How to Increase Energy Project NPV by 230% with Sophisticated Modeling: A Multi-Technology Case with NEM 3.0

As organizations look to design multi-technology distributed energy projects, what often is overlooked is how quickly the complexity can grow by adding just one additional technology such as a battery or EV charger on top of solar. If you're looking to model the expansion, replacement, or addition of solar, batteries, and possibly other Distributed Energy Resources (DERs) such as generators or Electric Vehicles, then understanding how to navigate and reduce this complexity is crucial to project success.

This blog will provide a foundational understanding of how and why this complexity occurs. And will show you how taking a sophisticated (not complicated) approach to modeling your energy project is critical, and can help you to achieve significantly higher NPV (Net Present Value).  In order to illustrate the best approach, we'll use an example case with solar and battery in a scenario that includes NEM 3.0. However, other examples that increase the complexity are Ancillary service (AS) market participation e.g. in PJM territory or projects that are built for resiliency via CHP, backup generation or batteries.

Since our example scenario is in California, below is some quick background on NEM 3.0 and the California Microgrid Incentive Program (MIP).  You can skip straight to the next section, Solar + Battery Adds Complexity, Especially During Project Design, if those programs aren't relevant to you or your projects.

Solar + Battery Adds Complexity, Especially During Project Design

Through our project experience and the projects of our clients, it is becoming clear that adding energy storage and other technologies to solar projects is important for viability. It’s also critical to recognize that this introduces multiple challenges, ranging from technical considerations to regulatory hurdles. Some of those challenges include:

• Batteries must be carefully matched with solar panels and integrated with inverters, energy management systems, and the grid. This requires precise engineering to ensure the system functions efficiently under varying energy production and consumption conditions. 

• Storage systems rely on smart software to manage when to store or release energy. These systems must analyze factors like weather forecasts, demand patterns, and utility rates to optimize performance, adding a layer of operational sophistication.

• Incorporating batteries requires specialized knowledge to manage the physical setup, wiring, and cooling systems. Compliance with electrical codes and best practices for storage technology increases the complexity of installation processes.

• Batteries substantially increase the upfront investment in solar projects, making financing more difficult. Justifying these costs often requires a detailed understanding of available incentives, projected savings, and system payback periods.

• And of course, in California, rules governing battery storage, such as interconnection standards and NEM 3.0 export compensation rates must be factored into the feasibility analysis and design stages to ensure optimal configuration.

While these challenges require additional expertise and planning, energy storage also amplifies the benefits of solar systems by increasing flexibility, reliability, and overall value.

Similarly, adding technologies such as biogas generators, hydrogen fuel cells, wind turbines, combined heating and cooling (CHP), and other technologies can improve the resiliency, carbon reduction, and cost savings of these systems, but will also increase the complexity exponentially during the planning and design phases.

For example, if we only consider the variable of technology sizing, the number of possible combinations grows quickly as you move from a two-technology configuration (solar + battery) to a 3-technology configuration (solar + battery + CHP), and so on, rapidly increasing the complexity of the analysis (see figure 1). 

Figure 1: This decision tree showcases a simplified view of the myriad of different microgrid design configurations that are possible just from the selection of a few technologies and their size options.

In reality, each time step for all future years is also a variable that will impact technology sizing since the operational levels will depend on utility rates and other factors that change over time and dictate the technology sizes needed.  If we were to add this variable to the graphic in Figure 1, then the complexity would skyrocket.

How Advanced Modeling Techniques Make Solar + Storage Projects Easier

As illustrated in Figure 1, adding even one or two different options significantly increases the number of possible solutions to a problem. When modeling an energy system over a year of operation, the number of possible options is in the millions - not only are we considering a vast range of PV and/or storage sizes, we’re also considering multiple options for meeting system demand every hour in the year.

A concrete way to grasp the scale of an energy system modeling problem is to see how this translates to the number of decision points that need to be made. In a model that studies the design and operation of a PV and battery system over a 20-year period, the number of decisions are: 

20 years x 8760 hours x (sizes of PV) x (sizes of battery) x (operation of PV) x
(operation of battery) x (utility purchase amount)

All of these possible solutions include decisions on how the system demand can be met, how the PV output can be used, how the storage can be charged, how much energy is lost during charging and discharging, and so on. This amounts to hundreds of thousands of decisions on how to design and operate a system.

In addition to solving the design and operation of a system, the costs have to be calculated to define the financial aspects and identify the decision points that will yield the overall solution with the most savings. Utility purchases and variable maintenance costs are calculated at each hour, and CAPEX equations calculate the investment costs for each tech. Different incentives and revenue streams are calculated based on hourly operation (e.g. renewable energy credits, ancillary market participation, demand response) and tech capacities (e.g. tax credits, grants).

Clearly, designing a microgrid requires tools capable of taking into account all of these variables in order to generate an optimal solution. Energy system modeling tools fall into two categories: 

1. Simulation tools 
2. Gradient-based optimization tools

Simulation tools model energy systems as a known^[1] combination of technology capacities, with rule-based approaches employed to fix technology operation. Since a simulation requires knowing the technology capacities in advance, most simulation-based tools automate the process of testing a range of tech capacity combinations. Some tools prompt users to iterate through different final selections, showing how the total costs change if they adjust the capacity of one tech. Other tools identify and present the user with the least-cost solution of the combinations simulated. Depending on the range and granularity of combinations tested, simulation based tools may identify solutions that are close to the optimal solution - however, gaps between simulation points, coupled with static rule-based dispatch approaches, make it impossible to find reliable, least-cost solutions. 

