GOVERNMENT SPONSORED RESEARCH OR DEVELOPMENT

[0001] This invention was made with government support under award number DE EE0001 1 13, awarded by the Department Of Energy (DOE). The government has certain rights in the invention.

CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. provisional application Serial No. 61/908,326 filed November 25, 2013 which is incorporated herein by reference in its entirety.

TECHNICAL FIELD

[0003] The present application is related to systems, methods, and apparatuses for statistically modeling and forecasting building energy and cooling demands. The technology described herein may be generally applied, for example, to optimize the operation of combined cooling, heating, and power (CCHP) systems.

BACKGROUND

[0004] In most real cases, building cooling and electricity loads are highly dynamic oscillating within a wide range of values during course of a day. This is mainly because several physically explicit or latent factors can instantaneously influence cooling and electricity demand patterns. These factors (variables) may include, for example, (i) static factors that are usually set at the design stage and only change due to time, e.g., ageing and wear and tear. Building characteristics, CCHP components, chiller types and generator nominal capacities are examples of such factors; (ii) Environmental variables extrinsic to the building, such as climate and weather data; (iii) operational variables, e.g. cooling/heating set point values, lighting, time schedule to operate various equipment and system components within plant or building; and (iv) uncontrollable dynamical variables, such as number of occupants at any time, noise due to structural variations (e.g., building dimensions). All of these factors and their impacts on energy consumption dynamics in order to optimally forecast and control cooling and electricity demands for a single building or a cluster of buildings. However, a complete forecast model is not practically attainable due to unknown significant dynamical variables, lack of tools to measure their effects, or that some of these variables are uncontrollable.

[0005] Researchers often apply two different approaches to model and forecast cooling and electricity load demands. The first approach is Box and Jenkins time series models where cooling and electricity load demands are forecasted based upon linear combinations of their past values. The major drawback of such models is that the future values are typically forecasted based upon the past and present values of cooling and electricity load demands without considering any exogenous factors in the model. In the second approach, the cooling and electricity load demands are modeled solely based upon exogenous factors. Multiple linear and nonlinear regression models, artificial neural networks, decision tree techniques are some examples of the second approach. These models often ignore the complex interactions between exogenous factors, which may result in less accurate forecast values. To overcome this problem, a number of studies use a hybrid approach, which employs the main components of both aforementioned approaches. Autoregressive with exogenous variable (ARX) and autoregressive moving average with exogenous variable (ARMAX) are two examples of this approach. Although these models perform effectively in some cases, they have many parameters to estimate since all input and output variables with their past and current values must appear in the forecast. Furthermore, as the number of parameters increases, the accuracy of estimates tends to decrease particularly when the model is used for forecast purposes.

SUMMARY

[0006] Embodiments of the present invention address and overcome one or more of the above shortcomings and drawbacks, by providing methods, systems, and apparatuses related to statistically modeling and forecasting building energy and cooling demands. This technology is particularly well-suited for, but not limited to, optimizing the operation of combined cooling, heating, and power (CCHP) systems used at sites such as university campuses.

[0007] According to some embodiments, a computer-implemented method for using an integrated methodology of forecasting and optimization to determine optimal values for a Combined Cooling, Heating, and Power (CCHP) system associated with a site includes constructing a time-series regression model for cooling and electricity load demand based on long-term historical data associated with the CCHP system. The time-series regression model for cooling and electricity load demand is applied to weather forecast data and short-term historical data associated with the CCHP system to yield one or more predictions of cooling and electricity load demand. A plurality of optimal CCHP data settings is determined for devices included in the CCHP system using the one or more predictions of cooling and electricity load demand.

[0008] Various additional features or enhancements may be applied to the aforementioned method of using an integrated methodology of forecasting and optimization. For example, the optimal CCHP data settings may be transmitted to a human-machine interface display and/or automatically applied to the one or more of the devices included in the CCHP system. Moreover, the implementation of the time-series regression model may vary according to different embodiments. The time-series regression model may apply, for example, a Cochrane- Orcutt procedure to the long-term historical data. Additionally, the time-series regression model may utilized occupancy and weather data associated with the site.

[0009] Additionally, the details of the long-term historical data used in the aforementioned method of using an integrated methodology of forecasting and optimization, may vary according to different embodiments. For example, the long-term historical data may include one or more of (i) past power output data corresponding to at least one of the devices included in the CCHP system (ii) schedule data corresponding to at least one of the devices included in the CCHP system; (iii) power consumption values corresponding to at least one of the devices included in the CCHP system; and/or (d) a measure of imported electricity from an electrical grid to the CCHP system.

[0010] Various techniques may be used for determining the plurality of optimal CCHP data settings for the devices included in the CCHP system in the aforementioned method of using an integrated methodology of forecasting and optimization. For example, an optimization process may be employed. In one embodiment, this optimization process includes determining a first set of optimal CCHP data settings corresponding to a chiller and a thermal energy storage included into the CCHP system based on an approximation of a Mixed Integer Nonlinear Programming (MINLP) problem using a set of MINLP problems. Additionally, a second set of optimal CCHP data settings corresponding to an onsite power generation system and a grid import control included into the CCHP system are determined based on a nonlinear program problem. In another embodiment, the optimization process includes determining a set of optimal set points corresponding to a chiller and a thermal energy storage included into the CCHP system, wherein the set of optimal set points minimize a chiller electricity consumption value and determining an optimal gas turbine operating schedule and an optimal power grid purchase plan based on the chiller electricity consumption value and electricity demand associated with the site. The plurality of optimal CCHP data settings then comprise the set of optimal set points, the optimal gas turbine operating schedule, and the optimal power grid purchase plan.

