KR102244443B1 - Real-time decision method and system for industrial load management in a smart grid - Google Patents

Abstract

The potential impact of evolving industrial load management on demand response (DR) programs has been widely recognized. The present invention proposes a real-time decision-making model for load management of an industrial production process facing continuously changing real-time price (RTP). Due to the inherent dependencies between successive operations in the manufacturing process, the decision model must take into account future load management. The problem lies in the uncertainty that real-time prices in the future cannot be known in advance. Taking this into account, a robust optimization has been adopted to deal with future price uncertainties, so the proposed model makes timely decisions about industrial load control when receiving real-time prices for the current time slot while taking into account load scheduling in future time slots. You can get off. The present invention was carried out in a steel powder manufacturing process. The simulation results confirmed the validity of the proposed real-time decision-making approach from various perspectives.

Description

Real-time decision method and system for industrial load management in a smart grid}

The present invention relates to a real-time determination method and system for industrial load management in a smart grid.

As the demand for electricity continues to increase, a huge burden is placed on the power grid. The traditional solution is to build the next generation of capacity to match supply and demand while facing the problem of increasing fuel and carbon emissions costs. With the advent of Smart Grid, demand response (DR) has begun to actively contribute to improving grid efficiency and reliability due to its ability to quickly respond to demand-supply mismatch by adjusting loads that are flexible in terms of demand. It has been widely recognized that industrial DR management relates to power-intensive industrial processes and has a potential impact on power-intensive industrial processes.

There are still few effective DR mechanisms designed for energy management in industrial manufacturing processes. In the past, there has been research on a day-ahead decision model. That is, assuming that the next day's electricity price (the usage time price) or the day's previous real-time price can be utilized in industrial facilities, the optimal energy consumption scheduling for the next day may be predefined by minimizing the cost of the day. Recently, an incentive-based real-time demand bidding strategy has been proposed for the energy management of the discrete manufacturing process, but it is assumed that the period during which incentives are issued to industrial facilities will be announced in advance. To some extent, the incentive-based real-time demand bidding method in the discrete manufacturing process is still deterministic and cannot accommodate the uncertainty under a dynamic smart grid environment. Overall, the definitive model described above faces unexpected events that can occur in the real-time phase.

In addition, with the deregulation of the electricity industry, electricity can be injected into the grid from various places at different times, and the intervention of demand-side resources in the electricity market further exacerbates the real-time imbalance between supply and demand. All of these factors lead to inaccuracies in price predictions the day before. In other words, the price a day ago (DA-RTP) cannot keep up with the real-time situation of the ever-changing wholesale electricity price. In these circumstances, innovative DR mechanisms must be devised to accommodate the volatility or uncertainty of dynamic real-time pricing (RTP).

Several portable real-time DR energy management methods have been proposed. This method responds to the received real-time price and can help the resident user make a timely decision to control the load. The DR method, which focuses on the simultaneous load control of equipment at the present time interval without considering future price uncertainty, cannot be applied directly to industrial facilities. Manufacturing processes are generally composed of different successive tasks, and since the tasks inherently work together and cannot be handled independently, an extended scheduling interval when modeling load management problems in industrial manufacturing processes. (For example, one day) is required.

Previous studies have proposed DR approaches for various applications that take into account future price uncertainties as well as making real-time energy management decisions for the current time slot (when provided with RTP) by formulating a time interval of one day. To deal with price uncertainty largely, two effective approaches were applied: a scenario-based probabilistic programming model and a robust optimization. Using the former approach, multiple scenarios must be created. Then, scenario-dependent action decisions are driven by the uncertainty in the electricity price of each scenario. It is necessary to use the probability distribution to generate the scenario, which is complicated to fit the probability distribution for uncertain data (e.g. future price). Moreover, the number of scenarios required to describe uncertain data is usually quite large, which incurs a significant computational burden and becomes difficult to handle, especially when the application program is complex.

Alternatively, a robust optimization has been adopted to model price uncertainty. Robust optimization is a useful tool for solving optimization problems with uncertain parameters, where a set of uncertainties (e.g. intervals) is used to account for the variability of uncertain parameters that can be easily generated. Using an uncertainty set, robust optimization guarantees feasibility for all realizations of the uncertain parameters within this set, and the computational burden is lighter compared to the probabilistic programming model.

However, all of these existing studies are designed for domestic load control or demand response operation at the grid level. There have been no prior studies dealing with how to manage real-time energy consumption in industrial production processes for instantaneously changing electricity prices.

US Patent Publication 2015/0206083

The present invention receives RTP for the current time slot in consideration of the dynamic characteristics of RTP (Real-Time Price) and timely (timely) for the industrial manufacturing process whenever robust optimization for handling future price uncertainties is adopted. To solve the problem of making load control decisions.