Rule-based dispatch approaches can miss opportunities for cost reduction through smart operation of techs, as shown in the example below. Additionally, even though most simulation-based tools provide selections of dispatch rules to enforce, choosing one that makes the most sense for a given project (let alone specifying your own rules) is difficult for modelers who are new to designing microgrids.

Gradient-based optimization tools, such as Mixed Integer Linear Programs (MILPs) like Xendee, can guarantee a globally optimal solution for both design and operation of a multi-tech energy system, including sizing and dispatch. Users set up the model by defining the technology characteristics, the system demand requirements, and the utility rates and incentive programs.

Given those parameters, a MILP will optimize how much tech capacity to invest in and how to operate it, therefore users don’t have to test different combinations of technology investments, or construct dispatch rules for the techs to follow. This will reduce the modeling and design times considerably since the MILP will do the work for the modeler. Since MILPs optimize capacities and dispatch simultaneously, the impact that design and operation have on each other is accounted for in identifying a least-cost solution at significantly reduced modeling and design times.

Through the example below that involves NEM 3.0, we demonstrate the power of advanced optimization tools like Xendee.

Lowering Costs for a California Hotel with PV and Battery

In a case study modeling a California-based large hotel subject to a time-of-use tariff, PV and batteries are considered for minimizing costs. The peak period with the most expensive energy and demand charges is from 4 pm to 9 pm. To compare rule-based dispatch against optimized dispatch, two versions of the energy system are modeled and compared against a reference case^[2]. 

In the first version, rule-based dispatch (typically used in current design and modeling tools) for storage is emulated by setting charging and discharging times manually to line up with the modeled tariff’s off-peak and on-peak hours. The storage can charge during cheaper times and can dispatch during expensive times. This manual rule-based approach will fail most modeling tasks since it is impossible to capture all assumptions manually. 

In the second version, storage dispatch is completely optimized by a MILP, and is not subject to user defined scheduling.

Rule-Based Results

In the first set of results, using outdated rule-base approaches, an overall savings of 3.9% in total costs (including annualized equipment costs) is realized through investing in 195 kW of PV and 127 kWh of storage. Over a 20-year study period, the NPV of the project is $89k.

Table 1: Costs and Emissions for Rule-Based Dispatch Results

Table 2: Tech Investments for Rule-Based Dispatch Results (Technology sizes are mostly found by a manual process or semi-automated process)

Table 3: Financial metrics for Rule-Based Dispatch Results

Figure 2: Shown above is the dispatch for an August day. The on-peak period is 4 pm to 9 pm (shown above as H16-H20), so storage discharging time is set to those hours. Storage can charge the rest of the day.

Optimized Dispatch Results

When MILPs are used to optimize the project design (not only the amount of energy charged/discharged by the battery, but also the timing) the project financials improve significantly. Total savings increase from 3.9% to 13.2% through smart operation of additional PV and battery. Although investments increased to purchase 296 kW of PV and 545 kWh of battery (increasing tech capacity by 52% and 329%, respectively), the optimized dispatch reduces utility costs more effectively, yielding a NPV of $299k, a 236% increase in project viability compared to the basic rule-based concepts!

Table 4: Costs and Emissions for Optimized Dispatch Results

Table 5: Tech Investments for Optimized Dispatch Results.  The technology sizes are calculated automatically without the need of user interaction.

Table 6: Financial metrics for Optimized Dispatch Results

The much higher savings with more tech investment is due to optimal leverage of the storage. Given the load shape, the best hours to dispatch are not just during the expensive on-peak period from 4 pm to 9 pm (shown as H16-H20), but later in the evening, as well as during a morning load peak, after charging the storage by purchasing inexpensive energy in the hours leading up to 6 am.

Figure 3: As seen from the Xendee bar chart, the blue bars have flattened to 90kW (meaning less demand and demand charges from the utility.  And this can only be achieved with a larger battery investment, which constitutes higher net present value compared with the rule-based approach with a smaller assumed battery size.

Conclusions from the Case Study

The case study tested with Xendee only imposed restrictions on timing for charging and discharging - the amount of energy going into/out of the battery was optimized by the MILP. Therefore, the case study described above had more opportunities to optimize storage operation than typical rule-based approaches.

The impact of switching from a rule-based strategy to an optimized dispatch is significant, a 236% increase in NPV! The focus on avoiding high energy and demand charges during the on-peak period missed additional opportunities for reducing energy purchases and non-coincident demand charges. The result of imposing storage dispatch rules via outdated simulation approaches is a more expensive project with undersized technologies.

This case study also illustrates how operation influences sizing, and vice versa: the rule-based restrictions limited the functionality of storage, and therefore, a lower storage capacity was installed. They also led to less PV capacity since the restrictive storage discharging window limited the usefulness of PV in charging up storage before the evening.

Overall Conclusion

When designing a renewable energy project, it’s critical to take into account multiple variables such as NEM 3.0’s net billing, types of technologies and their sizing, incentives, and even electricity dispatch schedules over time.  These types of projects become exponentially more complex as technologies and other variables are added, but gradient-based MILP tools like Xendee enable you to account for these variables and truly identify the optimal project solution for the short and longer-terms.

Authors:

• Michael Stadler
• Kelsey Fahy
• Trent Maw

Footnotes:

^{[1] Modeling decades into the future requires assumptions to be made around policy, new technological developments, and changing cost and materials markets. Knowing all the possible combinations 20 yrs into the future is not humanly possible unless you have an optimization software like Xendee to do this analysis.  Existing simulation tools for PV and battery projects cannot achieve that.}

^{[2] The reference case (or base case) is the starting point in the scenario without any distributed energy resources.  Each of the potential solutions are compared against this same reference case in order to measure improvement in desired outcomes.}