[0011] According to other embodiments, a computer-implemented method for determining optimal values for a CCHP system associated with a site includes training an energy demand forecasting model based on long-term historical energy demand data associated with the CCHP system and training a cooling demand forecasting model based on long-term historical cooling demand data associated with the CCHP system. Two types of data are acquired: weather forecast data providing near-term predicted weather information for a predetermined number of hours from a present time and short-term historical data associated with the CCHP system. The energy demand forecasting model is used to determine one or more forecasted energy values based on the weather forecast data and the short-term historical data. The cooling demand forecasting model is used to determine one or more forecasted cooling values based on the weather forecast data and the short-term historical data. An optimization process is applied to the one or more forecasted energy values and the one or more forecasted cooling values to yield a plurality of optimal CCHP data settings for devices included in the CCHP system.

[0012] Various additional features or enhancements may be applied to the aforementioned method for determining optimal values for a CCHP system associated with a site. For example, the optimal CCHP data settings may be transmitted to a human- machine interface display and/or automatically applied to the one or more of the devices included in the CCHP system. In some embodiments, training of at least one of the energy demand forecasting model and the cooling demand forecasting model incorporates occupancy data associated with the site. Additionally, various types of long-term energy historical data may be used. For example, in one embodiment, the long-term energy historical data comprises a measure of imported electricity from an electrical grid to the CCHP system.

[0013] Additionally, various optimization processes may be used with the aforementioned method for determining optimal values for a CCHP system. For example, in one embodiment, the optimization process determines two sets of CCHP data settings. The first set of optimal CCHP data settings correspond to a chiller and a thermal energy storage included into the CCHP system. Determination of the first set of optimal CCHP data settings is based on an approximation of a Mixed Integer Nonlinear Programming (MINLP) problem using a set of MINLP problems. The second set of optimal CCHP data settings correspond to an onsite power generation system and a grid import control included into the CCHP system. Determination of the second set of optimal CCHP data settings is based on a nonlinear program problem. In other embodiments, the optimization process includes determining a set of optimal set points, an optimal gas turbine operating schedule, and an optimal power grid purchase plan. The set of optimal set points correspond to a chiller and a thermal energy storage included into the CCHP system, such that the set of optimal set points minimizes a chiller electricity consumption value. The optimal gas turbine operating schedule and the optimal power grid purchase plan are determined based on the chiller electricity consumption value and electricity demand associated with the site. Then, the plurality of optimal CCHP data settings used in the aforementioned method comprises the set of optimal set points, the optimal gas turbine operating schedule, and the optimal power grid purchase plan.

[0014] According to other embodiments, system for determining optimal values for a CCHP system associated with a site includes a weather adapter, a forecast adapter, and an optimization adapter. The weather adapter is configured to acquire weather forecast data providing near-term predicted weather information for a predetermined number of hours from a present time. The forecast adapter is configured to use an energy demand forecasting model to determine one or more forecasted energy values based on the weather forecast data and short- term historical data associated with the CCHP system. The forecast adapter also uses a cooling demand forecasting model to determine one or more forecasted cooling values based on the weather forecast data and the short-term historical data associated with the CCHP system. The optimization adapter is configured to apply an optimization process to the one or more forecasted energy values and the one or more forecasted cooling values to yield a plurality of optimal CCHP data settings for devices included in the CCHP system. In one embodiment, the system also includes a network driver configured to provide the plurality of optimal CCHP data settings to a human-machine interface.

[0015] Additional features and advantages of the invention will be made apparent from the following detailed description of illustrative embodiments that proceeds with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The foregoing and other aspects of the present invention are best understood from the following detailed description when read in connection with the accompanying drawings. For the purpose of illustrating the invention, there is shown in the drawings embodiments that are presently preferred, it being understood, however, that the invention is not limited to the specific instrumentalities disclosed. Included in the drawings are the following Figures:

[0017] FIG. 1 provides a schematic of a CCHP plant that may be managed using one or more of the techniques described herein, according to some embodiments of the present invention;

[0018] FIG. 2 provides a high-level overview of a system which may be used to optimize CCHP systems, such as those illustrated in FIG. 1), according to some embodiments of the present intention;

[0019] FIG. 3 provides an illustration of a simple heuristic algorithm for the mixed integer nonlinear problem, according to some embodiments of the present invention;

[0020] FIG. 4 provides an illustration of a greedy algorithm that may be used to achieve a sub-optimal solution to integer problem discussed above with reference to FIG. 3, according to some embodiments of the present invention;

[0021] FIG. 5 provides an illustration of an iterative optimization framework which combines the heuristic approach and the power of commercial solvers to obtain sub-optimal solutions, according to some embodiments of the present invention; [0022] FIG. 6 provides an illustration of process for approximating the MINLP with MILPs to iteratively obtain a solution, according to some embodiments of the present invention;

[0023] FIG. 7 provides pseudocode illustrating an example dynamic programming algorithm for the chiller bank and TES optimization, according to some embodiments of the present invention.