To this end, the method for determining industrial load management in real time according to the present invention is a method for an energy management device of an industrial facility to determine industrial load management in real time, wherein electricity prices are received at regular time intervals from a grid supplying power. Steps, the electricity price of the current time interval (t) and the amount of electricity supplied from the grid at the current time interval, the minimum value of the predicted electricity price of the future time interval after the current time interval, and the amount of electricity supplied from the grid for each future time interval, the future The step of generating an objective function based on a robustness level that reflects the degree of influence of the deviation of the predicted electricity price at a time interval, and an operation selection (ST) task arranged in the manufacturing process of the industrial facility so that the value of the generated objective function is minimized. And adjusting the energy demand of the motion regulation (CRT) task.

In addition, the energy management device for real-time determination of industrial load management according to the present invention receives electricity prices at regular time intervals from the grid supplying power, and the electricity price at the current time interval t and the grid at the current time interval. An objective function based on a robustness level that reflects the degree of variation in the amount of electricity supplied from, the minimum value of the predicted electricity price for future time intervals after the current time interval, the amount of electricity supplied from the grid for each future time interval, and the predicted electricity price for future time intervals. And adjusting the energy demand of the motion selection task (ST) and the motion control task (CRT) arranged in the manufacturing process of the industrial facility so that the value of the generated objective function is minimized.

In addition, the energy management system for real-time determination of industrial load management according to the present invention includes an industrial facility having a manufacturing process consisting of an operation mandatory (NS) task, an operation selection (ST) task, and an operation control (CRT) task, and power. By receiving electricity prices at regular time intervals from the supplying grid, the electric cost of the current time interval (t), the predicted electric cost of all future time intervals after the current time interval, and the degree of influence of the deviation of the predicted electricity price of the future time interval Energy that generates an objective function with the reflected toughness level as a variable and adjusts the energy demand of the operation selection (ST) task and the operation control (CRT) task of the manufacturing process of the industrial facility so that the value of the generated objective function is minimized Includes a management device.

In addition, the program recorded in the storage medium for determining the industrial load management according to the present invention is a program recorded in the storage medium for real-time determination of the industrial load management in the manufacturing process of an industrial facility. When electricity prices are received at each time interval, the electricity price of the current time interval (t) and the electricity demand of the manufacturing process in the current time interval, the range of the forecast electricity price for a future time interval after the current time interval, and manufacturing by future time interval The step of generating an objective function with a robustness level as a variable reflecting the degree of variation influence of the electricity demand of the process and the predicted electricity price in the future time interval, and selecting the operation of the manufacturing process so that the value of the generated objective function is minimized (ST ) Adjusting the energy demand of the task and motion control (CRT) task.

As described above, since the present invention employs robust optimization to deal with future price uncertainty, the real-time determination model according to the present invention is implemented in real time for the current time slot while considering load scheduling in the future time slot. Receiving prices has the effect of being able to make timely decisions about industrial load control.

1 shows a storage location between tasks.
Figure 2 shows the manufacturing process of the steel powder.
3 shows the amount of electricity generated by the solar panel in each time slot.
4 shows the actual optimal ratio (AOP) and estimated reference parameter (RP) for each date.
5 shows the total electricity demand of all tasks in case 1 (No DR).
6 shows the total electricity demand for all tasks in case 2 (DR with robust optimization).
7 shows changes in cold water and nitrogen storage.
8 is a comparison of electricity demand for four cases.
9 shows the load configuration of an industrial facility.
Figure 10 compares the daily costs for four cases.
11 shows the configuration of an energy management system for real-time determination of industrial load management according to the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. The configuration of the present invention and its effect will be clearly understood through the detailed description below.

It should be noted that prior to the detailed description of the present invention, a detailed description will be omitted when it is determined that the gist of the present invention may be unclear for a known configuration.

First, the main contents of the present invention are as follows.

-We propose a real-time DR decision framework to manage energy consumption of industrial manufacturing processes in a timely manner.

-Linking the current and future time slots together to form an extended scheduling interval, where robust optimization is adopted to model future price uncertainties.

-In order to lower the daily electricity cost, we propose a method to assist system designers who specify the level of robustness by estimating a tuning parameter (referred to as Reference Percentage (RP)).

-The entire steel powder manufacturing process was adopted as a case study. The simulation results confirmed the effectiveness of the proposed real-time decision approach from various perspectives.

The present invention is described in the order of formulating the problem of industrial load scheduling, a real-time decision-making model based on robust optimization, and a case study of a steel powder manufacturing process.

The present invention provides a real-time decision model for industrial facilities participating in a real-time price (RTP) based DR program. The facility's Energy Management Center (EMC) receives real-time prices from power providers a few minutes before the time interval to arrive and then manages the energy consumption of the industrial manufacturing process according to the received real-time prices. Such activities are repeated each time a new price is offered to the energy management center. Hereinafter, the mathematical formulation of the decision model, including process modeling, various motion constraints, and objective functions, will be described in detail.

