To effectively harvest energy from a piezoelectric transducer, we need to move beyond a simple black-box view and model its behavior. The most widely accepted tool for this is the Butterworth-Van Dyke (BVD) model, an equivalent electrical circuit that represents the transducer's electromechanical properties near its resonant frequency.

This model has two main parts:

1. The Motional Arm: An RLC series circuit ($R_m$, $L_m$, $C_m$) that represents the mechanical properties of the transducer. $L_m$ is the equivalent mass, $C_m$ is the equivalent compliance (the inverse of stiffness), and $R_m$ represents the mechanical damping and energy losses.
2. The Static Capacitance: A parallel capacitor, $C_0$, represents the inherent capacitance of the piezoelectric material itself, as if it were a standard parallel-plate capacitor. This is often called the parasitic capacitance because it doesn't contribute to energy conversion and can create challenges for the harvesting circuit.

The magic of energy harvesting happens at resonance. The BVD model reveals two key frequencies. The series resonant frequency ($f_s$) is when the motional arm behaves purely resistively, as the reactances of $L_m$ and $C_m$ cancel each other out. This is the point of minimum impedance and maximum current.

The parallel resonant frequency ($f_p$), or anti-resonant frequency, occurs slightly higher. Here, the motional arm becomes inductive and resonates with the static capacitance $C_0$, leading to maximum impedance.

For most vibration harvesters, operating at or very near the series resonant frequency ($f_s$) is the goal, as this is where the mechanical-to-electrical energy conversion is most efficient.

The sharpness of this resonance is defined by the Quality Factor (Q), which measures how underdamped the system is. A high Q factor means low damping ($R_m$) and a very narrow, sharp resonant peak. This allows for high energy conversion, but only if the ambient vibration frequency perfectly matches the transducer's resonant frequency. A low Q factor means a wider, flatter response curve, which is more robust to slight frequency mismatches but less efficient at its peak.

Not all mechanical energy is converted to electrical energy. The effectiveness of this conversion is captured by the electromechanical coupling coefficient ($k^2$). This dimensionless figure of merit tells us what fraction of the stored mechanical energy can be converted into usable electrical energy per cycle. A higher $k^2$ value means a more efficient transducer. This coefficient directly relates the resonant and anti-resonant frequencies.

With a complete BVD model, we can treat the transducer as an electrical source to design the rest of our energy harvesting circuit. For analysis, it's often simplified into a Thévenin or Norton equivalent circuit.

• A Thévenin equivalent models the transducer as an ideal voltage source ($V_{th}$) in series with an impedance ($Z_{th}$).
• A Norton equivalent models it as an ideal current source ($I_{no}$) in parallel with an impedance ($Z_{no}$).

According to the maximum power transfer theorem, to extract the most power, the load impedance ($Z_L$) connected to the transducer should be the complex conjugate of the transducer's source impedance ($Z_{th}$). For a purely resistive load ($R_L$), the optimal value at resonance is equal to the source resistance.

Finding this optimal load is the critical first step in designing the interface electronics, like a rectifier or a DC-DC converter, that will ultimately power the microgrid.

By understanding and applying these models, we can precisely predict how a transducer will perform and design systems that efficiently capture and convert ambient vibrations into useful electrical power.