[0024] FIG. 8 provides example pseudocode that may be used to generate optimal actions based on output tables, according to some embodiments of the present invention;

[0025] FIG. 9 provides an illustration of a method for determining optimal values for a CCHP system associated with a site according to some embodiments of the present invention;

[0026] FIG. 10 provides a slightly more detailed example of a method for determining optimal values for a CCHP system associated with a site, according to some embodiments of the present invention; and

[0027] FIG. 1 1 illustrates an example of a computing environment within which embodiments of the invention may be implemented.

DETAILED DESCRIPTION

[0028] The following disclosure describes the present invention according to several embodiments directed at statistically modeling and forecasting cooling and electricity load demand for Combined Cooling Heat and Power (CCHP) systems. In some embodiments, a model is implemented for forecasting cooling and electricity load demand for a given time window. The model is a generalized form of Cochrane-Orcutt estimation technique, which combines a multiple linear regression model and a seasonal autoregressive moving average (ARMA) model. The technology described herein may be generally applied, for example, to optimize the operation of CCHP systems in single building or multi -building applications.

[0029] FIG. 1 provides a schematic of a CCHP plant 100 that may be managed using one or more of the techniques described herein, according to some embodiments of the present invention. Such a plant may be used, for example, for powering a university campus. However, it should be understood that the present invention is not limited to site systems; rather it may be generally applied to manage CCHP systems associated with various sites (e.g., office buildings, resort areas, etc.). The CCHP plant 100 includes a set of electric chillers 105A providing cold water, a thermal energy storage (TES) tank 1 10, an onsite gas turbine (GT) 120, and a steam turbine (ST) 1 15. In this example, GT 120 is the primary source of electric power providing electricity for the site and for the chillers 105A, 105B. As a byproduct, GT 120 generates the exhaust gas, which can be a source of extra thermal energy. Such energy is then used to produce steam using a heat recovery steam generator (HRSG) unit 125. The generated steam drives the steam turbine (ST) 1 15. The steam can also be used to produce hot water for the site needs. A portion of the produced steam is also transferred to use in a steam chiller unit.

[0030] Continuing with reference to FIG. 1 , the electricity produced by the two generators 1 15, 120 and imported from the Power Grid 120A may be sent either directly to the site for satisfying a portion of electricity demand or supplied as the energy needed for electrical chillers, which are mainly responsible to provide cold water. Cold water can be either directly supplied to the site via an absorption chiller 105B to meet site cooling needs or, alternatively, it may be stored in the TES tank 1 10 for later use. Hence, the chillers 105A, 105B and the TES 1 10 together are the main sources to meet site cooling demands. Any additional electricity demand may be met through purchase from the grid. Such a CCHP system 100 is able to produce thermal energy (cooling) along with electricity over time. The TES 1 10 is a flexible component of the plant, which allows the site reshaping the cooling demand particularly in peak hours. A key element for such optimal control is to have accurate information about the energy (electricity and cooling) demand over the course of a day, which is one focus of the technology described herein.

[0031] FIG. 2 provides a high-level overview of a system 200 which may be used to optimize CCHP systems, such as those illustrated in FIG. 1), according to some embodiments of the present intention. The system includes various components which may be implemented using any combination of software and hardware. For example, in some embodiments, the various adapters 225A, 225B, and 225C shown in FIG. 2 each correspond to a software module. The system 200 employs the "operator-in-the-middle" control strategy, suggesting to the system operator the best set-points based on the control algorithms determined by the optimization system 225. The operator can decide whether to apply them or not. Long-term historical data 2). This long-term historical data 210 may include, for example, past power output of the gas and steam turbine, chillers schedules, power consumption and cooling output, imported electricity from the grid and other parameters. In some embodiments, the long-term historical data 210 may also include occupancy information associated with the site. The long-term historical data 210 is used to train the energy and cooling demand forecasting algorithms.

[0032] Continuing with reference to FIG. 2, a weather data source 220 provides a weather forecast for a predetermined period in the future (e.g., the next 24 to 36 hours) to a weather adapter 225B in the optimization system 225. For example, the weather adapter 225B may be configured to automatically retrieve weather data at regular interfaces (e.g., using a pull methodology with the weather data source 220) or weather data may be sent to the device at regular intervals (e.g., using a push methodology with the weather data source 220). In some embodiments, the weather data source 220 is a website operated by an entity such as the United States National Weather Service. The weather data may include, for example, values for ambient temperature, an outdoor humidity ratio, luminous efficacy of sky diffuse, etc.

[0033] In the Optimization System 225, a forecast adapter 225A runs forecasting algorithms and computes the cooling and electricity demands of the site. In some embodiments, the Cochrane-Orcutt technique is used to construct a time-series regression model for cooling and electricity load demand. In addition to the long-term historical data 210, the forecast adapter 225A may utilize short-term historical data 205 from a central plant historian database (not shown in FIG. 2). This data may include various information related to the CCHP system. For example, consider a site with eight chillers at a central plant providing cooling energy for the entire site. The short-term historical data 205 may include the following list of available hourly data: the power associated to each chiller (tons); the temperature of chilled water supplied to the site and denoted by T _CHSC ; the temperature of return water to chillers denoted by T _CHlVR ; and the chilled water site flow m _ckw . Additionally, values may be derived from the known data. For example, to find the total cooling power at time t the following equation may be used:

W _coolin Akw) = m _chw x 500 x (T _CHmi - T _CHSC ) / 3412.142 (1) With this equation, one can calculate the hourly cooling demand and employ it for forecasting purposes. Note that multiple models may be constructed based on the environmental characteristics. For example, if it is observed that the average cooling demands for weekdays are statistically identical and significantly different from the average cooling demand of weekends, one can build two separate models for weekends and weekdays and integrate the models into a single forecasting policy.