Formulate the problem

1. Industrial load modeling

a) Constraints on the operation of other tasks

The industrial load can be viewed as a complex of other successive tasks. For each task k, I _k,t represents the operation state during the time interval t and is defined as a binary variable. If task k operates, I _k,t =1, if not, I _k,t =0. According to specific operating conditions, tasks are classified into three types.

1) A non-shiftable (NS) task must continue to run once it starts running. That is, I _k,t is always "1".

2) The shiftable (ST) task has two operation states of "on" and "off", that is, "1" and "0".

3) The controllable (CRT) task can operate at a different operation level during a _{predefined set Mk.} y Denotes _m,k,t as binary variables, and represents the state of the operation level m of task k at the time interval t. In one time interval, the task _{can be executed with only one operation level (m∈M k} ) and can be expressed as Equation 1.

b) material storage constraints

It is assumed that there is storage space between all two successive tasks. The storage space is denoted by s. _{The material storage S s,} t in the storage space s at the time interval t is as shown in Equation 2.

Here, P _s represents a set of tasks produced for the storage space s, and C _s represents a set of tasks consumed from the storage space s. P _s,k,t is the production amount of storage space s produced by task k at time interval t, and C _{s,k ',t} is the consumption amount of storage space s consumed by task k'at time interval t. It is the same as Equation 3 and Equation 4.

Here, P _m,k,t denotes the production rate of task k at operation level m, and C _m',k',t denotes the consumption rate of task k'at operation level m'.

1 shows a storage space s where task k produces material for s and task k'consumes material for s. At certain time intervals, in order to satisfy the operation of the processing machine, it is sometimes necessary to maintain a minimum amount of material flow. In addition, the stored material cannot exceed the maximum capacity limit. Therefore, S _s,t should be limited to the following equation (5).

c) final product constraints

For industrial applications, it is usually necessary to maintain a minimum amount of final product at the end of the scheduling interval. The final output demand is expressed as Θ, and the condition of Equation 6 below must be satisfied.

Here, F denotes a set of tasks that produce the final product. When task k is not a CRT, y _{m,k,t in} Equation 6 is replaced with I _k,t.

d) the electricity demand of the process

For a task k, if the electricity demand during the time interval t is expressed as e _k,t , e _k is called the power consumption rate of task k according to the task classification in a). Then, the electricity demand for each task type is modeled as follows.

The required operation (NS) task is shown in Equation 7.

The operation selection (ST) task is shown in Equation 8.

In the motion adjustment (CRT) task, the electricity demand is modeled as shown in Equation 9 according to the binary variable determined in Equation 1.

Here, e _m,k denotes the power consumption rate of task k at the operation level m.

Therefore, the total electricity demand (E _t ) for one hour interval can be obtained by summing all the demands of each task k.

2. ESS modeling

Energy storage systems (ESS) exist in industrial facilities for the purpose of mitigating the grid load by supplying energy to the facility during electricity shortage times. During the time interval t, the ESS state is modeled using Equation 11. Equation 11 is calculated from the state of the previous time interval and also considers the amount of charge or discharge of the ESS.

Equation 11 has a constraint expressed by the following equation.

Here, at any time interval, state ESS ^(ESS E _t) is the capacity as the constraints of equation (12) ^- should be in the (E ^ESS). Constraints (Equations 13 and 14) indicate that the possible charging or discharging amount (E _ch,t ^ESS or E _dis,t ^ESS ) exceeds the charging rate or discharging rate (E _ch ^rate or E _dis ^ESS ) for a unit time (1 slot). It indicates that it should not be done. y _t ^ESS is a binary variable that controls the transition between charge and discharge modes. Equation 15 categorizes the discharged electricity into two, where E _use,t ^ESS represents a part used to meet the industrial load, and E _sold,t ^ESS is the remaining part, which is resold to the grid for profit. η _ch and η _dis are charging and discharging efficiency.

3. DER modeling

Industrial facilities are also assumed to have distributed energy resources (DER). In the present invention, solar energy may be an example. The produced energy (E _pro,t ^DER ) is divided into two, one for satisfying the industrial load (E _use,t ^DER ) and one for resale to the grid (E _sold,t ^DER ), which is Equation 16 Is modeled as

4. Power balance

During the time interval (slot) t, the electricity demand (E _t ) of the industrial manufacturing process and the charging demand (E _ch,t ^ESS ) of the _{ESS are supplied by the grid (E t} ^grid ) or the ESS (E _use,t ^ESS ) and It is supplied by binding energy provided by DER(E _use,t ^DER).

5. Maximum Demand Constraints

Peak demand is considered an important factor when designing industrial load scheduling. This is because many industrial facilities have to comply with the maximum electricity demand limit. Therefore, electricity drawn from the grid (E _t ^grid ) cannot exceed the maximum limit (L _{1 ).}

6. Total power entering the grid

The total power that can be resold to the grid comes from both resources (ESS and DER).

