy cost, energy sales revenue, battery data, state of charge data, state of health data, weather data, and time of day data; and The operation of the battery energy storage system is optimized by the echo state network based on data input to and processed by the echo state network that is relevant to the operation of the battery energy storage system. The results of two different solvers are compared using an optimization result analyzer, wherein at least one of the two different solvers includes a mixed-integer nonlinear solver. Based on the data output from the optimization results analyzer and the data output from the echo state network, a day-ahead schedule for the operation of the battery energy storage system is generated.

6. The system of claim 5, wherein the echo state network comprises a neural network.

7. The system of claim 5, further comprising a controller for controlling the battery energy storage system after optimizing the operation of the battery energy storage system based on the data input to and processed by the echo state network.

8. A battery energy storage system, the system comprising: At least one processor and a memory, the memory storing instructions to cause the at least one processor to perform the following operations: An echo state network of a reservoir computer, consisting of a sparsely and operationally connected hidden layer with fixed and randomly assigned connectivity and weights of its hidden neurons, optimizes the operation of the battery energy storage system based on data input to and processed by the echo state network and related to the operation of the battery energy storage system. The results of two different solvers are compared using an optimization result analyzer, wherein at least one of the two different solvers includes a mixed-integer nonlinear solver. Based on the data output from the optimization results analyzer and the data output from the echo state network, a day-ahead schedule for the operation of the battery energy storage system is generated; as well as Data is received by the echo state network, wherein the data is related to the operation of the battery energy storage system, and wherein the data includes at least: load data, state of charge (SOC), state of health (SOH), renewable energy data, non-renewable energy data, carbon emission data, carbon credit data, and weighted average energy cost.

Images