[0034] In addition, it is easy to calculate the site electricity using the same method. To do this, one should find the total power consumed by chillers as well as total hourly power supplied to the site. This includes all the power generated by the gas turbine and the steam turbine as well as the power imported from the grid. Therefore, the total hourly electricity needed for site can be calculated as follows: w _dec (0 = w _Rrid ( + w _GT ( + w _ST (k) - w _CUUers (0, (2)

Where W _gri d(t), Woi(t) and Ws ) are the total power supplied by grid, gas turbine and steam turbine, respectively. Also, Wchni _e r _s it) is the total power consumed by all the available chillers at time t and can be given by:

⁸ (3)

W _ChiUers (t) =∑ (w _h ( x 0.293 x 12000 /(1000 x cop , )),

i=\

Where cop\ is the coefficient of performance for the f ^h chiller and w' _ch (t) is the power consumed by ¹ chiller at time t. The other constants are applied to covert tons into kw.

[0035] Furthermore, the heating load may be estimated in the same way cooling demand was calculated. The site hot water supply and return temperature are available and denoted by T _HTHWRT ^an d T _HTHWST · I ⁿ addition, the hot water site flow can be integrated by adding up hot water flows from Gas/Steam turbines (mmn and mma)- Therefore, total heating load consumed at time t can be estimated as follows: = (m _IBn +m _HX2 ) x 500x (T _HTHWST -T _HTHWRT )/34llU2 (4) [0036] Returning to FIG. 2, the forecast adapter 225A passes the forecasted values to an optimization adapter 225C that executes optimization algorithms and output set-points. These algorithms, and the optimization process generally, are described below. Once the set-point values are optimized, a network driver 225D (e.g., a TCP/IP driver) passes the set point values to the human-machine interface 215 (HMI) (e.g., computer, smart phone, specialized display device) for display to the operator. The operator observes the set-points and decides whether to set them or not to configure the CCHP system (see FIG. 1). It should be noted that presentation to the operator is not necessarily required. For example, in some embodiments, the system is further configured to automatically apply the optimal set-points to the CCHP system.

[0037] The objective of the optimization is to meet both the cooling and electricity demands, while keeping the operation cost minimized. Inputs may include, for example, demand forecast, rate information from utilities, and the current system status (TES), while outputs may include optimal set-points for components such as the gas turbine, steam turbine, electric chillers/TES, as well as electricity purchase recommendations. In some embodiments, the optimization problem corresponding to the "master controller" is re-formulated into a mixed- integer-nonlinear-optimization-problem (MINLP) as follows. The objective function is expressed as the summation of grid electricity cost and fuel cost in the 24-hours look-ahead period:

In addition to the dynamic constraints associated to TES, and various equality constraints derived from component models, the following constraints on demand and capacity also need to be enforced:

(6) Grid power purchased GT power produced ST power produced

iooo

esMus -sower demand

Chiller &β.ίϊ.& poiver csnsu pfton

[0038] Additionally, the optimization problem includes several unknowns (i.e., decision variables) that need to be decided at regular intervals. For example, in some embodiments, for each hour of the 24-hour look-ahead period, the following variables needs to be decided: ON/OFF states for each chiller; Charging/discharging mode of TES; Total chilled water supply to site; Total chilled water supply through TES; Operating power level of gas turbine; power purchased from grid; and temperature of returned water from site.

[0039] Various algorithms may be used for solving the aforementioned optimization problem, with each algorithm having different computation cost and performance. In the sections below three algorithms are discussed: a heuristic algorithm (see FIGS. 3 and 4); a MILP (mixed integer linear programming) based algorithm (see FIGS. 5 and 6); and a dynamic programming-based algorithm (see FIGS. 7 and 8).

[0040] FIG. 3 provides an illustration of a simple heuristic algorithm 300 for the mixed integer nonlinear problem according to some embodiments of the present invention. The idea is to treat the amount of cooling stored in TES as a resource, and design a greedy algorithm to allocate it to different time slots to meet the cooling demand and all other constraints. This algorithm constitutes the following steps:

[0041] At step 305, a charging and discharge plan is determined. The objective of the original optimization is to meet the cooling and electricity demand of the site while keeping the cost as low as possible. TES is the key component for shifting the cooling\ electricity demand away from peak hours. Intuitively we should charge TES during the night when electricity is cheap and discharge TES in the rest of day. In some embodiments, the charging hours are fixed (e.g., the first 9 hours of a day, up to 9:00 am, and discharging from 10:00 am to midnight). In other embodiments, different or dynamic charging\discharging time zones may be applied. We also set a target status of TES at end of the charging period, in terms of Ta, and a final status of TES end of the discharge period, in terms of Tb. When Ta is as low as the target value, the charging of TES is considered as finished. When Tb is as high as the final value, the TES is considered depleted. Target Ta is one of the control variables to be optimized. In some embodiments, it is manually set and tunable to find the best value in terms of lowest total cost. Tb may be decided by maximum flow rate through site and maximum flow rate through TES. This is because when the outlet temperature of TES (equal to Tb in our model) is too close to some temperatures (e.g., 60F), it cannot be easily utilized for site cooling.