7. Electricity trading constraints

Exceeding the maximum demand constraint of Equation 18, the power received from the grid or the power entered into the grid during one slot are mutually exclusive. This can be limited using Equation 20 and Equation 21.

Here, y _t ^grid is a binary variable ("0" or "1") used to control switching between the two options, and L ₂ represents the maximum amount of electricity that can be resold to the grid.

8. Objective function

The goal is to minimize the total cost of electricity consumption over the entire time interval (separated into current slot (t) and future slot (t+1, ..., T)), and the total cost is the cost of the current slot (RTP at slot t). π _t ) and the total cost of the future slot ( _{assuming the future price as π τ} , ∀τ∈[t+1, T]). The reason for considering the remaining Tt slots of the scheduling period is to maintain adaptability. This is because successive tasks of the manufacturing process are interdependent and cannot be handled independently. Using the extended scheduling interval, the proposed model can avoid solutions that are not feasible and difficult to implement.

Taking the current slot t as an example, the cost is calculated as the difference between the cost of purchasing energy from the grid at the _{price π t (E t} ^grid ) and the profit from selling energy to the grid at the _{price π sell (E sold,t ).} ε ₁ ·E _sold,t ^DER and ε ₂ ·E _sold,t ^ESS are defined as artificial penalties imposed on other energy sources. Here, ε ₁ and ε ₂ are assumed ^{to be small enough (e.g., e -7} or 2e ^-7 ) to ensure that the total cost is not affected. By adding two penalty terms, the priorities of selling energy to different resources can be differentiated. That is, as the value of ε is smaller, the energy management sensor (EMC) will preferentially sell all possible energy from that resource before considering other resources. The present invention is ε ₁ Assume that <ε _2. This is because the energy cost of DER is generally considered to be less than that of ESS.

Robust optimization methodology

1. Robust optimization model

Note that the objective function of Equation 22 is not officially defined. This is because the future electricity cost (π _τ ) is an unknown value. The present invention employs robust optimization to manage future price uncertainties. The core concept of robust optimization is to minimize the overall cost of Equation 22 without an _{accurate π τ value.} Price intervals are adopted to model the uncertainty of future prices without losing generality. The price interval is _{denoted by [π τ} ^mim , π _τ ^max ], π _τ ^mim and π _τ ^max are the lower and upper limits of the price interval, and π _τ varies during the future time slot. The price interval can be provided by a subscribed power provider, or can be obtained using a price prediction model (eg, a time serial model and a neural network). The present invention focuses on exploring the possibility of adopting robust optimizations in real-time decisions for load management of industrial manufacturing processes. The price prediction can be adopted from the current technology and is outside the scope of the present invention.

When a price interval is provided, the strong correspondence of Equation 22 is formulated as in Equation 23.

In Equation 23, Γ is a parameter specified by the system designer that controls the robustness level of the answer with respect to the uncertainty of future price. The Γ value can be adjusted from 0 to T-t (maximum value of future slots). For a particular Γ value, it represents the maximum number of allowable price deviations. In particular, Γ = 0 represents the most optimal case that completely ignores the effect of price deviation in the objective function. On the other hand, Γ = T-t represents the most conservative case, which takes into account the simultaneous effects of price deviation (or uncertainty) in all future time slots.

Looking at the inner maximization of Equation 23, T ₀ is a set of future time slots. That is, T ₀ ⊆ [t+1,T]. Here, the number of components of T ₀ must be less than Γ. This means that the maximum number of permissible price deviations is Γ. Internal Maximization attempts a worst-case simulation of an uncertain price. In other words, we want to find the worst case of price uncertainty that has the greatest impact on the overall cost (which causes the greatest increase in the overall cost). Then, the external minimization of Equation 23 aims to obtain an optimized solution for all cases. Such a solution is robust in the sense that the number of prices allowed to change in each interval [π _τ ^mim , π _τ ^{max] is up to Γ.}

It should be noted that the problem in Equation 23 is the minimum and maximum problem, and it is generally difficult to solve such a problem. According to the duality gap theory, this minimum and maximum problem can be converted into a single level of optimization problem as shown in Equation 24 below.

Equation 24 follows the constraints expressed by Equations 1 to 21 and the constraints expressed by Equations 26 to 30 below.

The variables λ and ξ _τ are double variables of the initial problem (Equation 22) used when considering the price limit. η _τ is an auxiliary variable used to obtain a uniform linear equation. In Equation 24, Γ may have any real number from 0 to the number of all unknown prices, and the greater the value of Γ, the greater the level of toughness when it is expensive.