[0042] The original optimization problem is now divided into two sub-problems: when TES operates in charging mode and when TES operates in discharging mode. At 310, the optimal set- points of all components during charging mode are determined. For example, if the charging time period is 12 midnight to 9am, the objective is to find optimal set-points of all components from 12 midnight to 9am, such that TES at 9am reaches the target status, the site cooling demand and electricity demand are both met, while the total monetary cost is minimized. At 315, the optimal set-points of all components for the time -period while the TES operates in discharge mode are determined. Continuing with the previous example time periods, the objective is to find optimal set-points of all components from 10am to 12 midnight, such that the TES at 12 midnight reaches the final status (meaning that the stored cooling capacity in TES is fully utilized), the site cooling demand and electricity demand are both met, while the monetary total cost is minimized.

[0043] At 320, the cooling optimization and electricity consumption optimization are separated. For each of the sub-problems described above with respect to steps 310 and 315, the optimization for hydraulic system/chillers/TES may be separated from the gas turbine/ co- generation plant. Now it becomes a two-phase optimization problem. In the first phase, the chiller operating schedule and TES set-points are determined such that all constraints are satisfied and the chiller electricity consumption (actually the electricity cost evaluated using electricity price from grid) is minimized. In the second phase, given the chiller electricity consumption and site electricity demand, the GT operating schedule W GT and purchase plan from grid W _gr i _d that is optimal (lowest monetary cost) is determined. This two-phase optimization will lead to sub-optimal solution, but it reduces the complexity of the problem. [0044] At 325, the complexity of the integer program is reduced. The aforementioned second phase for both sub-problems is not difficult since there are no dynamics involved. So we focus on first phase - deciding the chiller operating schedule and TES set-points. This is still a complex nonlinear mixed integer program. The complexity of this problem can be reduced by evaluating all possible configurations of chiller operation. At each time step (e.g., one hour), there are 2 ^number °f ^chlUers possible operating conditions. Some of these configurations generates the same amount of chilled water flow (thus same amount of cooling) as other configurations, but consumes more electricity. Such configurations may never be considered as an option because there are more efficient alternatives to them. By eliminating those inefficient configurations, the number of possible operating conditions may be reduced.

[0045] FIG. 4 provides an illustration of a greedy algorithm 400 that may be used to achieve a sub-optimal solution to integer problem discussed above with reference to FIG. 3, according to some embodiments of the present invention. At step 405, an initial solution is constructed which generates the most electricity consumption. For example, in one embodiment all chillers are staged on for the first 9 hours. TES is in charging mode for this time period. Then, for every hour from 10am to midnight, a configuration of the chillers is selected such that the cooling they provide is maximized but less than the site cooling demand. TES is in discharging mode and provides the rest part of cooling to meet site demand. Additionally, system constraints are set at 405 which will constrain the iterative operation of the remaining steps of the algorithm. Various constraints may be used such as, for example, those described above with reference to Equations 6 and 7.

[0046] At step 410, for each iteration, an hour t between 10am and 12 midnight is selected and the amount of cooling provided by chillers is reduced by dQ during that hour. In some embodiments of the greedy algorithm, chilled water from TES may be used to provide this gap of dQ to meet site cooling demand. Next, at step 415, a number, referred to herein as an "efficiency score" is computed to help deciding t and dQ. For example, at hour t, let the current chiller configuration be cl and an alternative configuration be c2. The amount of cooling they provide are Ql and Q2, respectively, and the amount of electricity they consume are PI and P2, respectively. Then, the efficiency score S is the ratio between the changes in cooling and electricity consumption: S = dP/dQ = (P1-P2)/(Q1-Q2); (8)

[0047] Continuing with reference to FIG. 4, at step 420, a search among possible alternative configurations is carried out for each iteration, and the one corresponding to highest score may be selected. This score favors those chiller configurations that provide more cooling at the cost of less electricity. Note that S can be negative, meaning that some chiller configurations provide less cooling but consume more energy than current configuration. Next, at step 425, after the alternative configuration is decided, the TES model is simulated again to decide how much cooling capacity is left at midnight using new set-points. That way, this algorithm works like resource allocation, it allocates TES cooling to different time slots, in the place of chiller cooling, to reduce electricity cost, until TES is depleted.

[0048] The greedy algorithm 400 terminates when one or more of the constraints set at 405 are violated. For example, the TES is depleted at midnight. The sub-problem for TES charging mode can be solved similarly. The only difference is that when computing the efficiency score, the dQ is not the amount of cooling provided by chillers, but the amount of cooling that is charged to TES (per each second). So the status of TES (Ta) needs to be taken into account: dQ = cp*(Ta - 40)*dm, (9)

Here dm is the difference of flow rate between two different chiller configurations.

[0049] FIG. 5 provides an illustration of an iterative optimization framework 500 which combines the heuristic approach and the power of commercial solvers to obtain sub-optimal solutions, according to some embodiments of the present invention. The idea is to divide the original optimization problem into two parts: the optimization of chiller and TES operation, which is a MINLP problem, and the optimization for the onsite generation system and the grid import, which is a nonlinear program. The first problem is solved by approximating the MINLP 510 by a set of MILPs based on inputs 505 which may include, for example, model parameters, tariff data, and cooling demands. In the first problem, TES and chiller operations are optimized, while minimizing factors such as energy consumption, cooling energy balance, TES / Chiller Dynamics, and Capacity Constraints. This results in a measure of chiller energy consumption 515 which is used with additional inputs 520 in the second problem solved by a second MINLP 525. These additional inputs 520 may include, for example, model parameters, tariff data, and site electricity demands. In the second problem, GT & HRSG operations are optimized, while minimizing based on factors such as 24-hour cost, electricity energy balance, GT/HRSG characteristics, and capacity constraints.

[0050] FIG. 6, provides an illustration of process 600 for approximating the MINLP with MILPs to iteratively obtain a solution, according to some embodiments of the present invention. At 630, nominal values are initialized for linearization using the previous heuristics methods. Next, at step 615, the TES & Chiller dynamics are linearized around the nominal values to produce inputs 605. In addition to the linearized dynamics, the inputs 605 may include a fixed TES profile, cooling load data, and other model parameters. At step 610, a MILP problem is solved via optimization solvers to generate optimal set-points 625. These solvers may include conventional solvers such as, for example, SCIP, Lp solve, GLPK, CBC, etc. At step 620, the TES & Chiller simulation is executed using the optimal set-points. Then, step 615 is repeated and the nominal values for linearization are updated based on the simulation results. During step 615 the optimal solution is compared to the nominal values. If the MILP optimal solution is "close" to nominal values (e.g., within a predetermined threshold value), the process 600 stops. Otherwise, the process continues iterating through the various steps discussed above.

[0051] Another approach to solve the MINLP optimization problem is to discretize all state variables and dynamics in the components, which give rise to a fully discrete problem. For this problem the only dynamic component is still the TES tank. After discretization, the TES dynamics is described by a finite state machine instead of differential equations. The optimal operation of chiller banks, TES and onsite generation can be computed through dynamic programming. In some embodiments, several modeling assumptions may be applied to the DP algorithm to simply processing. For example, the following conditions may be assumed to be approximately satisfied:

Assumption 1: If the i ^th chiller is on, where i = 1, . . . , n _c , its mass flow rate m _{chw i} its provided cooling amount Q _ChW, i and its consumed electricity W _{chw i} are all constants.

Assumption 2: For the chiller bank, the chilled water supply temperature T _chws is a constant. For the load, the site chilled water return temperature T _camr is a constant. Assumption 3: The total cooling capacity of all chillers is greater than the peak cooling demand of the site.

[0052] At any time t, given optimization horizon N, desired terminal state s ^* (expressed by upper level TES temperature and lower level TES temperature in a two-layer TES model, or expressed by the thermo-cline level in a thermo-cline TES model), and predicted cooling demand Qdem _r the algorithm will generate the following three tables as outputs:

• U ^* (s, i), the optimal action at time i and state s, for all i = t, · · · t + N - 1 and s e S.

• NS(s, i), the next TES state obtained by applying U ^* (s, i) at time i and state s, for all i = t, • · · t+N-land s e S.

• V (s, i), the minimal cost it takes to bring the system from state s at time i to final state s ^* at time t + N, for all i = t, ... , t + N - 1 and s e S.

Here S is the set of allowable states for TES model. FIG. 7 provides pseudocode 700 illustrating an example dynamic Programming Algorithm for the chiller bank and TES optimization, according to some embodiments of the present invention.

[0053] After the output tables are obtained, optimal actions can be generated from them, as explained in the pseudo code 800 shown in FIG. 8. Here V ^opt , u ^opt , s ^opt , a ^opt are the minimized total cost, optimal control sequence, optimal TES state sequence, and optimal TES charging and discharging status, respectively. The dynamic programming algorithm can be further extended to include various constraints. For example, to accommodate the special requirements of chiller operation, a constraint may be added that no chiller should be turned on between certain hours every day. So during this time period, chillers can only be turned off.

[0054] FIGS. 9 and 10 provide examples of techniques that may be used to apply the optimization techniques discussed herein. FIG. 9 provides an illustration of a method 900 for determining optimal values for a CCHP system associated with a site (e.g., site) according to some embodiments of the present invention. At step 905, a time-series regression model for cooling and electricity load demand is constructed based on long-term historical data associated with the CCHP system. This long-term historical data may include, for example, past power output data, power consumption, and/or scheduling data corresponding to at least one of the devices included in the CCHP system. Additionally, the long-term historical data may include an indication of how much electricity was previously imported from the electrical grid into the CCHP system.

[0055] At step 910, the time-series regression model (e.g., a Cochrane-Orcutt procedure) is applied to weather forecast data and short-term historical data associated with the CCHP system to yield one or more predictions of cooling and electricity load demand. In some embodiments, the time-series regression model is further based on occupancy data associated with the site. This will allow population fluctuations to be addressed. For example, where the site is a university campus, seasonal fluctuations can be large and have a great effect on CCHP demand.