The objective function of Equation 24 including constraints (Equations 1-21 and 26-30) formulates a mixed-integer linear programming (MILP) optimization problem that can be efficiently solved using a commercially available software package. By solving the problem for each slot t, not only energy consumption decisions for the current and remaining slots, but also conversion control variables such _{as y t} ^ESS and y _t ^{grid are obtained.} However, only decisions are made on the current slot, which can provide industrial facilities with optimal energy management guidance for the current slot.

From Equation 24, it can be seen that the daily cost varies according to the value of Γ. We propose a method that can help system designers in specifying the level of robustness to lower their daily electricity costs.

2. How to specify the level of toughness

Ideally, the toughness level Γ in Equation 24 may have a real number between 0 and the maximum number of unknown prices by multiplying the percentage by T-t, for example. We propose a method that can help system designers by estimating the adjustment parameter RP based on the daily criterion and controlling the daily cost to the lowest level when specifying Γ. Then, Γ may be determined as in Equation 31 by multiplying RP by T-t.

Before explaining the estimation method in detail, a distinguishable parameter, that is, an Actual Optimal Percentage (AOP) is introduced. A specified level of toughness can incur a minimum daily cost at an actual optimal rate. Actually, the AOP is not known in advance, but can be obtained in the reverse direction. That is, various percentages can be listed at the end of the day. For each percentage listed, the daily cost can be inversely calculated and then the percentage that incurs the least daily cost is considered an AOP. AOP is differentiated every day because it relies on constantly changing electricity prices and electricity demand. In this regard, the RP of the current date d is estimated as shown in Equation 32 based on past data of RP and AOP using an exponential smoothing model.

Here, RP _d-1 and AOP _d-1 mean the RP of the previous day (d-1) and the actual optimal ratio, respectively, and RP _d is the RP of the current date d. β(0<β<1) is specified by the system designer as the smoothing factor of the exponential smoothing model.

According to Equation 32, RP _d may be estimated based on a daily criterion, and then it is used to specify the toughness level Γ at the start of each slot t when the schedule interval proceeds.

Case study

1. Composition of the case

In order to verify the effectiveness of the robustness optimization-based method according to the present invention responding to RTP, a case study as shown in FIG. 2 is performed by adopting a steel powder manufacturing process composed of a plurality of tasks. According to the function and characteristics of each task, it is classified into 3 types, 2 NS tasks (blast furnace, finishing-reduction furnace), 4 ST tasks (spraying, dehydration, drying, magnetic separation), and 5 CRT tasks (crushing, sorting). , Mixing, water cooling, nitrogen generation). The last two CRT tasks provide cold water and nitrogen for the tasks of spraying, dewatering, drying and finishing-reduction furnaces. A furnace is an energy-intensive steel pile used to supply liquid iron to multiple production lines simultaneously. Since the focus was on modeling a single production line, the energy consumption of the furnace was excluded from this case study.

Table 1 or Table 2 lists three types of parameters as shown in FIG. 2. The minimum storage requirements for all tasks in Table 1 are set to zero in this manufacturing process. This is because if the material produced by the previous task is unavailable, the subsequent task stops working. The parameters of ESS are shown in Table 3. For simplicity, the DER in this example is assumed to be a solar panel. The amount of electricity produced in each time slot is shown in FIG. 3. 3 is shown based on the common sense that the electricity production rate is highest at noon when the solar irradiance is the maximum and electricity is not generated at night. In fact, the probabilistic properties of solar energy can be predicted using sophisticated methods based on weather forecast information.

The final production requirement Θ was set to 90 tons. The maximum amount of electricity L ₁ supplied from the grid is 500 kWh, and the maximum amount of electricity L ₂ sold to the grid is 100 kWh. In practice, these values are application specific and must be set by the industrial facility designer or the power supply side.

The total time interval is divided into 24 time slots representing the 24 hours of a day. The time interval starts at 00;00, and simulations were carried out from time to time. Assuming that the RTP in each slot is received in advance (eg, 5 minutes ago), the robust optimization problem (Equations 24-30) according to the present invention was solved using Gurobi software.

The computation time required to solve the single slot optimization problem was 1 second. This short time can sufficiently satisfy the requirements for deploying real-time DR management in industrial manufacturing processes. The output of the problem solver (Gurobi) includes the operating status or level of each task, the electricity demand of the ESS, and the energy transaction between the industrial facility and the grid.

Simulations were performed from July 1st to August 11th, 2015 based on the utility company's RTP. On the first day, RP _d-1 was assumed to be 50%. On the next day, RP _d was estimated using Equation 32. Here, β was set to 0.5 so that there was no discrimination _{between RP d-1} and AOP _d-1.

4 shows the AOP and estimated RP for each date. It can be seen that RP shows a very similar trend to AOP. This proves that the method of estimating RP is effective.

Table 4 shows the RTP of August 11. Here, the second column is the actual price, and the third and fourth are the corresponding price ranges, formulated as a percentage of the actual price (for example, 1±α). In the simulation, α was set to 20%.