[0056] Next, at step 915, one or more optimal CCHP data settings (e.g., set-points) for devices included in the CCHP system are determined using the predictions of cooling and electricity load demand. Various optimization models may be used, depending on the data settings being generated. For example, in some embodiments two sets of optimal CCHP data settings are determined: a first set of optimal CCHP data settings corresponding to a chiller and a thermal energy storage included into the CCHP system, determined based on an approximation of a MINLP problem using a set of Mixed Integer Linear Programming problems; and a second set of optimal CCHP data settings corresponding to an onsite power generation system and a grid import control included into the CCHP system, determined based on a nonlinear program problem. The data setting determined at 915 may also include schedules or purchase plans. For example, in one embodiment, set-points corresponding to a chiller and a thermal energy storage included into the CCHP system and minimizing chiller electricity consumption value are determined, along an optimal gas turbine operating schedule and an optimal power grid purchase plan.

[0057] Once these settings have been determined they may be transmitted to a human- machine interface (e.g., computer, smart phone, etc.) for display to an operator of the CCHP system. Alternatively, the optimal CCHP data settings may be automatically applied to the one or more of the devices included in the CCHP system. Thus, the system can intelligently self- regulate itself to optimal conditions. Even when automatic operation is used, one or more messages may also be provided to the human-machine interface to allow the operator to override the automatic operation if it is producing undesirable results.

[0058] FIG. 10 provides a slightly more detailed example of a method 1000 for determining optimal values for a CCHP system associated with a site, according to some embodiments of the present invention. This example may be implemented, for example, using the system 200 shown in FIG. 2. At 1005, models are trained by the forecast adapter 225A. These models may include, for example, an energy demand forecasting model based on long-term historical energy demand data associated with the CCHP system, and a cooling demand forecasting model based on the long-term historical cooling demand data associated with the CCHP system. At steps 1010 and 1015, short-term historical data (e.g., retrieved from an external central plan database) and weather forecast data (e.g., retrieved using a weather adapter 225B), respectively, are acquired. This weather forecast data provides near-term predicted weather information for a predetermined number of hours from a present time. In some embodiments, the number of hours is set by an operator, while in other embodiments the number of hours maybe dynamically changed based on factors such as the fidelity of the models being used by the method 1000. At step 1020, the forecast adapter 225 A uses an energy demand forecasting model to determine forecasted energy values based on the weather forecast data and the short-term historical data. Prior to, after, or in parallel with step 1020, at step 1025, the forecast adapter 225A uses a cooling demand forecasting model to determine forecasted cooling values based on the weather forecast data and the short-term historical data. Then, at step 1030, the optimization adapter 225C applies an optimization process to the forecasted energy values and the forecasted cooling values to yield optimal CCHP data settings for devices included in the CCHP system.

[0059] FIG. 1 1 illustrates an example of a computing environment 1 100 within which embodiments of the invention may be implemented. Computing environment 1 100 may include computer system 1 1 10, which is one example of a computing system upon which embodiments of the invention may be implemented. Computers and computing environments, such as computer system 1 1 10 and computing environment 1 100, are known to those of skill in the art and thus are described briefly here. Such a computing environment may be used to implement features associated with technology described herein such as, for example, the software and hardware components illustrated in FIG. 2. [0060] As shown in FIG. 1 1 , the computer system 1 1 10 may include a communication mechanism such as a bus 1 121 or other communication mechanism for communicating information within the computer system 1 1 10. The computer system 1 1 10 further includes one or more processors 1 120 coupled with the bus 1 121 for processing the information. The processors 1 120 may include one or more CPUs, GPUs, or any other processor known in the art.

[0061] The computer system 1 1 10 also includes a system memory 1 130 coupled to the bus 1 121 for storing information and instructions to be executed by processors 1 120. The system memory 1 130 may include computer readable storage media in the form of volatile and/or nonvolatile memory, such as read only memory (ROM) 1 131 and/or random access memory (RAM) 1 132. The system memory RAM 1 132 may include other dynamic storage device(s) (e.g., dynamic RAM, static RAM, and synchronous DRAM). The system memory ROM 1 131 may include other static storage device(s) (e.g., programmable ROM, erasable PROM, and electrically erasable PROM). In addition, the system memory 1 130 may be used for storing temporary variables or other intermediate information during the execution of instructions by the processors 1 120. A basic input/output system 1 133 (BIOS) containing the basic routines that help to transfer information between elements within computer system 1 1 10, such as during start-up, may be stored in ROM 1 131. RAM 1 132 may contain data and/or program modules that are immediately accessible to and/or presently being operated on by the processors 1 120. System memory 1 130 may additionally include, for example, operating system 1 134, application programs 1 135, other program modules 1 136 and program data 1 137.

[0062] The computer system 1 1 10 also includes a disk controller 1 140 coupled to the bus 1 121 to control one or more storage devices for storing information and instructions, such as a magnetic hard disk 1 141 and a removable media drive 1 142 (e.g., floppy disk drive, compact disc drive, tape drive, and/or solid state drive). The storage devices may be added to the computer system 1 1 10 using an appropriate device interface (e.g., a small computer system interface (SCSI), integrated device electronics (IDE), Universal Serial Bus (USB), or Fire Wire).

[0063] The computer system 1 1 10 may also include a display controller 1 165 coupled to the bus 1 121 to control a monitor or display 1 166, such as a cathode ray tube (CRT) or liquid crystal display (LCD), for displaying information to a computer user. The computer system includes an input interface 1 160 and one or more input devices, such as a keyboard 1 162 and a pointing device 1 161 , for interacting with a computer user and providing information to the processor 1 120. The pointing device 1 161 , for example, may be a mouse, a trackball, or a pointing stick for communicating direction information and command selections to the processor 1 120 and for controlling cursor movement on the display 1 166. The display 1 166 may provide a touch screen interface which allows input to supplement or replace the communication of direction information and command selections by the pointing device 1 161.

[0064] The computer system 1 1 10 may perform a portion or all of the processing steps of embodiments of the invention in response to the processors 1 120 executing one or more sequences of one or more instructions contained in a memory, such as the system memory 1 130. Such instructions may be read into the system memory 1 130 from another computer readable medium, such as a hard disk 1 141 or a removable media drive 1 142. The hard disk 1 141 may contain one or more datastores and data files used by embodiments of the present invention. Datastore contents and data files may be encrypted to improve security. The processors 1 120 may also be employed in a multi-processing arrangement to execute one or more sequences of instructions contained in system memory 1 130. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions. Thus, embodiments are not limited to any specific combination of hardware circuitry and software.

[0065] As stated above, the computer system 1 1 10 may include at least one computer readable medium or memory for holding instructions programmed according embodiments of the invention and for containing data structures, tables, records, or other data described herein. The term "computer readable medium" as used herein refers to any medium that participates in providing instructions to the processor 1 120 for execution. A computer readable medium may take many forms including, but not limited to, non-volatile media, volatile media, and transmission media. Non-limiting examples of non-volatile media include optical disks, solid state drives, magnetic disks, and magneto-optical disks, such as hard disk 1 141 or removable media drive 1 142. Non-limiting examples of volatile media include dynamic memory, such as system memory 1 130. Non-limiting examples of transmission media include coaxial cables, copper wire, and fiber optics, including the wires that make up the bus 1 121. Transmission media may also take the form of acoustic or light waves, such as those generated during radio wave and infrared data communications.

[0066] The computing environment 1 100 may further include the computer system 1 1 10 operating in a networked environment using logical connections to one or more remote computers, such as remote computing device 1 180. Remote computing device 1 180 may be a personal computer (laptop or desktop), a mobile device, a server, a router, a network PC, a peer device or other common network node, and typically includes many or all of the elements described above relative to computer system 1 1 10. When used in a networking environment, computer system 1 1 10 may include modem 1 172 for establishing communications over a network 1 171 , such as the Internet. Modem 1 172 may be connected to bus 1 121 via user network interface 1 170, or via another appropriate mechanism.

[0067] Network 1 171 may be any network or system generally known in the art, including the Internet, an intranet, a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a direct connection or series of connections, a cellular telephone network, or any other network or medium capable of facilitating communication between computer system 1 1 10 and other computers (e.g., remote computing device 1 180). The network 1 171 may be wired, wireless or a combination thereof. Wired connections may be implemented using Ethernet, Universal Serial Bus (USB), RJ-1 1 or any other wired connection generally known in the art. Wireless connections may be implemented using Wi-Fi, WiMAX, and Bluetooth, infrared, cellular networks, satellite or any other wireless connection methodology generally known in the art. Additionally, several networks may work alone or in communication with each other to facilitate communication in the network 1 171.

[0068] The embodiments of the present disclosure may be implemented with any combination of hardware and software. In addition, the embodiments of the present disclosure may be included in an article of manufacture (e.g., one or more computer program products) having, for example, computer-readable, non-transitory media. The media has embodied therein, for instance, computer readable program code for providing and facilitating the mechanisms of the embodiments of the present disclosure. The article of manufacture can be included as part of a computer system or sold separately. [0069] While various aspects and embodiments have been disclosed herein, other aspects and embodiments will be apparent to those skilled in the art. The various aspects and embodiments disclosed herein are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the following claims.

[0070] An executable application, as used herein, comprises code or machine readable instructions for conditioning the processor to implement predetermined functions, such as those of an operating system, a context data acquisition system or other information processing system, for example, in response to user command or input. An executable procedure is a segment of code or machine readable instruction, sub-routine, or other distinct section of code or portion of an executable application for performing one or more particular processes. These processes may include receiving input data and/or parameters, performing operations on received input data and/or performing functions in response to received input parameters, and providing resulting output data and/or parameters.

[0071] The functions and process steps herein may be performed automatically, wholly or partially in response to user command. An activity (including a step) performed automatically is performed in response to one or more executable instructions or device operation without user direct initiation of the activity.

[0072] The system and processes of the figures are not exclusive. Other systems, processes and menus may be derived in accordance with the principles of the invention to accomplish the same objectives. Although this invention has been described with reference to particular embodiments, it is to be understood that the embodiments and variations shown and described herein are for illustration purposes only. Modifications to the current design may be implemented by those skilled in the art, without departing from the scope of the invention. As described herein, the various systems, subsystems, agents, managers and processes can be implemented using hardware components, software components, and/or combinations thereof. No claim element herein is to be construed under the provisions of 35 U.S.C. 1 12, sixth paragraph, unless the element is expressly recited using the phrase "means for."