In the next part, we analyze the simulation results for the representative date (August 11). The price of selling surplus electricity to the grid was assumed to be 10% cheaper than each RTP.

2. Results with and without real-time demand response

To demonstrate the performance of the model according to the present invention, a criterion without real-time DR was made, where a fixed flat price was applied to the model. That is, the objective function of Equation 22 simply minimizes the _{daily cost (E t} ^grid π _{flat) by integrating the cost of each slot.} Let this be case 1 (No DR).

Compared to the baseline, we investigate the refined performance of the real-time decision model using robust optimization. Let this be Case 2 (DR with robust optimization applied). Here,'refined' means that the model considered only the energy demand of the manufacturing process itself, not ESS and DER.

5 and 6 show the total electricity demand of all tasks in Case 1 and Case 2, respectively. Clearly, when a fixed flat price (this value has the average RTP in Table 4) is applied, the industrial facility receives no incentive to reduce or move the energy demand, but simply schedule the electricity demand in the starting slot as early as possible. Complete the daily production target (see Equation 6). In comparison, in case 2, when a robust optimization-based real-time decision model was used, electricity demand was properly scheduled. Specifically, most of the electricity demand for CRTs and STs was scheduled with non-peak slots. So, because only the NS task needed to run and both other types of tasks were stopped, the overall demand was maintained at a fairly low level in the peak slot. This confirms that the model according to the invention can respond appropriately to RTP.

To gain insight into the manufacturing process, we chose two Water Cooling and Nitrogen Generating (CRT) tasks to describe the production process over time. 7 shows the process of accumulation and consumption of cold water and nitrogen. When the RTP is low, you can see that each store continues to grow. This stacking process correlates with the optimal load scheduling results of FIG. 6 since the main power demand was planned for a low cost period. In particular, it was found that nitrogen storage increased slightly in slot 11. The reason is that RTP has suddenly decreased compared to slot 10. Thus, this slight increase in nitrogen storage demonstrates that the robust optimization model of the present invention can respond to instantaneous changes in RTP in an effective way.

3. Results of considering ESS and DER

Here, the effects of ESS and DER were considered. Accordingly, two additional cases were designated: Case 3 (DR with Robust Optimization and ESS) and Case 4 (DR with Robust Optimization and ESS and DER). Figure 8 compares the demand (purchased on the grid) for the four cases. Because ESS stores energy when the RTP is low (more demand occurs during non-peak hours), more demand has shifted from peak to off-peak hours in Case 3 compared to Cases 1 and 2, and RTP is extremely high. When it is high, it can be seen that it supplies power to the system requirements. With the help of ESS, the peak demand for industrial equipment has been maintained at a very low level (eg slots 12, 14, 15 and 16) which is believed to be able to significantly relieve the grid burden. Considering DER in the case 4 model as well, the demand during peak hours in slots 12, 14, 15 and 16 was further reduced or even zeroed. As shown by negative values in FIG. 8, in slots 12 and 15, surplus electricity was sold back to the grid, generating profit.

In addition, case 4 was analyzed extensively by decomposing the total load of industrial equipment into several parts. As can be seen in FIG. 9, red and green bars represent electricity that is purchased from the grid and used in the manufacturing process to charge the ESS. The purple bars represent the electricity emitted by the ESS, and the blue bars show how the energy generated by the DER is distributed across the facility. Obviously, the electricity drawn from the grid was properly controlled in response to RTP. In other words, most of the power (red and green bars) was purchased from the grid during the off-peak period, and only a small amount was needed from the grid during the peak period as the system requirements were met by the energy stored in the DER and ESS. The energy generated by the DER (see Fig. 3) was used enough to meet the system requirements or to make a profit by selling to the grid in slots 12 and 15. In addition, electricity sold to the grid comes only from DER, not ESS. This is because the energy sales priority of other resources is distinguished by the objective function of Equation 22.

4. Comparison of cost results

Figure 10 shows the daily electricity costs for four cases on August 11, 2015. Compared to Case 1 to which DR was not applied, the daily cost of Case 2 (when the DR model of the present invention was deployed) was reduced by 13.5%. This can serve as a key motivation for industrial facilities to participate in RTP-based DR programs. In addition, when ESS and DER are added in Cases 3 and 4, the cost is further reduced by 9% and 10%, showing the additional benefits that industrial facilities can obtain by adding ESS and DER.

11 shows the configuration of an energy management system for real-time determination of industrial load management according to the present invention.

Referring to FIG. 11, the energy management system according to the present invention includes a grid 10 for supplying power, an energy management device 20 for determining industrial load management in real time, an industrial facility 30 having a manufacturing process, and the like. .

The grid 10 notifies the electricity price and supplies power at regular time intervals (for example, every hour).

The industrial facility 30 has an operation required (NS) task, an operation selection (ST) task, and an operation regulation (CRT) task placed within the manufacturing process.

The energy management device 20 is installed with a program to which the robust optimization model according to the present invention is applied.

When the energy management device 20 receives the electricity price from the grid 10 at regular time intervals, the purpose of Equation 24 according to the constraints expressed by Equations 1 to 21 and the constraints expressed by Equations 26 to 30 Create a function.

The energy management device 20 includes the electricity price of the current time interval t and the amount of electricity supplied from the grid 10 at the current time interval, the minimum value of the predicted electricity price of the future time interval after the current time interval, and the grid for each future time interval ( An objective function is generated based on the robustness level (Γ) reflecting the degree of influence of the difference in the amount of electricity supplied from 10) and the predicted electricity price in the future time interval.

That is, the objective function has a robustness level that reflects the degree of variation of the electric cost of the current time interval (t), the predicted electric cost of all future time intervals after the current time interval, and the predicted electric price of the future time interval as variables.

In this case, the energy management device 20 may add a profit generated when electricity stored in the energy storage device of the industrial facility 30 is sold to the grid 10 as a variable.

And the value of the robustness level, which is a variable of the objective function, is a real number that exists between 0 and the maximum number of future time intervals, and can be obtained by multiplying the value of the maximum number of future time intervals (Tt) by the daily estimated adjustment parameter (RP). have.

In addition, the daily estimated adjustment parameter (RP) is obtained based on the adjustment parameter of the previous day and the actual optimum ratio (AOP) of the previous day, and the actual optimum ratio (AOP) is a plurality of percentages (%) listed at the end of the last interval of the day. ) Is a percentage corresponding to the minimum value among the daily costs of the industrial facility 30 calculated for.

When generating such an objective function, the energy management device 200 adjusts the energy demand of the motion selection (ST) task and the motion control (CRT) task arranged in the manufacturing process of the industrial facility 30 so that the value of the objective function is minimized. .

In this way, the energy management device 20 controls the operation of the task of the industrial facility 30 so that the total cost is minimized while applying a robust optimization every time interval.

The above description is merely illustrative of the present invention, and various modifications may be made without departing from the spirit of the present invention by those of ordinary skill in the technical field to which the present invention pertains.

Therefore, the embodiments disclosed in the specification of the present invention are not intended to limit the present invention. The scope of the present invention should be interpreted by the following claims, and all technologies within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

10: grid 20: energy management device
30: industrial facilities

Claims (22)

In a method for an energy management device of an industrial facility to determine industrial load management in real time,
A step in which the rate unit price is changed according to the amount of electricity purchased by each customer at regular time intervals from the grid supplying electricity; receiving an electricity price that varies according to the electricity production conditions of the grid and the total electricity consumption of the customer, rather than a fee,
The electricity price of the current time interval (t) received from the grid in real time, the current cost by the amount of electricity supplied from the grid at the current time interval, the minimum value of the predicted electricity price of the future time interval after the current time interval, and the above for each future time interval The future cost by the amount of electricity supplied from the grid, a real number that exists between 0 and the maximum number of future time intervals, which generates an objective function by summing the values of the robustness level reflecting the degree of the effect of the deviation of the predicted electricity price of the future time interval. Step and,
For determining in real time industrial load management comprising the step of adjusting the energy demand of an operation selection (ST) task and an operation control (CRT) task arranged in the manufacturing process of the industrial facility so that the value of the generated objective function is minimized. Way. The method of claim 1,
A method for real-time determination of industrial load management, characterized in that when generating the objective function, a profit generated when electricity stored in an energy storage device is sold to the grid is added as a variable. delete The method of claim 1,
The value of the toughness level is calculated by multiplying the value (Tt) of the maximum number of future time intervals by an adjustment parameter (RP) estimated every day. The method of claim 4,
The daily estimated adjustment parameter is calculated based on the adjustment parameter of the previous day and the actual optimal ratio (AOP) of the previous day. The method of claim 5,
The actual optimal ratio (AOP) is a percentage corresponding to the minimum value among daily costs of industrial facilities calculated for a plurality of percentages (%) enumerated at the end of the last interval of the day. Way to do it. A step in which the price unit price changes according to the amount of electricity purchased by each customer at regular time intervals from the grid supplying electricity. The electricity price received in real time at the current time interval (t) and the current cost by the amount of electricity supplied from the grid at the current time interval, the minimum value of the predicted electricity price at the future time interval after the current time interval, and supply from the grid for each future time interval An objective function is generated by summing the value of the robustness level reflecting the degree of the deviation influence of the predicted electricity price in the future time interval as a real number between 0 and the maximum number of future time intervals. Energy management device for real-time determination of industrial load management, characterized in that the energy demand of the motion selection task (ST) and the motion control task (CRT) arranged in the manufacturing process of an industrial facility are adjusted so that the value of one objective function is minimized . The method of claim 7,
An energy management device for real-time determination of industrial load management, characterized in that when generating the objective function, a profit generated when electricity stored in an energy storage device is sold to the grid is added as a variable. delete The method of claim 7,
The energy management apparatus for determining industrial load management in real time, characterized in that the value of the toughness level is calculated by multiplying the value (Tt) of the maximum number of future time intervals by the adjustment parameter (RP) estimated every day. The method of claim 10,
The energy management device for determining industrial load management in real time, characterized in that the adjustment parameter estimated every day is calculated based on the adjustment parameter of the previous day and the actual optimal ratio (AOP) of the previous day. The method of claim 11,
The actual optimal ratio (AOP) is a percentage corresponding to the minimum value among daily costs of industrial facilities calculated for a plurality of percentages (%) enumerated at the end of the last interval of the day. Energy management device for use. An industrial facility with a manufacturing process consisting of an action required (NS) task, an action selection (ST) task, and an action control (CRT) task,
A step in which the price unit price changes according to the amount of electricity purchased by each customer at regular time intervals from the grid supplying electricity, but receives the electricity price that varies according to the electricity production conditions of the grid and the total power consumption of the customer, rather than a fee, and from the grid. The electricity price received in real time at the current time interval (t) and the current cost by the amount of electricity supplied from the grid at the current time interval, the minimum value of the predicted electricity price at the future time interval after the current time interval, and the supply from the grid for each future time interval An objective function is generated by summing the value of the toughness level reflecting the degree of the deviation influence of the predicted electricity price in the future time interval as a real number existing between 0 and the maximum number of future time intervals. Energy management for real-time determination of industrial load management including an energy management device that adjusts the energy demand of the operation selection (ST) task and the operation control (CRT) task of the manufacturing process of the industrial facility so that the value of one objective function is minimized system. The method of claim 13,
The operation essential task is a load that always operates at all time intervals, the operation selection task is a load that operates or does not operate at a specific time interval, and the operation control task is a load that can operate at a different operation level for each time interval. Energy management system for determining in real time industrial load management, characterized in that. The method of claim 13,
The electricity cost of the current time interval is a value obtained by multiplying the electricity price of the current time interval by the electricity demand of all tasks in the manufacturing process. The method of claim 13,
Real-time determination of industrial load management, characterized in that the predicted electricity cost of all future time intervals after the current time interval is the sum of a value obtained by multiplying the minimum value of the predicted electricity price and the predicted electricity demand of all tasks in the manufacturing process for each future time interval. Energy management system for The method of claim 13,
The industrial facility further comprises an energy storage device for storing electricity,
The energy management system for determining industrial load management in real time, characterized in that the energy management device adds a profit generated when the electricity stored in the energy storage device is sold to the grid as a variable of the objective function. The method of claim 13,
Energy for real-time determination of industrial load management, characterized in that the smaller the value of the toughness level, the less the effect of the price deviation in the objective function is evaluated, and the larger the value, the greater the effect of the price deviation in the objective function is evaluated. Management system. As a program recorded in a storage medium for real-time determination of industrial load management in the manufacturing process of an industrial facility,
The current time when the price of electricity that changes according to the electricity production conditions of the grid and the total amount of electricity used by the customer is received, rather than a fee, according to the amount of electricity purchased by each customer at regular time intervals from the grid supplying electricity. The electricity price of the interval (t) and the current cost due to the electricity demand of the manufacturing process in the current time interval, the range of the forecast electricity price for the future time interval after the current time interval, and the future cost due to the electricity demand of the manufacturing process by future time interval , Generating an objective function by summing the values of the toughness level reflecting the degree of variation influence of the predicted electricity price of the future time interval as a real number existing between 0 and the maximum number of future time intervals,
A program recorded in a storage medium, characterized in that performing the step of adjusting the energy demand of an operation selection (ST) task and an operation control (CRT) task of a manufacturing process so that the value of the generated objective function is minimized. The method of claim 19,
The electricity cost of the current time interval is calculated by multiplying the electricity price of the current time interval (t) by the electricity demand of the manufacturing process at the current time interval, and the minimum value and future time of the predicted electricity price range for a future time interval after the current time interval In a storage medium, characterized in that an objective function is generated by calculating the total predicted electricity cost, which is the sum of values multiplied by the predicted electricity demand of the manufacturing process for each interval, and adding the electricity cost of the current time interval, the total predicted electricity cost, and the toughness level. Recorded program. The method of claim 20,
A program recorded in a storage medium, characterized in that generating an objective function by adding a profit generated when electricity stored in the energy storage device of the industrial facility is sold to the grid as a variable. The method of claim 21,
The above revenue includes the total forecast revenue obtained by adding all of the current revenue generated in the current time interval and the forecasted revenue generated in the future time interval,
And subtracting the current revenue from the electricity cost of the current time interval and subtracting the total predicted revenue from the total predicted electricity cost, thereby adding the revenue as a variable.

Images